Navigation

Guidance and Navigation Systems CHAPTER 3: NAVIGATION
Navigation Problem Inertial Navigation Schemes Inertial Instruments Reference Systems and Models Navigation Equations and Error Sources Navigation Aids: Global Positioning System Integrated INS-KF, INS-GPS Systems Examples and Applications
3.1. The Navigation Problem The navigation problem answers two fundamental questions of motion and travel: what is my current position? where am I travelling to?
In navigation, a fundamental role is the precise determination of the current position and velocity in time. Such problem requires the definition of several reference systems that need to be used to relate the motion to, in order to compute motion components.
Figure 3.1: Navigation Problem
It is also important to note, that the variable time is not a dependent variable anymore, but it is an important independent variable, and reduction of all possible errors concurring to the evaluation of the current position is such that state estimation plays now a very critical role in the general navigation loop.
30/01/2004
Sistemi di guida e navigazione, a cura di Mario Innocenti
Navigation has acquired a new dimension in recent years, because of the capability associated to satellite-based technology (GPS), and expansion to low-cost mass motion vehicles such as automobiles. Early navigation occurred on land and sea. Even in the early times, the capacity of moving from one location to another was based on the availability of instrumentation, and the capability of reducing errors with respect to known objects. In the history of navigation, one great aid was the development of the magnetic compass. Although men had known of the magnetic properties of the lodestone for centuries before the Christian Era, the first use of the magnetic compass by navigators appears to have been in the 12th century. Navigators at this time also used the crossstaff and the astrolabe, two devices that the Greeks had invented to measure the altitudes of celestial bodies. From these measurements it was possible to determine the approximate latitude of the vessel. In the 17th century, Britain, France and other maritime countries actively began to aid the development of navigation. Astronomical observatories were established to provide almanacs. Mapmaking and the invention of required navigational instruments were also encouraged. In 1731 John Hadley, an Englishman, and Thomas Godfrey, an American, simultaneously invented a quadrant that made it possible to obtain accurate observations of celestial bodies. The instrument was similar to the sextant in common use today. Once latitude and longitude were established, the location was identified on a map or chart. During sea navigation, it was common practice to keep latitude information public, while longitude information was kept secret !!!
Figure 3.2: Latitude and Longitude Definitions
Independently of mechanization, the navigation solution always requires: Position: Velocity: Attitude - Latitude, longitude, altitude - X, Y, Z, components in Earth centered coordinate frame - North, East, Down - Roll, Pitch, Yaw - Quaternions - Cosine directors
30/01/2004
Figure 3.3: Example of referenced Position Variables
There are two main approaches to navigation, one that uses and refers to external aids, and a second, which is completely independent from them.
3.1.1. Dead Reckoning This is an example of navigation using external aids. Dead-Reckoning relies on the continuous updating of the position data derived from inputs of velocity components or speed and heading generated from a known start position. A simple example of this technique is a system, which uses a compass heading in combination with a device, such as an odometer, to measure the distance travelled over the ground. The accuracy of dead reckoning is largely influenced by the accuracy with which the initial position, velocity and heading are known. If the basic information comes from an AIR-DATA sensor, the measure velocity is:
n b n n = Cb Cw w + w
(1)
If the basic information comes from a RADAR-DOPPLER sensor, then:

n b s n = Cb Cs
(2)
From one of the above measurements we obtain latitude and longitude information:
north = Rmerid + h east = [Rnormal + h ] cos A possible implementation is shown in figure 3.4.
(3)
30/01/2004
Figure 3.4: Dead-Reckoning Mechanization
3.1.2. GPS Navigation With handheld receivers that cost less than $300, users can determine their position anywhere on the globe, provided they are outside and have a clear view of the sky. When combined with a computerized map, such receivers can pinpoint an individual or a vehicle's location on a given street within 30 feet. This revolution in navigation is the consequence of the development of a military satellite navigation system called the Global Positioning System/NAVSTAR, developed in the 1970s under the direction of Bradford W. Parkinson, now Stanford professor of aeronautics and astronautics. Commercial interest has grown, particularly with the spread of differential GPS. Differential GPS uses GPS receivers and satellites in conjunction with a ground station, or pseudo-satellite, at a known position, to provide high-precision tracking in specific locations.
3.1.3. Inertial Navigation Systems To a significant extent, inertial navigation is about coordinate frames. Inertial sensors measure rate information relative to an inertial frame of reference as shown in figure 3.5. An inertial coordinate frame does not rotate or accelerate with respect to any other system of reference. Accelerometers measure change of velocity with respect to an inertial frame. Gyroscopes measure change of rotation with respect to inertial space. In other words, they are inertial sensors. If one can relate measurements made in this inertial frame to some chosen navigation frame, then inertial sensors can be used to realize both 3dimensional coordinates (X, Y, Z) and 3-dimensional attitude (roll, pitch and yaw) in a world that makes sense to navigators. Inertial systems (accelerometers, gyroscopes, and computer) constitute a self-contained unit, with no relation with the outside world, and this is one of their main advantages. There are two implementations of the basic same principle: Stabilized platform and Strapdown platform.
4
30/01/2004
Figure 3.5: Inertial Navigation Structure
3.1.4. Navigation in Space Each "fix" on the orbit of an unpowered vehicle in free space (under the gravitational attraction of a single body, the Sun) requires six independent measurements of vehicle position or velocity. A number of combinations of quantities might conceivably be measured by optical or radar instruments to obtain the six independent measurements. Some possible examples are: Two sets of radar range measurements, separated by a sufficient interval of time, made on three celestial bodies moving in precisely known orbits around the Sun. Three photographs of a major planet against its background of stars, separated by sufficient intervals of time. Three radar range measurements and three radar velocity measurements made simultaneously, at a known instant of time, on three separate celestial bodies moving in precisely known orbits around the Sun. These measurements must be related to a known reference system, and this reference system must be maintained by some form of instrumentation on board the rocket. The best known way of implementing such a reference system is by a gyro stabilized platform. Over a long time, however, imperfections in the gyros will cause the platform and the reference to drift out of alignment. To remove these errors it is necessary periodically to correct the platform alignment by sightings on the fixed stars.
3.2. Inertial Navigation Schemes
Inertial navigation is implemented in two main ways as introduced earlier. The stabilized platform isolates the accelerometers from rotational motions of the vehicle and maintains the proper orientation of accelerometer axes. The computer operates mathematically on the accelerometer indications to determine the true position, velocity
5 30/01/2004 Sistemi di guida e navigazione, a cura di Mario Innocenti
of the vehicle, and steering commands for the control system. This solution finds its main application in space travel, the major advantages are: simpler gyros (platform rotates at small rates, lower dynamic range), high accuracy (North and East accelerometers do not see a component of gravity), self alignment by gyro compassing and sensor calibration by platform rotations. On the other hand some of the disadvantages are: complexity and cost, gimbal magnetics (torquers must not leak magnetic flux) and reliability (bearings and slip rings tend to wear).
Figure 3.6: Schematic Representation of Stabilized Inertial Platform
Figure 3.6 shows the general assembly of a stabilized platform, with sensors and state variables. The mechanization of the overall INU is shown in figure 3.7.
Figure 3.7: Mechanization of Stabilized Inertial Platform
Strap down platforms are characterized by components rigidly attached to the vehicle with benefits due to reduced size, cost and performance, as shown in figures 3.8 and 3.9. Among the advantages of Strapdown systems we have: simple structure, low cost; more rugged and lighter assembly; reliability (no gimbal magnetics, no slip rings, no
6
30/01/2004
bearings, electronics more reliable then machinery). There are of course disadvantages as: more difficult to align; more difficult to calibrate; motion induced errors which can only be partly compensated for; accelerometer errors (each accelerometer may feel 1 g from gravity) and requires a computer that can perform coordinate rotations in <.01 sec).
Figure 3.8: Schematic Representation of Strapdown Inertial Platform
Figure 3.9: Mechanization of Strapdown Inertial Platform
Improvement in navigation errors computation and reduction requires data estimation Kalman filters), and integration of inertial navigation output, with other schemes, namely radar and GPS systems. Synergy between inertial and GPS navigation is particularly effective due to the short term, long term characterization of the two methods. Prior to GPS, airborne radar ot other form of radar were used. A signal flow scheme of an inertial local level navigator is shown in figure 3.10 below.
7
30/01/2004
Figure 3.10: Signal Flow in Inertial Navigation
Example: Consider a simplified implementation of a 2D inertial navigation unit (Strapdown).
Figure 3.11: Basic Components of a Strapdown INU
The system contains two accelerometers and a single rate gyroscope, all of which are attached rigidly to the body of the vehicle. The sensitive axes of the accelerometers, indicated by the arrows directions are normal to each other and in the directions of motion (xb, zb) axes. The gyroscope is mounted with the sensitive axis orthogonal to both accelerometers, allowing it to detect rotations about an axis perpendicular to the plane of motion (yb). The navigation is assumed to occur with respect to an inertial system denoted by
8
30/01/2004
xi, yi, zi). Referring now to figure 3.11, body attitude is computed by integrating the measured angular rate yb with respect to time. This information is then used to resolve the measurements of specific force (fxb, fzb) into the reference frame. A gravity model, stored in the computer, is assumed to provide estimates of the gravity components in the reference frame. These quantities are combined with the specific force to determine true accelerations denoted by ( xi , zi ) . These are subsequently integrated twice to obtain estimates of vehicle velocity and position. The full set of equations needed to be solved is given by:
= ib fxi = fxb cos + fzb sin f = fxb sin + fzb cos zi xi = fxi + g xi = f + g zi zi zi xi = xi zi = zi
(4)
Now consider the specialization of (4) to the 2D case for the navigation in a rotating frame. The situation is shown in figure 3.12 for a vehicle moving in a meridian plane around the Earth.
Figure 3.12: Meridian View of a 2D Strapdown Navigation
We are concerned here with a vehicle operating in the vertical plane alone. Such a system would be required to provide estimates of velocity with respect to the Earth, position along the meridian, and possibly altitude. The mechanization given by (4) would work, but it would entail a further transformation of the velocity and position, derived in a space fixed coordinate, to a geographic frame. An alternative often used is to navigate directly with respect to the latter reference frame, defined in this case simply by the direction of the local vertical at the current location of the vehicle.
9
30/01/2004
It is therefore necessary to keep track of the navigation information with respect to the local geographic frame. This information can be extracted by differentiating successive gyroscopic measurements of body turn rate with respect to inertial space, and current estimate of the turn rate of the geographic frame with respect to the inertial space. The reference frames are depicted in figure 3.13 below.
Figure 3.13: Frames for 2D Navigation about the local Vertical
For a vehicle moving at a velocity x, in a single plane around a perfectly spherical Earth of radius R0, this rate is often called transport rate and is given by:
x R0 + h
The equations describing the navigation solution are now described by the set (5):
x = yb R + h 0 fx = fxb cos + fzb sin fz = fxb sin + fzb cos = f + x z x x R +h 0 2 x z = fz + g + R h 0 x = x = z z
(5)
If we compare equations (4) and (5) we can say: The attitude computation is modified to take into account of the turn rate of the local vertical as described above.
10
30/01/2004
The terms that appear in the velocity equations are included to take into consideration the additional forces acting as the system moves around the Earth (Coriolis terms). The gravity term appears only in the z-direction because of the assumption of spherical, uniform Earth.
3.3. Inertial Instruments
In the preceding sections we saw the basic components of an inertial navigation system. The overall package goes under different synonyms such as IMU, INU, INS, all with the same meaning. In both stabilized and Strapdown mechanizations, the three main elements are: accelerometer, gyroscope, and computer. This section will describe in some detail the physical functioning and mathematical modelling, leaving to later section a detailed analysis of the errors introduced. 3.3.1. Accelerometers The main element for measuring acceleration on board a navigation unit is the accelerometer. There are different implementations and different sensitivities allowing the measurement of higher order vibrational modes as well. A generic physical description, using linear modelling for potential and viscous damping, is shown in figure 3.14.
Figure 3.14: Physical Model of an Accelerometer
The basic principle consists of the measurement of the motion of a proof mass inside a case, when the case itself is subjected to an acceleration along the sensitive axis (input axis). The mass will tend to oppose the motion according to its inertia; when the steady
11
30/01/2004
state is reached, the force on the mass is balanced by the tension in the spring, and the net extension of the spring results proportional to the applied force, thus acceleration. Let us assume that the case has an instantaneous acceleration a, that is tilted up by an angle . Let y be the absolute acceleration of the proof mass, and define the relative motion of the mass with respect to the case by . The latter is the variable of interest, and its value is picked up by a variable electric circuit. Thus: a = x x y (6)
From the force free body diagram in figure 3.14, we can derive the equation of motion of the proof mass as: my + b + k mg sin = 0
(7)
Rearranging we have:
Suppose the inclination of the accelerometer is zero, the relationship between the vehicle (case) acceleration and the motion of the proof mass is trivial, leading to a second order transfer function:
m + b + k = mx (s ) 1 k G(s ) = x (s ) = mk ms 2 + bs + k
(8)
Typical high performance accelerometers have a break frequency acc = k m about four to five times higher than the frequency of the fastest signal measured. When an accelerometer is not oriented normal to the local gravity vector, the gravitational mass of the Earth will act on the proof mass stretching the spring and giving a non-zero acceleration signal as seen from (7) and repeated here:
m + b + k = mx + mg sin
(9)
12
30/01/2004
The acceleration bias due to gravity a bias = g sin must be then taken in proper account when extracting the vehicle true acceleration from the measurement. As mentioned before, the actual technological implementation of the basic principle takes many different forms depending on application, precision, dimensions, etc. The next figures describe some of the implementations. In many applications restrained pendulum accelerometers are used as shown in figure 3.15.
Figure 3.15: Schematic of a restrained Pendulum Accelerometer
Accelerometers of this type are capable of very high performance with good linearity, small biases, and high dynamic range (104-105 dimensionless acceleration per resolution). The dominant sources of error for this accelerometer are: fixed bias arising from residual spring torques; scale factor error caused by temperature effects; crossaxis coupling when g-loading is present; vibro-pendolous errors due to simultaneous acceleration and vibratory inputs; random bias due to assembly instabilities.
Figure 3.16: Pendulous fibre-optic Accelerometer

13
30/01/2004
Figure 3.16 shows the schematic of a 2-axis pendulumbased accelerometer with optical pick-off generated by a laser moving under the input acceleration. The performance of the accelerometer depend on several parameters such as: size of the proof mass, diameter of the optical fibre, length of the optical fibre, height of the suspension point, and size of the photo-sensitive array.
Figure 3.17: Vibrating Accelerometer 1
Figure 3.18: Vibrating Accelerometer 2
An accelerometer based on asymmetrically vibrating quartz beams is schematically shown in figure 3.17. Figure 3.18 shows another vibration-based accelerometer with a flat silicon proof mass cantilevered to the case. This solution is inexpensive, the material is elastic, and it has high strength and excellent electrical properties.
3.3.2. Gyroscopes
3.3.2.1. Brief History The earliest concepts for a gyroscope were put forth by G. C. Bohnenberger, a German inventor, in 1810. However, it was Jean Foucault who built the first functional gyroscope in 1851. Foucault, a French physicist, developed the instrument to demonstrate that the earth rotates on its axis. He named it from two Greek words, gyros meaning circle or
14
30/01/2004
ring, and skopein meaning to view, since the tool had allowed him to view the revolution of the earth. Hermann Anschutz-Kaempfe, a German manufacturer, developed a gyroscope in 1908 that could be carried on a submarine. At the time, submarines could barely tell where they were at sea, since traditional ship navigation techniques were unavailable under water. Anschutz-Kaempfes gyrocompass was itself problematic, since it had to be free to turn yet also able to withstand significant movements of a ship at sea. A continuous power supply was also necessary in order to keep the central disk rotating at a high speed. In the same year as Anschutz-Kaempfes first gyrocompass installation, an American named Elmer A. Sperry (figure 3.19) filed a patent for his own version of a gyrocompass that had significant improvement over the design of his German counterpart. His design allowed the axis of the rotor to move freely in a fore and aft direction, but it could not move side to side, as it was rigidly fixed to the ship. His first prototype was not completed until 1912, when it was installed in the USS Worden. It contained two gyro wheels, each weighing 4,000 pounds. The ship itself was a 433-ton torpedo-boat destroyer. The prototype used an electric motor to set the gyroscope precessing artificially just as the boat began to roll. Due to the massive weight of the gyroscope, its precessing forces would largely counteract the roll being experienced by the ship. As a result, the maximum roll experienced by the ship was reduced from 30 degrees to just 6.
Figure 3.19: Historical Review (taken by Sperry-Rand Corporation)
The gyroscope plays a fundamental role in navigation for two main reasons: a) it can be used for direct measurements of motion variables; b) it maintains inertial reference of the applied forces to the vehicle. Definition (Oxford Dictionary): An instrument designed to illustrate the dynamics of rotating bodies, and consisting essentially of a solid rotating wheel mounted in a ring, and having its axis free to turn in any direction.
15
30/01/2004
The two main physical principles behind the gyroscope are: Inertia: when the spinning portion of a gyroscope (called a rotor) is set in motion it will attempt to keep its axis of rotation continuously pointing in the same direction Precession: when a force is applied to a spinning rotor, it will attempt to compensate by rotating around an axis that is perpendicular to the applied force 3.3.2.2. Inertia The inertia principle is shown in figure 3.20. We can take a rotor and mount it in a set of gimbals (metal rings)., setting the rotor in motion, we can then hold on to an outer frame of the gyroscope and attempt to rotate the disc. Since inertia dictates that the rotating wheel desires to remain in the same plane of rotation, the rotor will cause the gimbals to rotate and will thus remain almost exactly where it started. The inertia or stiffness of a rotating disc is the gyroscopic principle used in recognizing an inertial direction in space.
Figure 3.20: Inertia Stiffness about Spin Axis
3.3.2.3. Precession Understanding the precession is perhaps a little less intuitive however it is equally if not more relevant since it is this property that is used in order to use a gyroscope as a sensor for angular motion in terms of applied torque to a vehicle. We start from a classical example. Imagine taking a bicycle wheel and dangling it from a string attached to one side of the center of its spokes. Then we lift some part of the wheel so that the entire wheel is now parallel with the string. Letting go of the wheel, it is quite obvious what would occur. Gravity would cause the upper portion to flop back toward ground, and the wheel would return to its original horizontal position.
16
30/01/2004
Figure 3.21: Stationary Wheel hanging from a Rope
The situation is depicted in figure 3.21. Now we spin the wheel as to rotate fast about the axle, the result is a vertical position rotating about the rope (precession). On a basic level, the physics of precession is fairly intuitive. To understand how precession works, we need to take a closer look at the forces acting on a gyroscope. Consider the three images in figure 3.22 below:
Figure 3.22: Precession Explanation
To continue with the bicycle wheel example, consider the first image a): the green arrows represent forces such as gravity pulling down and the string pulling up. As a result, the locations marked in red desire to travel in the directions marked a blue. This makes sense when thinking about the non-spinning wheel example, where the wheel simply flops over. If the wheel is set spinning before letting go, we saw that it remains vertical but turns around the hanging string. The second image b) demonstrates why this happens: the original points on the gyroscope that received left and right forces (marked as red) attempt to travel in those directions, but at the same time they are being spun. The last image c) shows the wheel as now spinning around a horizontal center line, while also rotating around a vertical axis (rather than simply flopping over) since each point receiving left and right forces is rapidly moving away from the point at which it received that force.
Figure 3.22: Relationship between Angular Momentum and Precession

17
30/01/2004
The main reason for precession is the induced variation of the total angular momentum of the spinning wheel in its vector form as shown above. A spinning wheel at constant angular velocity produces an angular momentum L given by:
L = I
From second Newtons law, the applied torque originating by the wheels weight mg satisfies the relationship:
d ( I ) dL = mgl = = = I = I dt dt inertial
(10)
Such torque changes the constant angular momentum from L to L+L, and the variation term L causes the precession about the rope as shown in figure 3.23.
Figure 3.23: Precession Torque Details
The principles of gyroscopes outlined above can be implemented in different ways, depending on the gyros degrees of freedom, or the number of gimbals used to hold the rotating mass In the simplest case, as seen before, the gyroscope resembles a spinning top and four point masses, A, B, C, D, represent the areas of the rim that are most important in visualizing how the workings. The bottom axis is held stationary but can pivot in all directions (see figure 3.24). When a tilting force is applied to the top axis, point A is sent in an upward direction and C goes in a downward direction (figure 3.24a). Since this gyro is rotating in a clockwise direction, point A will be where point B was when the gyro has rotated 90 degrees. The same goes for point C and D. Point A is still traveling in the upward direction when it is at the 90 degrees position in figure 3.24b, and point C will be traveling in the downward direction. The combined motion of A and C cause the axis to rotate in the "precession plane" to the right (figure 3.24b). This is called precession. A gyro's axis will move at a right angle to a rotating motion, in this case to the right. If the gyro were rotating counter clockwise, the axis would move in the precession plane to the left. If in the
18
30/01/2004
clockwise example the tilting force was a pull instead of a push, the precession would be to the left.
Figure 3.24: Another Look at the principle of Gyroscope
When the gyro has rotated another 90 degrees as in figure 3.24c, point C is where point A was when the tilting force was first applied. The downward motion of point C is now countered by the tilting force and the axis does not rotate in the "tilting force" plane. The more the tilting force pushes the axis, the more the rim on the other side pushes the axis back when the rim revolves around 180 degrees. Actually, the axis will rotate in the tilting force plane in this example. The axis will rotate because some of the energy in the upward and downward motion of A and C is used up in causing the axis to rotate in the precession plane. Then, when points A and C finally make it around to the opposite sides, the tilting force (being constant) is more than the upward and downward counter acting forces. The property of precession of a gyroscope is used to keep monorail trains straight up and down as it turns corners. A hydraulic cylinder pushes or pulls, as needed, on one axis of a heavy gyro. Sometimes precession is unwanted so two counter rotating gyros on the same axis are used. Also a gimbal can be used. The property of Precession represents a natural movement for rotating bodies, where the rotating body doesnt have a confined axis in any plane. A more interesting example of gyroscopic effect is when the axis is confined in one plane by a gimbal. Gyroscopes, when gimballed, only resist a tilting change in their axis. The axis does move a certain amount with a given force. Figure 3.25 shows a simplified gyro that is gimballed in a plane perpendicular to the tilting force. As the rim rotates through the gimballed plane all the energy transferred to the rim by the tilting force is mechanically stopped. The rim then rotates back into the tilting force plane where it will be accelerated once more. Each time the rim is accelerated the axis moves in an arc in the tilting force plane. There is no change in the RPM of the rim around the axis. The gyro is a device that causes a smooth transition of
19
30/01/2004
momentum from one plane to another plane, where the two planes intersect along the axis.
Figure 3.25: Gimballed Gyroscope
In figure 3.25, the precession plane in the gimballed gyro functions differently than in the example of figure 3.24, and has been renamed "stop the tilting force plane". The point masses at the rim are the only mass of the gyro system that is considered. The mass and gyroscope effect of the axis are ignored. At first consider only of the rim, the left half. The point masses inside the "stop the tilting force plane" share half their mass on either side of the plane, and add their combined, 1/4kg, mass to point mass A of 1/2kg. So then the total mass on the left side is the total mass of all 4 point masses, or 1kg. The tilting force will change the position of point mass B and D very little and change the position of point mass A the most. So we must use the average distance from the axis of all the mass on the lefthand side. The mass on the left side is 1kg. T he average distance the mass is from the "stop the tilting force" plane is 1/2 meter. Figure 3.25b shows a profile of the average mass in the tilting plane and the average distance from the axis that the mass is situated. We are concerned at how far the mass at the average distance will rotate within the tilting plane when a given force is applied to the axis in the direction indicated. Point mass A is rotating at 5 revolutions per second. This means that it is exposed to the tilting force for only 0.1 seconds. The tilting force of 1 Newton, if applied for 0.1 second, will cause the mass at the average distance to move .005 meter in an arc, in the tilting force plane. Since the length of the axis is twice as long as the average distance of the rims mass, the axis will move .01 meters in an arc. At the end of .1 seconds the point mass will be in the "stop the tilting force plane" and all the energy transferred to point mass A is lost in the physical restraint of the gimbal bearings. The same thing happens when point mass A is on the right side of figure 3.25. Only now, the tilting force will move point mass A down, and the axis will advance another .01meters. .01 meter every .1 second is not the whole story because the mass on the right side of the gyro hasnt been considered. The right side has the same mass as the
20
30/01/2004
left and has the same effect on the axis as the left side does. So the axis will advance half as much, half of .01 meter, or .005meters. Both halves of the rim mass will pass through the stop the tilting force plane 10 times in one second. Each time a half of the rim passes though the "stop the tilting force plane", it losses all its momentum that was added by the tilting force. The mass has to undergo acceleration again so we continually calculate the effect that 1 Newton has for .1 second on the rim mass at the average distance, 10 times a second. So then; at the point that the 1 Newton force is applied, the axis will move 5cm per second along an arc. The gyro will rotate at .48 RPM within the tilting force plane.
3.3.2.4. The Mathematics of the Gyro The mathematics of the gyroscope is derived from the mechanics of a spinning body.
Figure 3.26: Example of a Two-Axis Gimbaled Gyroscope
Figure 3.27: Free Body Diagrams of a Two-Axis Gimballed Gyro

21
30/01/2004
We refer to the previous figures, and consider the modeling of a two-axis gyroscope. The rotor is driven at a constant angular rate n with respect to inner gimbal g. The angle describes the angular position of the inner gimbal with respect to the outer gimbal o; and the angle defines the angular position of the outer gimbal with respect to the Earth (or with respect to the vehicle to which the gyro is mounted). The following assumptions are made in deriving the model: the rotor is perfectly balanced and symmetrical the bearings are rigid and perfectly centered a moment Mb is exerted about the Y axis between gimbals g and o a moment Me is exerted about the Z axis on the outer gimbal by the fixed base The Earth (or the vehicle) is considered to be inertial.
Referring to figure 3.26, the XYZ frame is considered the inertial frame; the frame is attached to the inner gimbal g, and the frame to the outer gimbal o. The angular velocities of the relative frames are given by:
= 1z = / + = 1y + 1z = = 1x ( sin ) + 1y + 1z ( cos )
(11)
()
The angular velocity of the rotor I then:

r r = r / + = = = 1x n + 1y + 1z =
(12)
= 1x ( n sin ) + 1y + 1z ( cos )
()
The equations of motion of the gyro are derived according to the free body diagrams in figure 3.27 and the application of Dalembert principle. The dynamic equilibrium equations in vector form for the system g+r from figure 3.27b are
( M )
*
= 0 = Mir + Mig
+ Mb
(13)
For the system g+r+o we have in vector form (see figure 3.27c):
( M )
*
= 0 = Mir + Mig + Mio
+ Mc
(14)
The inertial moments can be expressed in terms of angular momentum:
22
30/01/2004
M r = H r = H r + H r i i g g g g M i = H i = H + H Mio = H o = H r + H o i
(15)
Where the angular momentum expressions are:

H r = 1 J r ( n sin ) + 1 J r + 1 J r cos x x y y z z g g g g H = 1x J x ( sin ) + 1y J y + 1z J z cos o o H = 1z J z
( ) ( )
h J x J y J z
r Jx n r g = Jx + Jx r g = Jy + Jy r g = Jz + Jz
(16)
Rewriting (14) as:

0 = M ir + Mig
cos Mir + Mig
sin + Mio
( )
+ Mc
Using (15) and (16), yields:

Mir + Mig = J y + h cos + ( J z J x ) 2 sin cos y M r + M g cos = J cos2 sin cos h cos i i z z + ( J x J y ) sin cos Mir + Mig z sin = J x sin2 + sin cos + + ( J z J y ) sin cos
( ( (
) ) )
(17)
This becomes finally:

2 J y + h cos + ( J z J x ) sin cos = Mb o 2 2 J z + J z cos + J x sin h cos + 2 ( J x J z ) sin cos = Mc
(18)
In many applications of interest, it may be assumed that the angular motions will be very small (of the order of milliradians or arc seconds) and that n , therefore linearized analysis can be carried out. In this case, equations (18) become simply:
J y + h = Mb o J z + J z h = J Z h = Mc
(18)
23
30/01/2004
Because of the high quality construction of bearings and viscosity, we may assume:
Mb = b Mc = c
In many applications, the outer gimbal is fixed to the vehicle and only the inner gimbal is free to move. In this case, we have the so-called single-axis gyro:
J y = Mb h
Several arrangements exist, for instance depending upon Mb; a single axis rate gyro has then the linearized form:
Mb = k b + Mu J y + b + k = Mu h
(19)
Based on the above linearized analysis, we can make some comments on the motion characteristics of a gyroscope. To this end, consider a two-axis configuration described by (18) and repeated here with a little different nomenclature:
J y + h = M y JZ h = MZ
(20)
The axes y and Z are shown in figure 3.27b and 3.27c. System (20) can be solved using Laplace Transform methods yielding a natural motion with four poles:
s = 0,0, jh / J y JZ The two poles at the origin correspond to constant angular motion and , whereas the imaginary pair is an harmonic motion with natural frequency
= h J y JZ = Irx n J y JZ = 2n (21)
I where : J y = J Z = rx 2 Therefore the oscillation is roughly twice the spin speed n and it is undamped since the gyroscopic coupling terms do not produce energy dissipation. Consider the response to an impulse response My(t)=(t). The initial velocity produced by the moment can be computed to be
24
30/01/2004
yo
Jy
And the angular motions are given by

yo t = ( t ) [sin ()u(t )] d 0 yo J y t = t ( ) [(1 cos ())u(t )] d J Z 0
(22)
Figure 3.28 shows the time histories of the angular displacements. The motion is called coning and it is a property of the rotor spin axis. The coning motion is circular if the two moments of inertia in (22) are equal. This motion is small compared to the spinning rate of the gyro, and it becomes noticeable only at low rotor speed (similar to a top slowing down).
Figure 3.28: Coning Motion of a two-axis Gyro
The motions of precession (described earlier) and nutation (present in the case of a two-axis gyro) can also be derived from (20) by applying a constant torque Myo acting on the inner gimbal of its y axis. In this case the time histories of the angular motions can be computed solving (20) as: M yo t [(1 cos ())u(t )] d (t ) = J y 2 0 M yo t sin ( ) ( ) ( )u(t ) d t = h 0
(23)
The time histories of precession and nutation are shown in figure 3.29. The spin axis follows a cycloid path. The steady behavior of (t):
25
30/01/2004
Is the precession amount, and the up and down nodding motion superimposed to it is the nutation. The precession ss (t ) is steady about an axis normal to the axis about which the torque is applied, as we know already, and for a lowly damped case, a gyro may precess many times, nutating all the way. The results are shown in figure 3.30.
Figure 3.30: Precession and Nutation of a Two-Axis Gyro
The steady value of precession is also called gyro drift, and its amount dictates the actual gyro performance. ( t ) = M yo h
Figure 3.31: Commercial Gyrocompass

26
30/01/2004
Depending on the use the tolerable amount of drift varies of course. For missiles and control antennas typical values are degrees per minute, for aircraft stabilization are degrees per hour, inertial navigation units must have drifts of the order of degrees per year. As in the case of accelerometers, there are many technological implementations for single axis and free gyros. Figure 3.31 shows a high quality gyrocompass with a price of about 15,000 USD.
Figure 3.32: Floating SingleAxis Gyro
A general view of a floating single axis gyro is shown in figure 3.32. The gimbal is sealed in a case; the float is suspended in a liquid is just sufficient to relieve the gimbal bearings of the rotor weight reducing friction by up to 1000 times.
Figure 3.33: Electrically suspended spherical Gyro

27
30/01/2004
A better implementation for reducing bearing friction is shown in figure 3.33. It employs a spherical rotor suspended electrically or magnetically in an extremely good vacuum. Support is effected by controlling the plate voltages with position-feedback signals, and the location of spin axis ( and ) is sensed optically. In recent years, technology advances in Strapdown inertial units have converged to two types of gyros: (RLG) Ring Laser Gyro (FOG) Fiber-optic Gyro
Ring Laser Gyro (RLG) The ring laser gyro was developed in the 60s and 70s for commercial, space and military guidance, navigation and control systems. The first commercial application came when Boeing chose Honeywell to supply inertial reference systems for its 767 and 757 aircraft in 1978. Since that time, the ring laser gyro has provided higher accuracy and reliability at lower cost than other mechanical and fiber optic gyro technologies. Laser gyros have a number of advantages over more conventional ones, such as: No moving parts Simple design less than 20 components Very rugged-insensitive to g and g*g vibrations Wide dynamic range Output inherently digital and TTL compatible Fast update rate-less than 50ms to measure a rotation of 0.5o/hour Long and reliable lifetime Low total cost A single DOF gyro is schematically shown in figure 3.34 in a triangular configuration (other configurations have four sides).
Figure 3.34: Schematic of a 3-side RLG
The gyro is made up of a laser, a closed path cavity, mirrors at each intermediate corner in the path, and an interferometer/photo detector.
28
30/01/2004
The operation of the gyro is based on optical and electronic phenomena, rather than mechanical. As a result, these gyros exhibit little of the error associated with acceleration. The optical principle used by these gyros is the so-called Sagnac effect shown in figure 3.35.
Figure 3.35: Sagnac experiment used in RLG
Two laser beams from the same source propagate in opposite directions: clockwise (cw) and counter clockwise (ccw). Three mirrors S with a beam splitter form a square enclosed in a circle of radius r. The effect can be described with geometrical optics, referring to the right scheme in figure 3.35. As a result of rotation , the effective path length traveled in the direction shown is shortened (assuming small angles) by:
l =
2 2 s = r 2 2
For a rotation . The time to propagate around the closed path in both directions is: (l + l ) + = c = (l l ) c The difference in time to propagate in the two directions is therefore:
= + = 2
l r = 2 c c
If we define the rotation period above as the time to transient around the closed path without rotation, and we relate the time difference in wave-length change, we have:
29
30/01/2004
= 2
r c
From the geometry in figure 3.35, the enclosed area A formed by the path and its perimeter P are given by: A = 2r 2 ; P = 4 2r Therefore:
= 2 r 4A = c Pc /
Recalling that the frequency is related to the wave-length by =c, we obtain finally:
= and = 4A Pc 4 A = c Pc (24)
Equation (24) shows that the frequency change due to the gyros rotation is more sensitive with the higher ratio of the area to the perimeter. This equation holds for other RLG configurations as well as circular fiber-optic gyroscopes. Two commercial RLG are shown in figure 3.36.
Figure 3.36: RLG Configurations (courtesy of Honeywell)
Fiber-Optic Gyro (FOG) Another application of laser technology is implemented in FOGs. A scheme of a single DOF FOG is shown in figure 3.37.
30
30/01/2004
Figure 3.37: Schematics of a FOG.
The gyros mechanization also uses Sagnacs effect. The interferometric output of phase shift can be determined from (24). The relationship is given by: = 2 NL c (25)
Where N = number of windings around the spool, L=length of optical fiber. Substituting (24) in (25) we have that the phase change due to rotation is more sensitive with longer length fiber.
D 2 2 4 4 2LD = NL = c DN c
(26)
Figure 3.28: Example of FOG as Gyrocompass (courtesy of Sperry Rand Corp.)

31
30/01/2004
An application of the FOG technology is shown in figure 3.28, showing a fiber-optic gyrocompass used in marine applications.
3.4. Reference Systems and Models
In inertial navigation, the motion of a point is initially referred to an inertial system. Such reference however may not be the final reference frame of use for the navigator, nor the easiest to use in terms of implementation. Varieties of coordinate systems are used, and coordinate transformations take place within or outside the INU, in order to represent position, velocity and orientation in the most convenient way for the end-user. 3.4.1. Coordinate Transformations A coordinate transformation is a relationship between components of a vector expressed in two different reference frames. In a 3D environment, they usually consist in 3x3 square matrices. The main approaches to coordinate transformation are the direction cosines and Quaternions. 3.4.1.1. Direction Cosines Direction cosines use a 3-angle sequential rotation for relating two different reference frames. The most common direction cosines matrix (DCM) is certainly the one based on the standard Euler angles triad for a right-handed transformation:
3 = Yaw Angle 2 = Pitch Angle = Roll Angle 1
(27)
The basic transformations deal with a single rotation 3-2-1 according to figure 3.29.
Figure 3.29: Director Cosine Transformation Sequence
And they are given in (28):

32
30/01/2004
T (a) 1
BA
0 1 0 ca BA = 0 ca sa , T (a) 2 = 0 0 sa ca sa ca BA T (a) 3 = sa 0
0 sa 1 0 0 ca
(28)
sa 0 ca 0 0 1
The general transformation is therefore a linear combination of the three above as:
T 321 = T ( ) 1 T ( ) 2 T ( ) 3
And:
BG
B2
21
1G
(29)
c c cs + c ss ss + c c s s c + s c s [T ]321 = cs c c + s ss s s c c c
BG
(30)
(30) is shown schematically in figure 3.30.
Figure 3.30: Euler Angles
In the case of time varying Euler angles, it is of interest to compute the time variation of (30) or one of the 1-rotation DCMs in (29). We refer to figure 3.31 for the time rate of change of yaw angle. Recall the DCM relative to the angle from reference X to reference Y:
33
30/01/2004
[T ( )]3
BG
s 0 c = s c 0 0 0 1
(31)
Figure 3.31: Geometry of Time Rate of Change of Yaw Angle Transformation
The effect of a variation in a time t is shown in figure 3.31. In analytical form we have:
y (t ) = T ((t ))x (t ) = T (t )x (t ) y (t + t ) = [I ]T (t )x (t ) = T (t + t )x (t ) with :
(32)
[I ] =
Then the time rate of change of T ((t )) is defined as:

T [ (t )] = lim T (t + t ) T (t ) T (t ) T (t ) T (t ) = lim t 0 t 0 t t T (t ) = lim t 0 t
(33)
In compact form:
y Txy (t ) = y x / yTx (t )
(34)
Where is a skew-symmetric rotation matrix and the subscripts and superscript are interpreted as a rotation of the Y frame relative to the X frame in Y frame coordinates.
34
30/01/2004
0 y x/y = 3
3 0
(35)
The transpose of (34) is: Tyx (t ) = Tyx (t ) y x/y And the general result for a 3-axis transformation is given by: Tyx (t ) = Tyx (t ) y x/y
y x/y
0 = 3 2
3 0 1
2 1 0
(36)
Coordinate transformations based on Euler angles or similar triples suffer of singularities in certain situations. This problem is solved by the introduction of Quaternion representation.
3.4.1.2. Quaternions The Quaternions are members of a non commutative division algebra first invented by William Rowan Hamilton. The idea for Quaternions occurred to him while he was walking along the Royal Canal on his way to a meeting of the Irish Academy, and Hamilton was so pleased with his discovery that he scratched the fundamental formula of quaternion algebra, i 2 = j 2 = k 2 = ijk = 1 (37)
into the stone of the Brougham bridge. The set of Quaternions is denoted, H, and the Quaternions are a single example of a more general class of hyper-complex numbers discovered by Hamilton. While the Quaternions are not commutative, they are associative, and they form a group known as the quaternion group. By analogy with the complex numbers being representable as a sum of real and imaginary parts, a 1 + bi , a quaternion can also be written as a linear combination H = a 1 + bi + cj + dk Or represented by complex 2x2 matrices: z w a + ib H= = w z c + id c + id a ib (37)
The use of Quaternions in coordinate transformations is based on the following figure:

35
30/01/2004
Figure 3.32: Quaternions Coordinates
The direction cosine matrix in (30) can be written in terms of the orientation [, , ] of a unit vector and a rotation about this vector in figure 3.32.
Tyx (t ) = cos I + sin ( x ) + (1 cos ) T Using the relationship T = I + ( x )( x ) we have: Tyx (t ) = I + sin ( x ) + (1 cos ) ( x )( x ) Define the Quaternions from (27) and figure 3.32 in their standard notation as:
q0 = cos 2 q = sin = cos sin 1 1 2 2 q = sin = cos sin 2 2 2 2 q3 = 3 sin = cos sin 2 2 Where: cos2 + cos2 + cos2 = 1. With this and (39), (38) becomes
(38)
(38)
(39)
36
30/01/2004
2 2 2 2 With orthogonality condition q0 + q1 + q2 + q3 = 1 , and time rate of change given by:
q0 0 q 1 = 1 q2 2 q3 3
1 0 3 2
2 3 0 1
3 q0 2 q1 1 q2 0 q3
(40)
3.4.2. Reference Systems The main navigation reference frames for motion on, under and in proximity to the Earth are: Earth-centered Local level geodetic (geographic) Wander azimuth 3.4.2.1. Earth-Centered, Earth-Fixed Frame (ECEF) The ECEF frame is fixed within the earth, and it is centered at the earths center. The next 3 figures show different axis definitions. Figure 3.33 has the Z-axis parallel with the earths spin axis, figure 3.34 has the Y-axis parallel to the spin axis, and figure 3.35 has the X-axis in the direction of earths spin.
Figure 3.33: ECEF1, Z=Spin Axis, X=Greenwich Meridian, NED Local Geodetic
37
30/01/2004
Figure 3.32: ECEF2, Y=Spin Axis, Z=Greenwich Meridian, ENU Local Geodetic
Figure 3.35: ECEF3, X=Spin Axis, Y=Greenwich Meridian, NWU Local Geodetic
For each of the above frame, an inertial reference can be established as well (ECI) if needed.
3.4.2.2. Local Geodetic (Geographic) Reference Figures 3.33 3.35 also show a local reference system associated with the corresponding ECEF. They are labelled NED (North-East-Down), ENU (East-NorthUP), and NWU (North-West-Up) respectively. These navigation reference frames are non-inertial, and their relationship with the vehicles attitude requires the presence of a body-fixed system rigidly attached to the navigation system itself. For strap-down INU the body frame is usually identical to the navigation system inertial sensor triad. Figure 3.36 shows the body-fixed and navigation frames and their corresponding coordinate transformation matrix.
38
30/01/2004
Figure 3.36: Body Coordinate Frame and Transformation Matrix
3.4.2.3. Wander Azimuth Frame The Wander Azimuth (WA) frame is currently used in navigation, and has several definitions. The WA frame plays an important role for navigating at high latitudes without incurring in transformation singularities associated with North-based implementations, for instance. Figure 3.37 shows three different WA definitions depending on the local geodetic reference used. For example, in figure 3.37 1) the WA frame is rotated with respect to NED by an angle about the Z-axis. For the ENU and NWU frames, the Wander angle is an up rotation about the respective Z-Axes and it is shown in figure 3.37 2) and 3).
Figure 3.37: Wander Azimuth (WA) Frames for NED, ENU, and NWU
It will be shown later that the time rate of change of Wander angle satisfies [see 1) and 2), 3) respectively]:
= sin = sin
(*)
39
30/01/2004
The body frame WA frame relationship (NED case) is shown in figure 3.38, with the same yaw-pitch-roll sequence of figure 3.36.
Figure 3.38: Body Frame Wander Frame Relationship (NED)
3.4.2.4. Geographical Local Horizon LOS A line of sight can be defined on the local horizon plane, if there are sensors used for navigation or targeting such as cameras pointing a specific direction. Figure 3.39 shows the orientation of a LOS in terms of Azimuth (Az) and Elevation (El) angles relative to a NED geodetic frame.
Figure 3.39: Geographical Local Horizon LOS Reference Frame
3.4.2.5. Specific Coordinate Transformations The basic single-axis rotations given in (28) are used to go from one reference system to another with the appropriate rotation sequences. Using (28), we can compute the transformation matrix from ECEF to the local geodetic frame NWU as:
40
30/01/2004
0 c 0 s 1 0 T ( x ) = T1( ) T2 () = 0 c s 0 1 0 0 s c s 0 c
G E
(41)
If we want to use the WA frame, as navigation reference, then an additional type-3 rotation is necessary, yielding: c s 0 G T ( x ) = T ( x ) T ( x ) = s c 0 TE ( x ) 0 1 0 Note :
N E N G G E E N TN (x) = TE ( x ) T
(42)
In a similar fashion, we can derive the transformation matrix between the ECEF and NED via 3 rotations: Rotation 1: about Ze Rotation 2: - about Ye Rotation 3: alignment of axes from Up-East-Down to North-East-Down
s 0 0 0 1 c 0 s c T ( z ) = T3 (i ) T2 ( ) T3 ( ) = 0 1 0 0 1 0 s c 0 1 0 0 s 0 c 0 0 1 sc ss c = c 0 s c c c s s
G E
(43)
The transformation between earth centered inertial and local geodetic is obtained introducing earths angular displacement given by = I / E t .
sc ( + ) ss( + ) c G ( z ) TI E ( z ) = 0 TIG ( z ) = TE c ( + ) s ( + ) ( ) ( ) c c c s s + +
(44)
A relationship between body frame to navigation frame was obtained earlier, and shown in figure 3.36, for the earths spin axis defined along the Z ECEF direction. A similar transformation can be computed assuming earths spin axis along the X direction as shown in figure 3.38. This transformation requires four single rotations in a sequence: Rotation about xb Rotation about yb
41
30/01/2004
Rotation az about znb Rotation of about xnb.
The result is shown in figure 3.40.
Figure 3.40: Body Coordinate Frame and Transformation Matrix (X Spin Axis)
The transformation matrix from geodetic to LOS reference frame is computed with 2 simple rotations with reference to figure 3.39:
0 sEl cAz sAz 0 sElcAz sElsAz cEl cEl = T2 (El ) T3 ( Az ) = 0 1 0 sAz cAz 0 cAz 0 = sAz sEl 0 cEl 0 0 1 cElcAz cElsAz sEl
LOS G
The above expression can also be used to derive the relative (linearized) position r of an object at range , being tracked by a station with respect to the stations position:
sElcAz sElsAz cEl x 0 = sAz y cAz 0 0 cElcAz cElsAz sEl z
(45)
3.4.2.6. Rate of Change of Coordinate Frames Let us recall some facts from direction cosine matrix rates algebra. Consider Figure 3.3.1, and equation (36) repeated here for convenience: Tyx (t ) = Tyx (t ) y x/y
y x/y
0 = 3 2
3 0 1
2 1 0
(36)
42
30/01/2004
The above expression is obtained by expressing a cross product between two vectors as a matrix (skew-symmetric) vector product. Given two vectors a = [a1 a2 Then
0 a b = a3 a2 a3 0 a1 a2 0 a1 b = b3 0 b2 b3 0 b1 b2 b1 a = b a 0
a3 ] , b = [ b1 b2
T
b3 ]
With this, the skew-symmetric matrix in (36) is equivalent to the angular velocity vectors cross product y x/y . With reference to (35), we can establish the rate of change in earth-to-navigation transformation matrix. From (42), we can obtain (*) with the following procedure: recall that
N TEN ( x ) = N E / NTE ( x )
(46)
The navigation-to-earth angular rotation vector in navigation coordinates is given by:

x = y z
N E /N
(47)
Substitute in (45), use (42), and solve the second equation for yields: = c y + s x Substitute the above in the third equation:
= (s y c x ) c
(48)
(49)
Substituting (48) and (49) in the first equation, = sin = sin + z for z 0
(50)
43
30/01/2004
It is sometimes necessary to compute angular position and latitude/longitude wander angle errors. With reference to Figure 3.40, define the errors as follows (the overbar indicates the computed value):
= = =
For the angular position error vector we have then:

N TEN ( x ) = I ( x ) TE ( x ) 1 z y I ( x ) = z 1 x y x 1
(51)
And for the other errors:

= y c + x s =
x y
( c s )
c = z s
(52)
We can also express Tilt errors s in terms of Euler angles errors in body-navigation coordinates. Consider the relationship:
= [ ] = given
T
(52)
T
] = to be found
Tilt errors are expressed in terms of the computed and true body-to-navigation frame TBN ( x ) shown in Figure 3.40.
0 T (x) = I ( ) T I ( ) = z y
N B N B
z 0 x
y x 0
Using the above and following the procedure of equations (51)-(52), we have:
44
30/01/2004
= y c + x s = y s c c = y ( x c y s ) t
(52)
3.4.3. Earth Models Precise modeling of the Earth is a fundamental problem in the design of inertial navigation systems. The sources of errors come from the shape of the Earth, and the variation of gravity due to the non uniform density of Earth itself. The most widely used model today is the World Geodetic System (WGS-84). The main characteristics of WGS-84 are given in figure 3.41.
Figure 3.41: WGS-84 Ellipsoid Constants and Parameters
3.4.3.1 Ellipsoid Geometry
45
30/01/2004
Figure 3.42: Earth Shape Model Geometry
The Earths surface, from figure 3.42 is described by the equations for an ellipse: 2 z 2 2 + 2 =1 re rp 2 rp 2 r 2 = ( 1) e And also: z 2 = (1 ) tan 2 tan c = (1 ) tan
(53)
(54)
Using the variables in figure 3.43 relative to the Equatorial plane, we can compute the position coordinates in the ECEF frame for a point on the surface.
Figure 3.43: Earth Model in the Equatorial (X-Y) Plane
Combining (54) and (53) and from figure 3.43 we have:

46
30/01/2004
x = cos y = sin 2 z = re (1 )sin (1 2 sin2 )
(55)
Approximating the dependence on the altitude above the surface we finally obtain the following expressions for X-Y-Z coordinates:
r (1 2 ) e x ECEF = + h cos cos (1 2 sin2 ) 2 ECEF re (1 ) cos sin = + y h (1 2 sin2 ) ECEF re (1 2 ) sin = + z h (1 2 sin2 )
(56)
The radii of curvature (along the lines of constant longitude and latitude respectively) can be computer well. They are needed to determine the orientation changes of the navigation frame with change in position. The computation is carried out from the analytical definition of radius of curvature:
dz 2 1 + d re (1 2 ) = d 2z (1 2 sin2 ) 2 d
Rmeridian =
(57)
Rnormal =
re = cos (1 2 sin2 )
(58)
During the motion on the surface, the values of latitude and longitude vary with time with respect to the inertial frame, and with respect to the arc-length along meridians and parallels. In the latter case we can write therefore:
north = Rmeridian + h east = ( Rnormal + h ) cos
(59)
47
30/01/2004
Note: equation (59) is always valid except for poles operation, where a singularity exists. Position evolution in this instance must use other methods (DCM or Quaternions).
The longitude rate [second of (59)] is an angular rate relative to the ECEF frame about the Z-axis as shown in figure 3.33. We re-write (59) as time rate of change of the Earth geographic frame using an intermediate frame G transformed through the latitude . Thus:
c 0 s 0 0 1 0 = 0 0 + s 0 c 0
G' E /G '
After substitution we get: 0 0 1 G' = 0 1 0 E / G ' 1 0 0 east ( Rn + h ) = north Rm + h east tan ( Rn + h )
G E /G
(60)
If we write the Earths rotation in the above ECEF frame, and subsequently into the geographic frame, we obtain: 0 = 0 I / E I / E cos E 0 , and I / E = I / E sin
IE/ E
(61)
Finally, in vector form:
G I /G
east I / E cos + ( Rn + h ) = north Rm + h east I / E sin tan ( Rn + h )
(62)
3.4.3.2 Ellipsoid Gravity The mutual mass attraction due to Earth, or Gravitation, can be found as an approximated series of terms that include a first term representing spherical Earth, and
48
30/01/2004
successive terms approximating the oblateness. If we stop after the first two terms we have:
(63) Gravity is the acceleration on a rotating Earth; therefore (63) must be adjusted to account the centripetal acceleration due to Earths rotation. The gravity vector is therefore:
g = G I / E I / E r
(64)
A model of the gravity magnitude at the surface of the WGS-84 ellipsoid is given by: g = gWGSo (1 + gWGS1 sin2 ) (1 2 sin2 ) (65)
If the Earth were associated to a homogeneous ellipsoid, the local gravity direction would be normal to the ellipsoid. Gravity anomalies are present however and they are stored in the navigation unit, depending on the unity quality. Gravity deviations from the nominal value given by (64) and (65) are usually represented in vector form with respect to the local level geographic frame:
g (66) g= g g Where is the meridian deflection about the vertical; is the normal deflection of the vertical; g is the gravity magnitude from (65).
3.5. Navigation Equations and Error Sources
In this section we will derive the Navigation Equations for a Strap-Down inertial system. The Local Level mechanization frame will be used. The mechanization of a gimballed unit can be found in the references. The resulting differential equations are nonlinear and will be formed in terms of measured accelerations and angular rates, rather than standard Newtonian form. In order to integrate the INU with independent Kalman Filter data, a linearization of the equations is necessary, and error sources will be described. Note that linearization can
49
30/01/2004
be performed in different ways, depending upon the application and also on the industrial proprietary capabilities of INU makers. A general block diagram of a StrapDown INU is shown in figure 3.44. The core of the INU is a set of 3 gyros and 3 accelerometers mounted in orthogonal triads and rigidly attached to the vehicle. The gyros sense the motion IB/ B in body-fixed coordinates. The accelerometers output fB is transformed into the navigation frame by the
K , which is generated in the computer by the body rates navigation computer using CB coming from the gyros, and the navigation frame rates created by the vehicles velocity. In a gimballed INU, the sensors are mounted on a gimbal, whose alignment coincides with the navigation frame. Therefore accelerometers output fK are integrated to get velocity and position. As said before, each solution has advantages and disadvantages, however strapped down sensor information can not be used by a gimballed system due to the much higher level of measurement error.
Figure 3.44: Functional Diagram of a StrapDown INU
3.5.1. Local Level Mechanization Equations We will derive the translation equations first, and then the attitude equations. In the following sections the transformation matrices are indicated with a C(.) instead of a T(.). 3.5.1.1. Velocity Equation The velocity vector in the rotating navigation frame is given by the position rate of change in ECEF frame as:
N N CE rE
(67)
Taking the time derivative:

N N N CE rE + CE rE
(68)
50
30/01/2004
The position vector and its derivative, related to the inertial frame are:
r E = CIE r I E E I E I E I I I r = CI r + CI r = CI (r I / E r ) E E I with CI = CI I / E
(69)
Compute the second derivative of r E yields: r E = CIE (r I 2II / E r I + II / E II / E r I ) I I / E = [ 0 ] N N N CE = E / NCE
(70)
Note that in (70), the Earths rotation rate is assumed constant at one revolution per day. Substituting (68), (69), and (70) into (67) gives:
I I I E N I I I N = CIN r (E / N + 2I / E )CI I / E I / E r
(71)
The specific force is a combination of inertial and gravitational acceleration (gravity and centripetal contributions), the former sensed by the accelerometers and reduced in the navigation frame: f
N
= CIN r I G N
G N = g N + IN/ E IN/ E r N Using a similarity transformation:

I I IN/ E IN/ E = CIN IIIECN CIN IIIECN
(72)
From the above, the first of (70) becomes in inertial coordinates:

r I = CIN (f
N
+ g N ) + II / E II / E r I
(73)
Substituting (73) into (71) yields: N = f

N N N N N E / N + 2I / E +g
(74)
Equation (7) gives the basic description of velocity change in a local navigation frame. (74) is directly applicable to local-level geodetic and Wander frames (only if specifications in coordinates is needed), and also to either gimballed or StrapDown mechanizations.
51
30/01/2004
3.5.1.2. Attitude Equations The attitude equations are based on the gyros output. With reference to equation (36), the rate of change of the DCM matrix is given by:
N N CB = N B / N CB
(75)
N Where the skew symmetric matrix N B / N is formed from the rotation vector B / N as
N B/N
0 = 3 2
3
0 1
2 1 0
The rotation vector is obtained from the gyros measurements IB/ B as:
N N B N B / N = B / N CB I / B
(76)
Time evolution (solution) of (75) with (76) can be performed in several ways: Integration of 6 DCM elements and computation of the other 3 Integration of 4 Quaternions Integration of (76) directly
All techniques require Initialization.
3.5.1.3. Position Equations The position is obtained using DCM by solving the third of (70), that is:
N N CE = N E / N CE
(77)
Once initialized, (70) can be solved by direct integration as described above.
3.5.2. Perturbation Equations and Errors The nonlinear navigation equations are now linearized to obtain linear error models. Linearization is performed as perturbation on position, velocity, and attitude. The error system provides a useful way to study INS error propagation using linear methods and as the basis for designing Kalman filters to implement the various aiding techniques. Consider the navigation equations as a nonlinear system given by: x (t ) = f ( x , u ) (78)
52
30/01/2004
Define the error as: e=xINS -xSIM u=uINS -uSIM Then the error dynamics are given by: e (t ) = Ae+Bu Let us recall the navigation equations derived before:
N N N = f N N N E / N + 2I / E + g N N N CB = B / NCB N N C = N E / N CE E
(79)
(80)
(81)
The first step is the expression for the velocity error equation. Define:
a = computed a = real
The computed value has errors due to initialization, sensor measurements, biases, etc. For the first of (81), we can write: N = N + N Similarly we set: f N = I ( ) f N + f N N E / N = E / N + N E /N N N N I / E = I / E + I / E g N = g N + g N
N
(82)
(83)
Substituting (82) and (83) in the first of (81), subtracting the true value from the computed, yields: N = f
N
+ f
N N N + g N E / N + 2I / E
N N N E / N + 2I / E
(84)
53
30/01/2004
Equation (84) requires the expression for the rotation vectors errors (perturbations) expressed in terms of position and velocity perturbations, so that it can be solved in a closed form. Linearization here implies approximating the Earth radii with a constant nominal value R as done when we introduced the Wander Azimuth frame. From this, using the notation N E / N ; R h , we obtain equation (85). Remember that z represents the vehicle transport velocity about the vertical, which is zero for Wander Azimuth mechanizations [see equation (50)]. y x h R R = x y h R R z
N N N E / N = E / N E / N
(85)
As far as IN/ E = is concerned, its expression in vector cross product for is given by: IN/ E = IN/ E (86)
Substituting (85) and (86) in (84), the error equation in the Wander Azimuth mechanization is obtained. The next computation refers to the attitude error equation. This equation will be derived using the tilt error representation from (52), and (52). Recall the transformation from body to navigation frame:
N N CB (x) = I ( ) CB
(87)
The error component in the transformation matrix is given by:

N N N N CB = CB CB = ( ) CB
(88)
The rate of change of the error is then:

N N N N N CB = CB ( ) CB = CB + ( ) N B / NCB
( )
( )
Or:
N N CB = ( ) N B / N CB From (88) and (89) and using (81) we obtain
( )
(89)
54
30/01/2004
( ) =
And in vector form:
N B/N
( ) + ( )
N B/N
N N + B / N B / N
N N = N B / N + B / N B / N
(90)
In (90), we can introduce the gyro error through the measured angular velocity term:
N N B N N B / N = B / N CB I / B +
(91)
Equation (90) is then transformed into a vector equation where the tilt error is represented in a coordinate transformation form from inertial to navigation frame.
= + IN/ E + IN/ N +
(92)
The position error equation computation is the last one. Position errors are computed from the Earth to navigation transformation matrix:
N N CE = I ( ) CE
(93)
The procedure is similar to the other error equations. Define the error matrix:
N N N N CE = CE CE = ( ) CE
Taking the derivative yields

N CE = ( ) N E /N
( )
And, after some manipulations, in vector form we have:
= N E / N
Summary:
(94)
The entire set of linearized error equations can then be re-written summarizing the above results:
N = f
+ f
N N N + g N E / N + 2I / E
N N N E / N + 2I / E
= + IN/ E + IN/ N +
55
30/01/2004
= N E / N
For the case of zero transport rate, ( z = 0 ), we have:
Figure 3.45: Summary Error Equations
The error equations derived earlier used the Tilt angle formulation. We can however use an equivalent psi representation, which in some cases results in a reduced computational load, therefore attractive for onboard computation. In this alternate formulation, from Figure 3.40, the attitude error is defined as:
=
Taking the time derivative and using (93) we obtain:
(95)
= IN/ N +
Which is simpler than (92). 3.5.2.1. Alternate Error Equations Formulation Recall the velocity equation expression from (81) and (83):
N N N N = f N N E / N + 2I / E +g N 1 N = I ( ) +
(95)
(96)
The procedure for computing the velocity error equation is similar to that, which led to (84). Using z = 0 and (86) we obtain:
56
30/01/2004
1 = ( ) f + N
+ f
N 1 N + g N E / N + 2I / E
(97)
Using (95), and with g 1 = g 1 g N :

N N 1 1 = E / N + 2I / E + f N
+ f
+ g 1
(98)
The attitude error equation remains the same in this alternate formulation, that is:
= IN/ N +
The alternate position error equation is given by (94), and using (98) yields:
(99)
( 1 v y + z x x z ) + y2 h R R 1 + 0 ( v x z y y z) = x2 h + z R R y z
z 0 x
y x x y 0 z
(100)
Assuming Wander azimuth implementation ( z = 0 ) and using:
N E /N
y R = x R z
We obtain:
y z 1 1 x + 2 h y R R x R x 1 1 z y = R y R 2 h + R x h y x + x y + 1 z
A summary of the previous results is given by:
(101)
57
30/01/2004
Figure 3.46: Summary Alternate Error Equations
The above representation requires less error state than the equations in figure 3.45, which may be useful for computational purposes.
Figure 3.47: Error Equations in Wander Azimuth Reference
The alternate error description can be used to transform the error equations in a wander azimuth terrestrial navigation system (assuming z=0). The linear position error is related to the angular position error by the relationship:
58
30/01/2004
ry R r = x R z
Using equation (101), we obtain yet another representation as in figure 3.47. 3.5.2.2: Schuler Period The direction of the local vertical on the surface of the Earth can be determined using a simple pendulum consisting of a mass suspended by a string. If however the support is subjected to an acceleration a, the string will be deflected by an angle = arctan[a/g] and therefore no longer providing correct information. This could be solved if we were to design a pendulum having the string of length equal to the radius of the Earth; this hypothetical pendulum is called Schuler pendulum from the name of his inventor. Figure 3.48 shows the possible mechanization of a single-axis navigation system known as a Schuler tuned system.
Figure 3.48: Single-Axis tuned Schuler System
The measured specific force is resolved into the reference frame stored within the INU. The resolved component is integrated twice to get velocity and position. Transport rate is computed by dividing the indicated velocity by the radius of the Earth. This signal is then used to modify the stored attitude reference as the inertial system moves around the Earth. In the event that the stored attitude reference is in error by a factor , the horizontal direction indicated by the INU will be tilted by with respect to the true horizontal, and the resolved accelerometer measurement will include a gravity component g. Referring to figure 3.48, the resulting closed loop (Schuler loop) is unstable because of the presence of the two integrators. Its dynamics are governed by a simple harmonic
59
30/01/2004
motion with natural frequency called Schuler frequency and a period called Schuler period.
g = 0.00124 rad / sec s = R Ts = 2 = 84.4 sec s
(102)
Therefore, the Schuler oscillation can be considered as the motion of a hypothetical pendulum with length equal to the radius of the Earth. A pendulum, tuned to the Schuler frequency will always indicate the vertical on a moving vehicle, provided it has been initially aligned to it.
3.5.3. Error Sources
A precise evaluation of errors is critical to the precision of any INU. In the preceding section, the error dynamics were described in terms of linearized equations, resolved in different reference frames. Sensor performance, sensor models, and integration algorithms are all concurring to the error characteristics. Consider as an example, the Schuler pendulum mechanization from figure 3.48. The frequency of oscillation will be excited by the presence of errors in the initial estimates (o, o, xo) and fixed biases in the sensors (yb, fxb). The propagation of these errors can be computed by direct integration of the linearized error equations in one of the mechanizations (see figs. 3.45, 3.46, 3.47). Figure 3.49 shows the Schuler pendulum component in a 2D INU simulation.
Figure 3.49: 2D INU Simulation Diagram

60
30/01/2004
Sensor errors due to fixed biases, tolerance, g-sensitive drift, scale-factor asymmetry are usually included at time of purchase, according to the vendors data sheets. Other errors can, however, be modelled in a rather general form using random constant and random process dynamic models. An example is shown in figures 3.50 and 3.51 for gyros and accelerometers respectively.
Figure 3.50: General Gyro Sensor Model
Figure 3.51: General Accelerometer Sensor Model
An example of navigation simulation with error propagation due to biases in the sensors is shown in figure 3.52.
61
30/01/2004
Figure 3.52: Navigation Error Propagation Example
3.6. Navigation Aids
The precision of an INU can be improved by independent, redundant information. In addition, external navigational aids can be used as backup during partial or total failures of the inertial unit. Navigational aids are several, and they use different principles. Among the many, we can mention: Doppler velocity sensors Tactical air navigation range Forward looking and LOS radars Global Positioning Systems
3.6.1. Doppler Velocity Sensors
These devices use the same Doppler Effect principle; however they may have different mechanizations depending on the vehicle and medium. Consider a source emitting waves with period . A stationary observer is met by N waves over a time period t. t N= For an observer moving towards a stationary source with velocity V, the number of additional oscillations is
62
30/01/2004
V t If the source is emitted from the observer and reflected back, the number doubles (reflected moving source at observer) 2V t N = The frequency change is the number of additional oscillations during the time interval:
N =
N 2V 2V = = c t If there is a slant angle q between emitter/observers velocity and reflecting source, the Doppler frequency shift is given by: = =
As shown in figure 3.5.3. 2V cos c (103)
Figure 3.53: Doppler Effect Example
Usually a 4-beam configuration (Janus) is used, computing the vehicles velocity from their frequency changes. Errors in the Doppler unit (usually body-fixed) are due to scale factors, bias and noise, and installation misalignment. A representation of these errors can be given in a manner similar to figures 3.50 and 3.51.
dBx dSFx dBSxy dBSxz b b (104) DOP = dBy + dBSyx dSFy dBSxy DOP + dW dBz dBSzx dBSxy dSFz In (104), there is bias, scale factor-bore sight, noise components. In general applications, the output from Doppler systems is combined with INU data for additional processing by a Kalman filter. The Doppler divergence is defined as: DOP = INU DOP
(105)
63
30/01/2004
Eq. (105) can be expanded in a generic navigation frame n about the nominal velocity, yielding:
n n n n b n DOP = INU + INU Cb DOP + DOP
= ... =
n INU
(106)
C
n b
b DOP
C
n b
b DOP
3.6.2. TACAN Range Measurement 3.6.3. FLIR 3.6.4. Global Positioning System
An extensive description of the history and characteristics of GPS as navigational aid can be found in the literature. A tutorial example is given by the following link at the University of Colorado. http://www.colorado.edu/geography/gcraft/notes/gps/gps_f.html. 3.6.4.1. General Description The Global Positioning System (GPS) is a space age navigational system that can pinpoint your position anywhere on the globe, usually within a few yards or meters. This amazing technology is available to everyone, everywhere, day and night, and best of all, at no cost for use of the navigational data. GPS uses a constellation of 24 satellites in precise orbits approximately 11,000 miles above the earth. The satellites transmit data via high frequency radio waves back to Earth and, by locking onto these signals, a GPS receiver can process this data to triangulate its precise location on the globe.
Figure 3.54: GPS System
GPS operates 24 hours a day, in all weather conditions, and can be used worldwide for precise navigation on land, on water and even in the air. Some of its many current applications include: boating, fishing, hunting, scouting on land or from the air, hiking, camping, biking, rafting, pack trips by horseback, hot air ballooning, general aviation, snowmobiling and skiing, search and rescue, emergency vehicle tracking, 4 wheeling,
64
30/01/2004
highway driving and a host of other outdoor activities where accurate positioning is required. GPS uses satellite ranging to triangulate your position. In other words, the GPS unit simply measures the travel time of the signals transmitted from the satellites, then multiplies them by the speed of light to determine exactly how far the unit is from every satellite own sampling. By locking onto the signals from a minimum of three different satellites, a GPS receiver can calculate a 2D (two-dimensional) positional fix, consisting of your latitude and longitude. By locking onto a fourth satellite, the GPS can compute a 3D (threedimensional) fix, calculating your altitude as well as your latitude/longitude position. In order to do this a solution is to use a 12 parallel channel receiver; three of the channels lock on to satellites for triangulation, another channel locks on to a fourth satellite for 3D navigation, which lets the unit calculate altitude in addition to latitude and longitude. These four channels continuously and simultaneously track the four satellites in the best geometrical positions relative to you. The additional eight channels track all other visible satellites, and then add this data to the data from the original four satellites. The unit then over-resolves a solution, creating an accuracy-enhanced reading. The additional channels also ensure reliable, continuous and uninterrupted navigation, even in adverse conditions such as valleys or dense woods.
Figure 3.55: GPS Satellite Constellation
GPS was conceived in the 1970s, and is controlled by the United States Department of Defense. Although GPS was initially envisioned for military use, the Government realized early on that there would be numerous civilian applications as well. Subsequently, the Department of Defense (DOD) created two transmission codes; the P code (Precision code) for military use, and the C/A code (Civilian Access code) for civilian use. The highest accuracy levels were to be reserved for the military so as to prevent hostile enemy attacks against the U.S. using our own navigational system. However, once in operation, the civilian GPS receivers using the C/A code proved to be more accurate than the DOD had intended. Consequently, the military developed a system for randomly degrading the accuracy of the signals being transmitted to civilian GPS
65
30/01/2004
receivers. This intentional degradation in accuracy is called Selective Availability or S/A. This reduced the civilian GPS accuracy levels to being within 100 meters or less, 95% of the time. However, typical accuracy for most users averaged between 20 and 50 meters the majority of the time. You could easily see the effects of S/A on a GPS receiver when you were not moving. Typically, there would be random movements in speed, altitude and position readings, along with slow position "wandering" on the plotter trail. This was easily seen when you were on a .1 or .2 mile zoom range, and not moving. For example, while parked at the dock in your boat, you would see unexplainable changes in your digital speed readings up to a few miles per hour, even though you were not moving.
Figure 3.56: Early GPS Military and Civilian Codes
Effective May 2, 2000 selective availability (S/A) has been eliminated. The United States Department of Defense now has the technology to localize the control system to deny GPS signals to selected areas. It is not often that your electronics products increase in value after you've purchased them. Now boaters, aviators, drivers, hikers, hunters and outdoor enthusiasts of all types can locate their position up to ten times more precisely (within 10 to 20 meters) and navigate their way through unfamiliar terrain. Anglers can now return to their favorite spot on a lake or river instead of just their favorite area.
Figure 3.57: Current and Future GPS Precision

66
30/01/2004
The decision to allow civilians so much accuracy in location information was finally made because GPS is continually playing a more important role in the lives of people around the world - it's becoming an international utility. GPS is the global standard in navigation because it is completely free of charge to the public.
Figure 3.58: Differential GPS Precision
Differential GPS, or DGPS, has been developed to improve GPS accuracy to within a few meters. DGPS was originally initiated by the U.S. Coast Guard to counter the accuracy degradation caused by Selective Availability. Even with S/A now eliminated, DGPS continues to be a key tool for highly precise navigation on land and sea. DGPS technology adds a land-based reference receiver located at an accurately surveyed site to the other GPS components. This non-moving DGPS reference station knows where the satellites are located in space at any given moment, as well as its own exact location. This allows the station to compute theoretical distance and signal travel times between itself and each satellite. When those theoretical measurements are compared to actual satellite transmissions, any differences represent the error in the satellite's signal. All the DGPS reference station has to do is transmit the error factors to your DGPS receiver, which gives the information to the GPS receiver so it can use the data to correct its own measurements and calculations. GPS errors are a combination of noise, bias, and blunders. Noise errors are the combined effect of PRN code noise (around 1 meter) and noise within the receiver noise (around 1 meter). Bias errors result from Selective Availability and other factors. The potential accuracy of the C/A code of around 30 meters is reduced to 100 meters (two standard deviations). The SA bias on each satellite signal is different, and so the resulting position solution is a function of the combined SA bias from each SV used in the navigation solution. Because SA is a changing bias with low frequency terms in excess of a few hours, position solutions or individual SV pseudo-ranges cannot be effectively averaged over periods shorter than a few hours. Differential corrections must be updated at a rate less than the correlation time of SA (and other bias errors). Other Bias Error sources;
67
30/01/2004
SV clock errors uncorrected by Control Segment can result in one meter errors. Ephemeris data errors: 1 meter. Troposphere delays: 1 meter. The troposphere is the lower part (ground level to from 8 to 13 km) of the atmosphere that experiences the changes in temperature, pressure, and humidity associated with weather changes. Complex models of troposphere delay require estimates or measurements of these parameters. Unmodelled ionosphere delays: 10 meters. The ionosphere is the layer of the atmosphere from 50 to 500 km that consists of ionized air. The transmitted model can only remove about half of the possible 70 ns of delay leaving a ten meter un-modelled residual. Multipath: 0.5 meters. Multipath is caused by reflected signals from surfaces near the receiver that can either interfere with or be mistaken for the signal that follows the straight line path from the satellite. Multipath is difficult to detect and sometime hard to avoid. Blunders can result in errors of hundred of kilometers. Control segment mistakes due to computer or human error can cause errors from one meter to hundreds of kilometers. User mistakes, including incorrect geodetic datum selection, can cause errors from 1 to hundreds of meters. Receiver errors from software or hardware failures can cause blunder errors of any size. Noise and bias errors combine, resulting in typical ranging errors of around fifteen meters for each satellite used in the position solution.
Figure 3.59: GPS Error Sources
Figure 3.60: DGPS Error Reduction
The predicted precision increase using differential and real-time data link is shown in the next figure:
68
30/01/2004
Figure 3.61: Current and Predicted RT Kinematics
Figure 3.61: GPS Usage Projections
3.6.4.2. GPS Functionality GPS data can be given in different formats: typically they consist of time difference and Doppler Effect in terms of pseudo range and delta range between GPS receiver and GPS satellites. Some units have additional capability of providing the information in local geodetic (latitude and longitude) or ECEF coordinates. The satellite motion model used by a generic GPS receiver is based on central motion approximation as shown in figure 3.6.2, this leads to Keplerian orbits and equations:
2 r r = 2 r r + 2r = 0
69
30/01/2004
Figure 3.62: Central Force Motion
This yields the following solution: r = p 1 + e cos 2 p = a 1 e (107)
M = E e sin E M= 2 t= a3 = G [M + m ]
Where M is the Mean Anomaly and E is the Eccentric Anomaly.
Figure 3.63: Mean and Eccentric Anomalies
70
30/01/2004
The GPS satellite position is defined in terms of its Orbital Elements, a unique set of parameters that identify its position at any given time. The previous table and figure 3.64 give the standard definition of orbital parameters.
Figure 3.63: Orbital Elements
With the orbital parameters known, the position in ECEF coordinates can be computed:
x ECEF = r [cos cos sin cos sin ] ECEF = r [cos sin + sin cos cos ] y ECEF = r sin sin z = +
(108)
Provided that 4 satellites are needed for complete 3D information, the choice depends mainly on their visibility with respect to the user. One parameter selected for visibility is the elevation angle, which is determined by geometric considerations from figure 3.64.
Figure 3.64: Satellite Selection Parameters

71
30/01/2004
In addition to selecting a satellite based on its elevation angle, satellites are chosen based on their quadruple relative geometry with respect to the receivers antenna. This geometry is defined by the geometric dilution of precision parameter or GDOP. In general, ranging errors from the SV signals are multiplied by the appropriate GDOP term to estimate the resulting position or time error. Various GDOP terms can be computed from the navigation covariance matrix. ECEF XYZ DOP terms can be rotated into a North-East Down (NED) system to produce local horizontal and vertical DOP terms. The standard GDOP Components are: PDOP = Position Dilution of Precision (3-D), sometimes the Spherical DOP. HDOP = Horizontal Dilution of Precision (Latitude, Longitude). VDOP = Vertical Dilution of Precision (Height). TDOP = Time Dilution of Precision (Time). While each of these GDOP terms can be individually computed, they are formed from covariances and so are not independent of each other. A high TDOP (time dilution of precision), for example, will cause receiver clock errors which will eventually result in increased position errors.
Figure 3.65: Qualitative Locations for poor and good DGOP
For a selected group of satellites, the chosen quadruple is the one having the lowest GDOP. Define the i-th satellite LOS unit vector with respect to the local horizontal navigation frame:
L L e eLOS = Ce eLOS
Assume that the ranges are biased by a users clock error b, the relative pseudo-range is given by:
i =
( xi p1 )
+ ( y i p2 ) + ( zi p3 ) + b
2 2
(109)
If the receivers antenna position has an error p , we can compute the linearized value of pseudo-range error (normalized to the pseudo-range) i for the 4 satellites:
72
30/01/2004
1 p1 p 2 = H4 x 4 2 = H4 x 4 x z= 3 p3 b 4
If the assumed corrections to position and time are random, resulting from variable satellite positions, the expected error in position and time is given by its covariance: Cov { x } = E x x T = H 1E z z T H T T 2 E z z = PSEUDO I4 x 4
(110)
Where equal pseudo-range variance for all 4 satellites rewritten: LT eLOS 1 T L eLOS 1 2 2 T H H = = ( ) LT PSEUDO eLOS 3 LT eLOS 4 From = ( )
T
is assumed. Eq. (110) can be

1 1 1 1
(111)
( )
, we can compute for instance:
4 GDOP = ii i =1 2 HDOP = ii i =1 A summary of the previous derivations is shown in figure 3.6.6.
(112)
Figure 3.66: GPS Measurement Model

73
30/01/2004
A simulation of GPS performance in navigation setting is shown below.
Figure 3.67: GPS Navigation Example
3.7. Integrated INS-KF, INS-GPS Systems
The use of state estimation with static filtering techniques (Least-Squares, weighted LS) and dynamic estimators (KF, EKF,..) is common practice in order to improve INU performance. With the application of Kalman filtering, independent redundant sources of navigation information are combined with a reference navigation solution, to obtain optimal estimate of navigation states and other variables contributing to navigation solution error. In certain instances, the navigation solution coming from different device can also be integrated in order to compensate errors due to drift (INU) and/or loss of signal (GPS). Some of the benefits and properties of integration are: INS gives accurate estimates of aircraft orientation. GPS provides accurate estimates of aircraft position. INS solutions are generally computed 100 times per second. GPS solutions are computed once per second. GPS in subject to jamming, INS is not.
74
30/01/2004
Combining GPS and INS provides accurate and robust determination of both translational and rotational motion of the aircraft. Both translational and rotational motions are required to locate targets on the ground from the aircraft.
In a graphical format, INS-GPS integration, combined with Kalman filtering estimation is shown in figure 3.68.
Figure 3.68: INS-GPS integration with KF
The Kalman filter structure is standard, and operates on the covariance matrix update. The equations describing the filter are given by:
Figure 3.69: Kalman Filter Equations
The filter processing flow begins at some time instant (k-5), the state and covariance matrix are propagated in time to the next step (k-4). At this time instant, the
75
30/01/2004
measurements are available for use in updating, and one or more measurements may be processed by the state vector and covariance matrix measurement update equations. With the measurement update completed, there is a further time propagation to the time (k-3). The process is summarized in figure 3.7, and an example of simulation is shown in figure 3.71.
Figure 3.70: Processing Flow of KF Update
Figure 3.71: KF Estimate Simulation
76
30/01/2004
In practice, the KF implementation can be mechanized in two ways: feedforward and feedback mechanizations. In the feedforward mechanization, the Kalman filter has no influence on the INU and operates on the output only. This simplified implementation may have problems in the long run.
Figure 3.72: Feedforward KF Mechanization
An alternate solution is the feedback mechanization shown in figure 3.73. In this case, the Kalman Filter is capable of estimating a variety of signals characterized by non zero mean biases in the sensors measurements and GPS data.
Figure 3.73: Feedback KF Mechanization
77
30/01/2004
Simulations can be carried out, to verify how the integrated INU/GPS system is capable of dealing with changing noise characteristics.
Figure 3.74: Feedback KF Mechanization Simulations
3.8. Examples and Applications
Two examples are described on separate files:
3.8.1. INU Calibration The performance of inertial navigation system (INS) is largely dependent on calibration errors of gyroscopes and accelerometers. A typical gimballed INS is calibrated while stationary by multi-position rotating the inertial platform relative to the earths spin axis and local gravity vector in order to excite the gyroscope and accelerometer parameters. And, there is another technique which inertial sensor parameters are estimated by least square method using velocity errors obtained from so called single-axis Schuler calibration test in navigation mode. These have led to lengthy calibration procedures. In this study, we propose the fast calibration technique by Kalman filtering with the modified level axis Schuler error loop in which the Schuler factor is introduced in order to be able to change the Schuler period. The simulation results show that the calibration accuracy is improved as the Schuler period is decreased. Furthermore, fast calibration results from reducing the Schuler period. Consequently, our proposed calibration technique has led to enhanced accuracy in the instrument errors estimation while reducing the calibration test time. (Fast Calibration Technique for a Gimballed Inertial
78
30/01/2004
Navigation System, Yong-Jin Shin, Jeong-Hwa Park, Cheon-Joong Kim, Korea Aerospace Research Institute, Yuseong P.O.Box 113, 305 600 Daejeon, Korea, ICAS2002)
3.8.2. Design of a 3DOF Sensor Unit The scope of the application is to describe the design procedure of a low-cost strapdown inertial unit capable of giving motion information on a plane. (Progetto e Realizzazione di na piattaforma Inerzile low-cost di tipo Strapdown, Cellini Manuele, Mati roberto, DSEA, Universit di Pisa, 2004).
79
30/01/2004

Navigation

Încărcat de

Informații document

Descriere originală:

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Navigation

Încărcat de

Drepturi de autor:

Formate disponibile

Guidance and Navigation Systems CHAPTER 3: NAVIGATION

Figure 3.1: Navigation Problem

Sistemi di guida e navigazione, a cura di Mario Innocenti

Figure 3.2: Latitude and Longitude Definitions

Sistemi di guida e navigazione, a cura di Mario Innocenti

Figure 3.3: Example of referenced Position Variables

If the basic information comes from a RADAR-DOPPLER sensor, then:

Sistemi di guida e navigazione, a cura di Mario Innocenti

Figure 3.4: Dead-Reckoning Mechanization

Sistemi di guida e navigazione, a cura di Mario Innocenti

Figure 3.5: Inertial Navigation Structure

3.2. Inertial Navigation Schemes

Figure 3.6: Schematic Representation of Stabilized Inertial Platform

Figure 3.7: Mechanization of Stabilized Inertial Platform

Sistemi di guida e navigazione, a cura di Mario Innocenti

Figure 3.8: Schematic Representation of Strapdown Inertial Platform

Figure 3.9: Mechanization of Strapdown Inertial Platform

Sistemi di guida e navigazione, a cura di Mario Innocenti

Figure 3.10: Signal Flow in Inertial Navigation

Example: Consider a simplified implementation of a 2D inertial navigation unit (Strapdown).

Figure 3.11: Basic Components of a Strapdown INU

Sistemi di guida e navigazione, a cura di Mario Innocenti

Figure 3.12: Meridian View of a 2D Strapdown Navigation

Sistemi di guida e navigazione, a cura di Mario Innocenti

Figure 3.13: Frames for 2D Navigation about the local Vertical

Sistemi di guida e navigazione, a cura di Mario Innocenti

3.3. Inertial Instruments

Figure 3.14: Physical Model of an Accelerometer

Sistemi di guida e navigazione, a cura di Mario Innocenti

Sistemi di guida e navigazione, a cura di Mario Innocenti

Figure 3.15: Schematic of a restrained Pendulum Accelerometer

Figure 3.16: Pendulous fibre-optic Accelerometer

Sistemi di guida e navigazione, a cura di Mario Innocenti

Figure 3.17: Vibrating Accelerometer 1

Figure 3.18: Vibrating Accelerometer 2

Sistemi di guida e navigazione, a cura di Mario Innocenti

Figure 3.19: Historical Review (taken by Sperry-Rand Corporation)

Sistemi di guida e navigazione, a cura di Mario Innocenti

Figure 3.20: Inertia Stiffness about Spin Axis

Sistemi di guida e navigazione, a cura di Mario Innocenti

Figure 3.21: Stationary Wheel hanging from a Rope

Figure 3.22: Precession Explanation

Figure 3.22: Relationship between Angular Momentum and Precession

Sistemi di guida e navigazione, a cura di Mario Innocenti

Figure 3.23: Precession Torque Details

Sistemi di guida e navigazione, a cura di Mario Innocenti

Figure 3.24: Another Look at the principle of Gyroscope

Sistemi di guida e navigazione, a cura di Mario Innocenti

Figure 3.25: Gimballed Gyroscope

Sistemi di guida e navigazione, a cura di Mario Innocenti

Figure 3.26: Example of a Two-Axis Gimbaled Gyroscope

Figure 3.27: Free Body Diagrams of a Two-Axis Gimballed Gyro

Sistemi di guida e navigazione, a cura di Mario Innocenti

The angular velocity of the rotor I then:

= 0 = Mir + Mig + Mio

The inertial moments can be expressed in terms of angular momentum:

Sistemi di guida e navigazione, a cura di Mario Innocenti

Where the angular momentum expressions are:

Rewriting (14) as:

cos Mir + Mig

Using (15) and (16), yields:

This becomes finally:

Sistemi di guida e navigazione, a cura di Mario Innocenti