Proceedings of the 18th World Congress

The International Federation of Automatic Control

Milano (Italy) August 28 - September 2, 2011

A dynamic model for recycling: optimization of solid waste separate collection

D. Anghinolfi*, M. Paolucci*, M. Robba*, A.C. Taramasso*

*Department of Computer, Communication and System Sciences, University of Genova, via Opera Pia 13 16145 Genova
(email: davide.anghinolfi@unige.it;massimo.paolucci@unige.it, michela.robba@unige.it, act@dist.unige.it)

Abstract: Recycling is an important part of waste management (that includes different kinds of issues:
environmental, technological, economic, legislative, social, etc.). In this work, differently from many
works in literature, attention is focused on recycling management and on the dynamic optimization of
materials collection. The developed dynamic decision model is characterized by state variables,
corresponding to the quantity of biomass in each bin per each day, and control variables determining the
quantity of material that is collected in the area each day and the routes to be followed. The developed
decision model is integrated in a GIS-based Decision Support System that allows calculating daily waste
generation.
Keywords: Mathematical Model, Optimization, Environmental Engineering, Dynamic Model, Inventory
Control

completely different concepts. Optimization tools provide

1. INTRODUCTION
decision-makers with optimal policies for any region based on
Solid waste management includes different kinds of issues economic and environmental costs, capacity limitations,
(environmental, technological, economic, legislative, social, location, etc. LCA models are not structured to advise
etc.) that should be taken into account in order to find decision makers about waste management policies: they
solutions that are economically appealing and present the environmental and economic impacts of various
environmentally sustainable. policies so that decision-makers can take better decisions.

In literature, there are works that deal with the integration of Finally, waste generation is another aspect present in the
the different aspects to find effective planning and correlate literature that is important for waste management
management strategies. Two main classes of problems can be and that can be used for the modelling, simulation and
found: waste collection, and waste disposal and treatment. In optimization of policies. Ojeda-Benitéz et al. (2008) define
this work, attention is focused on the optimization of separate mathematical models that correlate the residential waste
collection for recycling. Both decision problems can be seen generation per capita per day to the following variables:
as planning (i.e., the setting of the necessary facilities) or education, income per household, and number of residents.
management problems (i.e., the dynamic modelling and
The aim of this work is the definition of a decision model for
optimization of the activities). From a planning point of view,
recycling based on data collected in the territory and
different Decision Support Systems (DSS) have been
geographical considerations (mainly related to the distance
developed and tested in the field of municipal waste
between buildings and bins already positioned in the studied
management (Costi et al. (2004), Minciardi et al. (2008), Li
area) through the use of GIS and optimization tools. In
and Huang (2006)) that integrate GIS, optimization
addition, a simple GIS-based waste generation model to
techniques, and technological issues. Few works can be found
forecast the dynamic filling of the different bins for recycling
about management problems that involve a dynamic
is adopted and described in the following.
optimization of the different operations.
In literature there are different attempts to optimize the
Other approaches may be found to evaluate impacts and
logistics of materials to be sent to recycling. For example,
strategies of waste management options and to assess
Lopez Alvarez et al. (2008) propose a methodology for the
sustainable development criteria. Various assessments
design of routes for the ‘‘bin to bin’’ collection of paper and
methods are present, such as Modified Cost Effectiveness
cardboard waste from small business premises. The tool used
Analysis (MCEA), Life Cycle Assessment (LCA), Risk
is ArcGIS. The objective is to estimate the number of paper
Assessment (RA), Swiss Technical Ordinance Waste (STVA),
and cardboard bins, identify the needs of waste collection in
Sustainability Potential Analysis (SPA) (Lang et al., 2007). In
the examined area, and draw the best route that links the main
the same way calculation models based on methodologies
points along the main street of the city. Lin and Chen (2009)
from Life Cycle Assessment (LCA) have been developed to
develop an optimization model to assist local authorities with
evaluate different waste treatment options, and scenarios for
the identification of regions that require additional recycling
municipal solid waste. LCA is also used as a tool of
depots for better access and integration with private facilities.
sustainable design for waste management. Technically,
Abou Najm and El-Fadel (2004) assess the importance of
optimization models and assessment methods are based on

978-3-902661-93-7/11/$20.00 © 2011 IFAC 12940 10.3182/20110828-6-IT-1002.01891

18th IFAC World Congress (IFAC'11)
Milano (Italy) August 28 - September 2, 2011
optimization models for waste management in order to
integrate all the different issues and help decision maker in
defining strategies. Furthermore, they stress the importance to
develop software tools which are user-friendly, and they
introduce an interface that can be used to work with an
integrated waste management optimization model.
Solid waste collection and hauling account for the greater part
of the total cost in modern solid waste management systems.
It is shown and demonstrated by Johansson (2006) that
dynamic scheduling and routing policies exist that have lower
operating costs, shorter collection and hauling distances, and
reduced labour hours compared to the static policy with fixed
routes and pre-determined pick-up frequencies currently
employed by many waste collection operators. To this end, in
a recent initiative, 3300 Swedish recycling bins have been
fitted with level sensors and wireless communication
equipment, thereby giving waste collection operators access to
real-time information on the status of each bin.
In this work attention is focused on recycling management
and, in particular, on the dynamic optimization of materials
collection. Results are reported on maps through the use of
GIS tools and shown to decision makers through figures and
simple tables coming from a relational database. The
developed models and presentation of the results have been
developed and tested together with the end users responsible
for waste collection in the case study area.
Section 2 describes the overall architecture of the DSS for the
optimization of separate collection. Then, the system model
(i.e., the waste generation estimation methodology, and the
logistic problem) is presented. In Section 4, the optimization
model is described in detail. Finally, results and conclusions
are reported.
2. THE GIS-BASED DSS ARCHITECTURE
The developed DSS is characterized by the integration of an
optimization model, a GIS system, a database and a model to
predict the daily waste generation for each material to be
recycled. In particular, through the GIS system, the area of
reference for a specific bin in a given location is calculated.
The forecasting model is able to predict the waste generated in
an area and daily transported by citizens to the collection
points (bins) on the basis of the GIS calculations and the
available data for the study area. Once all data are stored in
the database, the optimization model determines an optimal
collection plan taking into account both the costs of collection
and transportation, and the benefits due to recycling. A
collection plan determines for each day and for each of the
available vehicles how many trips must be performed, the
route to follow during such trips, and the quantity of material
to be collected.
The proposed architecture of the DSS, reported in Figure 1,
includes the following modules:
12941
• a GIS module to evaluate the area of reference for each
bin;
• a database to store data and save results;
• a simple waste generation module;
• an optimization module.
GIS Waste
Generation
Model
Data
Base
Optimization
Figure. 1. The DSS architecture.
The area related to a specific bin is calculated on the basis of
a distance criterion, having divided the territory in polygonal
areas through the Thiessen approach. This method calculates
the areas that are the closest to each bin, then assigning to
them a number of inhabitants and commercial activities.
Let As be the reference area related for a bin s, with s=1,…,S,
and Pst the number of inhabitants in the area during day t, with
t=0,…,T-1 (in this way we distinguish between week-end and
working days). Similarly, N sm t
represents the number of
commercial activities of type m, m=1,…, M, in the area As in
the day t. The unit waste production for each person and each
commercial activity has to be defined for computing the waste
daily transported to each bin. Of course, this value depends on
each area (i.e. on the number of inhabitants, the collection
system, the bins disposal, the participation of population to
recycling). Let Dptfs be the waste of material f, with f=1,…,F,
generated per capita in the area As during day t [m3/inh. day],
t
and Dcmfs the waste of material f generated in the area As
during day t by a commercial activity m [m3/act. day]. Then,
the waste generated for each material in each day and in each
area D tfs [m3/day] is given by
M
D tfs = Pst Dp tfs + t
∑ Dcmfs t
N ms t=0,…,T-1 f=1,..,F s=1,...,S (1)
m =1
3. THE SYSTEM MODEL
The system is characterized by a number of bins in a territory,
possible routes, a depot and a recycling area where all
collected materials are transported. Figure 2 shows the
considered system.
18th IFAC World Congress (IFAC'11)
Milano (Italy) August 28 - September 2, 2011
CELL 2 CELL 3
S
D plastic): vehicles can serve both kinds of waste, but only once
CELL 1
Figure. 2. The possible vehicle routes in the considered
system (three cells, one depot (D), one recycling area (S)).
The following assumptions are made:
• the position of the bins in the territory is known;
• the daily waste generation is predicted from the GIS and
available data, through relation (1);
• the bins are grouped in cells (i.e. macro bins); Si denotes
the set of bin grouped as cell i;
• in each cell the path within the bins is known and fixed
(i.e., it is a-priori known the sequence of bins that are
emptied);
• each cell is modelled as a stock with a specific state
equation (the state variables are related to waste volume
over time); for each cell we compute D tfi = ∑ s∈S i D tfs the
total waste of each material f generated during day t in the
its bins, and estimations of the distance travelled and time
spent in the cell by a vehicle during a trip collecting a
certain amount of waste (Figure 3 depicts the detail of a
cell);
• the control variables are: the possible path variations
among cells (path stopping and entering), the daily
quantity of waste to be collected in a specific cell during
a specific trip, during each day, and through a specific
vehicle kind.
A vehicle enters the generic cell i; the waste is
generated in each bin
Si
The waste is collected; its volume in the cell
diminishes; the vehicle exits the path or continues
with a next cell
Figure. 3. The detail of a generic cell.
4. THE OPTIMIZATION PROBLEM
The problem can be represented by a directed graph G=(N, A)
as depicted in Figure 2. Node 0 corresponds to the depot D,
12942
node n to the recycling facility S, whereas the other nodes to
cells. The availability of a set of vehicles with different
capacities is assumed. For each vehicle used during the
planning period a fixed cost is taken into account, which is
charged for each day a vehicle is used. The model also
assumes a set of different kinds of waste (e.g., paper and
at a time (that is, each trip is devoted to collect a single type
of waste). The purpose is to determine for each day in the
planning horizon the collection plan for vehicles (i.e., the type
and quantity of waste collected during each trip and the route
followed) so that both the cell and the vehicle capacities are
not exceeded and the overall minimum percentage
requirement for collection is reached. Several objectives must
be taken into account: the value of recycling (benefit) and the
cost of service and residual, i.e., not collected, waste at the
end of the planning horizon. Such objectives are all included
in a scalarized objective function modeling the overall cost.
The designed mathematical programming model is a mixed
integer linear programming (MILP) one which is detailed in
the following, after having defined the relevant quantities, i.e.,
sets, constants and variables.
4.1 Sets
T set of days
C set of cells
K set of vehicles
F set of kinds of waste (e.g., F={1,2} where 1=paper,
2=plastic)
V set of trips
N set of nodes N={0,…,n} 0=depot, n=recycling facility
A set of arcs
4.2 Constants
− W fi0 initial waste level of material f in node (cell) i
− D tfi waste of type f generated in node i in day t (m3)
− Bfi maximum waste level of type f in node i
− CPfk capacity of vehicle k for waste of type f
− CPk = max CPfk maximum capacity for vehicle k
f
− TTij travel time of arc (i, j)
− TTCi average travel time for collecting one m3 of waste of
type f in cell i, computed as (TC+TRi)/VCf, where TC is the
time for loading a bins, TRi the average time between two
bins in cell i and VCf the average size of a bin for material f
− TS time for unloading a vehicle in the recycling facility
− Lij length of arc (i, j)
− LCi average distance for one m3 of collection of waste of
type f in cell i, computed as Li/VCf, where Li average
distance between two bins in cell i
− PC minimum percentage of total collected waste
− CVk overall cost for kilometer for vehicle k
− CFk fixed cost of vehicle k
− BCf price for m3 of collected waste of type f
4.3 Variables
18th IFAC World Congress (IFAC'11)
Milano (Italy) August 28 - September 2, 2011

− wtfi ∈ ℜ + i ∈ C , t ∈ T , f ∈ F waste level for waste type f 0 ≤ wtfi ≤ b fi i ∈ C, f ∈ F , t ∈ T (16)

t t
−r fikv ∈ ℜ + f∈F, i∈C, t∈T, k∈K, v∈V waste collected in i q ikv ≤ CPfk i ∈ C, f ∈ F , k ∈ K , t ∈T (17)
specifying type
− qikvt
∈ ℜ + i∈C, t∈T, k∈K, v∈V waste in vehicle k trip v after The objective (2) minimizes the sum of vehicles’ costs
leaving cell i (traveling and fixed costs) and costs due to positive w|Tfi | , i.e.,
− l kt ∈ ℜ + i∈C, t∈T total travel distance for k in t the quantity of waste remaining in the cells at the end of the
planning period, and maximizes the revenue for collected
− akt ∈ ℜ + i∈C, t∈T total travel time for k in t waste. Constraints (3) determine the level of waste of the
t t different types in the cells during the planning horizon (note
− xijkv ∈ {0,1} (i, j)∈A, t∈T, k∈K, v∈V, xijkv = 1 if vehicle k
t
during trip v on day t travel on arc (i,j) that r fnkv = 0 as node n is the recycling facility). Constraints
− u k ∈ {0,1} k∈K, uk = 1 if vehicle k is used (4) define the initial levels of waste in the cells. Constraints
(5) impose that each vehicle k in day t can execute at most
− c tfkv ∈ {0,1} f∈F, t∈T, k∈K, v∈V, c tfkv = 1 if waste collected once the trip v. Constraints (6) model travel continuity
by k during trip v in day t is of type f conditions, i.e., if a vehicle k in day t enter a node j during trip
t
− γ = ( r fikv t
, x ijkv , u k , c tfkv ; i ∈ C , f ∈ F , t ∈ T , k ∈ K , v ∈ V ) v, then the same vehicle must also leave the node, whereas
constraints (7) the travel sequence, i.e., trip v in day t of
vector of all the defined decision variables.
vehicle k can be executed only after the completion of trip v-1
in t. Constraints (8) impose that the vehicle’s load leaving cell
i must be not lower than the amount of refuse collected in i.
4.4 MILP model
Constraints (9) are the VRP sub-tour elimination constraints
The problem is formulated as follows. which also allow the computation of the vehicles’ load after
serving cells. Constraints (10) establish that only used
vehicles can leave the depot. Constraints (11) and (12)
min ∑ ∑ l kt ⋅ CVk + ∑ u k ⋅ CFk +
γ t∈T k ∈K k ∈K determine the waste type collected by a vehicle during a trip.
(2) Constraints (13) and (14) define respectively the total travel
− ∑ ∑∑∑ t
∑ BC f ⋅ r fikv + ∑ ∑ BC f ⋅ w|Tfi | ) time and the total travel distance for vehicles in t. Constraints
t∈T k∈K v∈V i∈C f ∈F i∈C f ∈F
(15) impose the minimum percentage threshold for waste
collection. Finally, constraints (16) define bounds for the level
subject to of waste in cells during time and (17) for the maximum
vehicles’ capacity.
wtfi = wtfi−1 + D tfi − ∑ ∑ r fikv
t
i ∈ C, f ∈ F , t ∈T (3) Due to the complexity of mixed integer programming, which
k∈K v∈V is NP-hard, in general modeling a problem as a MILP would
w0fi = W fi0 i ∈ C, f ∈ F (4) not guarantee that acceptable solutions are generated within
suitable times; however, as shown in the experimental
∑ x0t jkv ≤ 1 t ∈ T , k ∈ K , v ∈V (5)
analysis, the dimension of the considered case study allowed
( 0 , j )∈A
t the MILP approach to produce good solutions in short
∑ xijkv = ∑ x tjikv t ∈ T , k ∈ K , v ∈ V , j ∈ N \ {0} (6) computation times.
( i , j )∈ A ( j ,i )∈A

∑ x0t jkv ≤ ∑ x tj 0 k ,v −1 t ∈ T , k ∈ K , v ∈V , v > 1 (7)

( 0 , j )∈A ( j , 0 )∈A
t t 5. APPLICATION TO A CASE STUDY: RESULTS
∑ r fikv ≤ qikv i ∈ C , t ∈ T , k ∈ K , v ∈V (8)
f ∈F In this section, we present the application of the proposed
q t
≥q t
+ CPk ⋅ ( x t
− 1) + ∑ r t optimization model to the case of the Cogoleto municipality,
jkv ikv ijkv fjkv
f ∈F (9) located in the Liguria region in the north of Italy, close to the
t ∈ T , k ∈ K , v ∈ V , (i, j ) ∈ A, i ≠ 0, j ≠ 0, i ≠ j city of Genoa. Specifically, Figure 4 reports the bins (each
circle includes different bins with different types of material)
x0t jk1 ≤ u k (0, j ) ∈ A, t ∈ T , k ∈ K (10) and the areas calculated with the Thiessen method.
t
∑c =1 fkv t ∈ T , k ∈ K , v ∈V (11)
f ∈F The system described in Section 3 is characterized for the case
r t
≤ CPfk ⋅ c t
f ∈ F , i ∈ C , t ∈ T , k ∈ K , v ∈V (12) study by three cells (Cell 1: Cogoleto center, Cell 2:
fikv fkv
Sciarborasca, Cell 3: Lerca).
akt = ∑ ( ∑ TTij ⋅ xijkv
t t
+ ∑ ∑ TTCi ⋅ r fikv t
+ ∑ TS ⋅xinkv )
v∈V ( i , j )∈ A i∈C f ∈F ( i , n )∈ A (13)
k ∈ K ,t ∈T
t t t
l = ∑ ( ∑ Lij ⋅ xijkv
k + ∑∑ LCi ⋅ r fikv ) k ∈ K , t ∈T (14)
v∈V ( i , j )∈ A i∈C f
t
∑ ∑ ∑ ∑ ∑ rfikv ≥ PC ⋅∑ ∑ ∑ D tfi k ∈ K , t ∈T (15)
k ∈K v∈V t∈T i∈C f ∈F t∈T i∈C f ∈F

12943
18th IFAC World Congress (IFAC'11)
Milano (Italy) August 28 - September 2, 2011

(2856 binary variables). We allotted 900 seconds as maximum

time for producing a solution as we verify that this duration
allows generating a good quality solution, so leaving human
planners the possibility of performing some what-if analysis.
The overall results obtained are reported in Table 3 and 4. In
particular, Table 3 shows for each scenario and waste type the
average daily collection, defined as the average over the days
when at least one trip was performed, the total waste
generated and collected during the 4 weeks, the total value of
the waste collected and the total collection cost, and finally
the quantity of waste remaining in the cells at the end of the
considered period. Table 4 focuses on the way the vehicles are
used in the two scenarios, reporting number of collection
Figure. 4. The area associated to each bin through the days, the average and maximum number of vehicles used
Thiessen method in the Cogoleto area. during the days when collection was performed, the total and
average number of trips per day, and finally the average trip
duration and length.
To calculate the waste generation in each cell and in each day As can be observed, Scenario 2 appears much more
(i.e., the time interval), the methodology reported in Section 2 challenging as for example a trip for plastic collection is
is used. Moreover, we consider two different scenarios over a performed almost every day and also paper is collected more
4 weeks (28 days) time horizon, assuming the same waste than the double of times with respect to Scenario 1.
generation level in each week. The time horizon is chosen as
regards the seasonality (i.e., changes in touristic population)
of the specific case study. Moreover, data from the waste Table 3. Aggregate performance indicators for the
manager are used to calculate waste generation and they are collection in the two scenarios.
available on a monthly scale. The first scenario is associated
Average Daily Total Waste Total Value of Total Cost Residual
with the current levels of waste generation for two types of Waste Type Scenario
Collection (m3) Generated (m3) Collection (m3) Collection (€) (€) Waste (m3)
waste, paper and plastic. The second scenario is defined Paper 1 22,00 177,65 176,00 774,40 955,79 1,65
assuming a substantial increase of waste generation and it is Plastic 1 42,75 384,72 384,72 1169,55 1457,45 0,00
used to analyze the impact of the optimization model under Paper 2 34,11 616,47 614,00 2701,60 3134,22 2,47
more critical operational conditions. We show the data for the Plastic 2 49,26 1335,07 1330,00 4043,20 5542,90 5,07

two scenarios respectively in Table 1 and 2, where, for each

waste type, the waste generated (m3) in the three macro cells
of Cogoleto municipality in each week day is reported. Table 4. Aggregate performance indicators for vehicle
trips

Number of Average Number Max Number Total Average Average Average

Table 1. The case study waste generation for scenario 1 Waste Type Scenario Days of of Vehicle Daily of Vehicles Number Number of Trip Trip
Collection Used Used in a Day of Trips Daily Trips Time Length
Paper 1 8 1,00 1 8 1,00 120,28 203,78
Waste type Cells Mon Tue Wed Thu Fri Sat Sun
Cogoleto 4.27 5.36 4.72 4.64 5.88 5.61 4.90 Plastic 1 9 1,00 1 18 2,00 104,56 150,63
Paper Sciarborasca 0.68 0.72 0.70 0.66 0.90 0.90 0.84 Paper 2 18 1,11 2 29 1,45 118,56 192,24
Lerca 0.49 0.51 0.47 0.47 0.62 0.57 0.53 Plastic 2 27 1,11 2 61 2,03 105,17 148,92
Cogoleto 9.25 11.60 10.22 10.05 12.74 12.15 10.60
Plastic Sciarborasca 1.47 1.56 1.51 1.43 1.95 1.95 1.82
Lerca 1.06 1.09 1.01 1.01 1.33 1.23 1.15 It can be noted that the total waste generated in Scenario 2 is

about three times greater than the one in Scenario 1.
Generated refuse is collected and only a small part is left on
Table 2: The case study waste generation for scenario 2 the territory. What is evident is that the recycling collection,
even if it is optimized, remains a cost (i.e., the value of
Waste type Cells Mon Tue Wed Thu Fri Sat Sun collection is less than the total cost for each Scenario and each
Cogoleto 14.84 15.93 15.31 15.21 18.58 18.30 17.59
Paper Sciarborasca 2.92 2.96 2.94 2.91 3.82 3.82 3.76 recovered material). The resulting net costs are:
Lerca 1.96 1.97 1.94 1.93 2.52 2.48 2.44
Cogoleto 32.15 34.50 33.15 32.95 40.23 39.64 38.09 Scenario 1: plastic 287.9 €, and paper 181.39 €;
Plastic Sciarborasca 6.33 6.42 6.37 6.29 8.26 8.26 8.14
Lerca 4.24 4.27 4.19 4.19 5.46 5.36 5.28 Scenario 2: plastic 1499.7 €, and paper 432.62 €.

The collection management obtained applying the
optimization model for Scenario 1 can be compared to the
We performed the experimental tests on an Intel Core 2
current operational policy. This policy prescribes for each
T4550 1.66 GHz 2Gb RAM notebook. The optimization
kind of material one day of collection a week during which
model was coded using IBM/ILOG commercial solver Cplex
exactly two trips are performed. Unfortunately no actual
12.2 (IBM, 2010). The MILP problem dimensions for both the
performance index values (i.e., costs and benefits) for the
scenarios are the following: 9414 constraints, 7986 variables
current operation policy in Cogoleto municipality are

12944
18th IFAC World Congress (IFAC'11)
Milano (Italy) August 28 - September 2, 2011

available for a comparison. However, an estimation of the
current operational policy performance was obtained by
imposing such a policy in the optimization model through an
appropriate set of constraints. Table 5 shows that the total cost
of optimized collection for Scenario 1 is higher than the
estimation of the current policy cost. However it is apparent
from Table 5 that (a) the total fixed charge of the optimized
collection is almost half the one of the estimated current
policy; (b) the value of collection of the optimized collection
is almost twice the one of the estimated current policy; (c) the
resulting net cost due of the estimated current policy turns out
to be more than 2.5 time greater than the one of the optimized
collection.
Table 5. Optimized collection vs current collection
Vehicle trip Vehicle fixed
Total cost
cost charge
Optimized collection 1,692.54 720.00 2,412.54
Current collection (est.) 1,011.38 1,280.00 2,291.38
6. CONCLUSIONS
A dynamic optimization model integrated in a GIS-based DSS
is presented to manage the recycling in the Cogoleto area
(Liguria Region, Italy). The model allows considering
different types of materials and vehicles for collection, a
variable number of bins. It optimizes the daily materials
recycling in order to minimize collection and transportation
costs, and to maximize benefits coming from materials sales.
Optimal results are reported for the small Cogoleto
municipality. Future developments will regard the testing of
the proposed model for a more complex case study (i.e., the
number of cells can be augmented till arriving to the number
of bins on the territory). Moreover, some noise for waste
generation variation can be taken into account.
IBM ILOG Cplex Optimization Studio, 2010
http://publib.boulder.ibm.com/infocenter/cosinfoc/v12r2/i
ndex.jsp
Johansson, M. (2006). The effect of dynamic scheduling and
routing in a solid waste management system. Waste
Management 26, 875–885.
Lang, D.J., Scholz, R.W., Binder, C.R., Wiek, A., and Staubli,
B. (2007). Sustainability Potential Analysis (SPA) of
landfills e a systemic approach: theoretical
considerations. Journal of Cleaner Production 15, 1628-
1638.
Li, Y.P., and Huang, G.H. (2006). An inexact two-stage
mixed integer linear programming method for solid waste
management in the City of Regina. Journal of
Environmental Management 81(3), 188-209.
Lin, H.Y., and Chen, G.H. (2009). Regional optimization
model for locating supplemental recycling depots. Waste
Value of
Net cost
management 29, 1473-1479.
collection Lopez Alvarez, J.V., Aguilar Larrucea, M., Ferna´ndez-
1,943.95 468.58
Carrion Quero, S., and Jimenez del Valle, A. (2008).
1,081.60 1,209.78 (258.18%)
Optimizing the collection of used paper from small
businesses through GIS techniques: The Legane´s case
(Madrid, Spain). Waste Management 28, 282–293.
Minciardi, R., Paolucci, M., Robba, M., and Sacile, R. (2008).
Multi-objective optimization of solid waste flows:
environmentally sustainable strategies for municipalities.
Waste Management 28 (11), 2202-2212.
Ojeda-Benıtez, S., Carolina Armijo-de Vegab, C., and
Marquez-Montenegroc, Y. (2008). Household solid waste
characterization by family socioeconomic profile as unit
of analysis. Resources, Conservation and Recycling 52,
992–999.

ACKNOWLEDGEMENTS

The authors would like to thank AMIU S.p.A. (the society

responsible for waste collection), with specific reference to
Dott. Paolo Cinquetti, for data collection and elaboration.

REFERENCES
Abou Najm, M., El-Fadel, M. (2004). Computer-based
interface for an integrated solid waste management
optimization model. Environmental Modelling &
Software 19, 1151–1164.
Costi, P., Minciardi, R., Robba, M., Rovatti, M., and Sacile,
R. (2004). An environmentally sustainable decision
model for urban solid waste management. Waste
Management 24, 277-295.

12945