Contents lists available atScienceDirect
Resources, Conservation and Recyclingj o u r n a l h o m e p a g e :w w w . e l s e v i e r . c o m / l o c a t e / r e s c o n r e c
An interactive optimization system for the location of supplementary
, Guan-Hwa Chenb
, Pei-Hao Leeb
, Chun-Hsu Linc
aDept. of Environmental Engineering and Management, Chaoyang Univ. of Technology, 168 Ji-Fong E. Rd., Wufong Township Taichung County, 41349 Taichung, Taiwan, ROC bInstitute of Environmental Engineering, National Chiao Tung University, Hsinchu, Taiwan, ROC
cEnergy and Environmental Research Center, Chung-Hua Institution for Economic Research, Taipei, Taiwan, ROC
a r t i c l e i n f o
Received 25 December 2008
Received in revised form 30 October 2009 Accepted 4 November 2009
Recycling depots allocation Decision support system Optimization model Recycling proximity Batteries recycling
a b s t r a c t
In order to increase participation in residential recycling, convenient access to recycling depots is essen-tial. Currently, most vendors have established various take-back systems in local retail shops. The integration of the recycling depots established by local authorities with these private facilities can sig-niﬁcantly improve access for residents. As a response to this need, an integrated system capable of providing optimization analysis and geographical information system functions has been designed to assist local authorities in identifying the regions where supplementary recycling depots need to be located to improve access to. A customized program which allows online optimization analysis has been devel-oped and embedded in an integrated system. Through a user-friendly interface, users can interactively locate the supplementary depots within the desired regions. The interactive process improves the ﬂexi-bility of the integrated system and avoids the shortcomings of impractical locations for recycling depots. A case study of battery recycling in central Taiwan is presented to demonstrate the effectiveness of the integrated system. By using the proposed integrated system to analyze the existing conditions, the loca-tions of supplementary points to improve access to the existing regional recycling collection points can be determined. In short, the proposed methodology provides local authorities with an uncomplicated and streamlined system with which to determine the number and locations of supplementary recycling points.
© 2009 Elsevier B.V. All rights reserved.
The treatment of municipal solid waste (MSW) is one of the pri-mary tasks for environmental authorities. Traditional treatments of MSW using landﬁlling and incineration have become more difﬁcult and expensive due to land scarcity and environmental concerns. Thus, the recycling of resources has become important as either an alternative or addition to traditional MSW treatments. Researchers (McDonald and Ball, 1998; Tilman and Sandhu, 1998) have iden-tiﬁed that the success of MSW recycling is decisively dependent upon the active participation of residents, which in turn is critically inﬂuenced by the proximity of drop-off depots. González-Torre and Adenso-Díaz (2005)also maintain that the distance between the drop-off depot and the residence affects participation and fre-quency in the recycling process: a shorter distance will signiﬁcantly improve the level of participation and also increase the quantity of material recovered.
∗ Corresponding author. Tel.: +886 4 23323000x4513; fax: +886 4 23742365. E-mail address:firstname.lastname@example.org(H.-Y. Lin).
The problems associated with drop-off depots have been stud-ied by researchers in the ﬁeld of MSW management, such asChang and Wei (2000),Kao and Lin (2002), andGautam and Kumar (2005). Typically, these models simulate the collection services provided by local authorities. A primary factor analyzed in these models is spatial proximity, represented in this case by walking distance for residents. Other factors including cost and vehicle capacities are accounted for in ﬁnding the optimal plan for collection systems or/and recycling depot locations. In essence, these methods are designed to make the modeling objective more attractive by sort-ing out the best from the available sites. However, three obstacles may discourage local authorities from utilizing these models. First, the identiﬁcation of possible choices is usually required for these models, which is a demanding step in itself, especially if the plan-ning area is vast. Second, ﬁnding the personnel with expertise and experience in optimization techniques who can manipulate these models is also impracticable for local authorities. Finally, due to the variance in the status of merchandise stipulated vendors, these take-back points may change frequently. As a result, a re-evaluation of the supplementary drop-off depot plans is required but may be impeded because of the amount of time usually needed to resolve these models.
0921-3449/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.resconrec.2009.11.001
616 H.-Y. Lin et al. / Resources, Conservation and Recycling 54 (2010) 615–622
The work involved in evaluating possible locations can be reduced by starting with the existing take-back points provided by the private sector, as only the regions with poor access to recy-cling facilities need to be analyzed. In Taiwan, because of the trend of Extended Producer Responsibility in waste management and the regulatory requirements, vendors of the products with stipulated recyclables assume the responsibility for the provision of drop-off containers or depots for recycling goods after use. These recyclables have either or both of the following properties: they are arbitrarily discarded (e.g. beverage bottles) and contain hazardous materials (e.g. batteries and ﬂuorescent lights). These private containers are widespread and effective in the collection of the designated recy-clable material and in providing closer proximity to the recycling depots in most regions. In other words, the local authorities only need to install supplementary recycling facilities in less serviced or less accessible regions. In previous work (Lin and Chen, 2009), an optimization model aiming to resolve these issues has been developed.
In order to address the obstacle posed by local authority per-sonnel who lack the ability to implement optimization models, an integrated system consisting of a user-friendly interface and ﬂex-ible functions has been developed. In addition, a tool capable of presenting and analyzing spatial information is essential to iden-tify regions with poor access to recycling facilities. Geographical information systems (GISs), featuring ﬂexibility, speed, accuracy and clarity, are widely utilized in studies with geo-referred objects. Chang et al. (1997)has asserted that GIS are ﬂexible in scenario analysis and applications. Clarke and Maantay (2006) has also applied GIS to analyze many recycling indicators in New York City and adjoining regions, and he has concluded that GIS are versa-tile in the presentation of geo-referred index values presentation. Utilizing GIS, factors that inﬂuence recycling efforts can be ana-lyzed clearly and presented intuitively. Other experiences using GIS in MSW management include landﬁll siting (Kao et al., 1996; Lin and Kao, 2005), collection routing (Shih and Lin, 1999; Sahoo et al., 2005), and the location of recycling depots (Kao and Lin, 2002; Chang et al., 2005). In addition to the GIS functions, an information system should also provide an approach to assessing different alter-natives for efﬁcient and realistic applications. For example,Chang et al. (2008)has provided a successful experience in the evaluation of landﬁll sites. GIS has also been used to screen potential landﬁll sites and they, in turn, have utilized the analytic hierarchy pro-cess to rank potential sites.Vaillancourt and Waaub (2002)have utilized GIS and multi-criteria analysis to evaluate waste facilities. They asserted that if spatial analysis functions and assessment tools can be combined or at least work more closely together, their sys-tem will be more effective in dealing with problems on a global scale.
While the application of optimization models are commonly used to solve MSW management problems (e.g.Koa and Chang, 2008; He et al., 2008; Chaerul et al., 2008), it is rare for them to be embedded as the resolving tool in MSW information management systems. The reason for this, as mentioned in previous sections, is the fact that the processing time for optimization modeling is often prohibitively long and this impedes its application in reality. An enumeration technique developed in previous work (Lin and Chen, 2008) has been proven to be effective in signiﬁcantly reduc-ing the solution time and thus has been embedded in the proposed integrated system. The procedure involved in the implementation of the integrated system is described in later sections.
Fig. 1presents the ﬂowchart used in the determination of the locations of supplementary recycling points using the integrated
Fig. 1. Flowchart of the integrated system.
system. In the procedure, MSW recycling data are ﬁrst collected, which can then be utilized for producing GIS map-layers or for developing statistic functions of recycling performances. With the assistance of these statistics, decision makers can evaluate if sup-plementary recycling points are necessary for a speciﬁc district or for a speciﬁc recyclable material. Once the scenario is deter-mined, the decision makers can deﬁne the indicators/weights and then apply the optimization model to locate the regions in which a pressing need for supplementary recycling points exists. Decision makers can then determine the locations of the recycling points within the region via an interactive GIS interface. A report recording the locations of the recycling points can be generated after the deci-sion makers have determined all of the recycling points. A detailed description of each step is described in later sections.
2.1. MSW recycling data
The MSW recycling data required for the integrated system include the quantity of each recyclable material collected, the loca-tions and acceptable materials for the private vendor recycling points, and the population distribution of the focus area. These data can then be either transferred into GIS map-layers for presentation or linked to the database functions for the generation of recycling performance statistics.
2.2. GIS functions
The GIS functions of the integrated system provide intuitive and clear presentations. Information concerning a recycling point can be sought using a web-based graphic interface. In addition to the map-layers generated by the MSW recycling data, the background GIS map-layers, such as streets and landmarks, are also included in the integrated system to assist users to realize the surround-ing environment of the locations concerned.Fig. 2illustrates the GIS interface of the integrated system with recycling points repre-sented by different symbols depending on the stipulated category to which they belong. The user can move the mouse cursor to a
recy-Fig. 2. Query interface of recycling points of the integrated system.
cling point mark in order for an associated table, which includes detailed information concerning the recycling point, as indicated by the black arrow inFig. 2.
2.3. Recycling performance statistics
With the statistical charts of MSW recycling data, users can have a clear picture of recycling status for an administrative district or of a speciﬁc recyclable material. In addition to creating routine reports for ofﬁcial records, three statistical graphs are also made available: trends, districts and category analyses. The trend analysis utilizes the periodical data and linear regression to provide a regression line which represents the recycling amount during a time period in the area of interest. The district analysis provides a bar chart illustrat-ing the recycled quantities for different administrative districts. The category analysis provides pie charts representing the contribution ratio of the recycling efforts of each political district.Fig. 3(a)–(c) illustrates the trend analysis, districts analysis and category anal-ysis, respectively. Customized charts can be made by setting up three options including time periods, recyclable material types and administrative districts to assist decision makers in the identiﬁ-cation of recyclable materials or administrative districts requiring additional attention.
2.4. Indicators and weights
Indicators and weights reﬂect the preferences for the appli-cation of the optimization model to identify the regions in need of supplementary recycling points. Once the decision makers determine to add supplementary recycling points for a speciﬁc material/district, different strategies may be employed to install the recycling facilities. In a previous study, Lin and Chen (2008) used three indicators to analyze the various demands of local authorities. To simplify the illustration of the integrated system, only the pop-ulation loading (PL) indicator is implemented in this work, which is deﬁned as the total number of recycling points over the total population in the analyzed region. A low PL value indicates that res-idents in this region share few recycling facilities, pointing toward
insufﬁcient recycling storage that may discourage residential par-ticipation. Customized indicators and weights can also be deﬁned via the interface of the integrated system, should these be desired. 2.5. Optimization model
The optimization model is a mixed integer programming (MIP) model, which was originally designated for landﬁll siting (Lin and Kao, 2005) and has been improved to ﬁnd regions in need of recy-cling facilities (Lin and Chen, 2008). The objective function of the improved model is to ﬁnd the region with the minimal value of PL indicator. To mitigate the computational efforts required for solv-ing a MIP model and to embed the improved optimization model in the integrated system, the improved model has been coded by C++ (Bell Labs, 1979) in an enumeration manner and a friendly interface to implement the parameters of the improved model is also devel-oped in this study, which not only improve the time burden to solve the improved model but also avoid the shortcoming that users have to be expertized in mathematical programming to implement the model.
2.6. Regions in need of recycling points
The priority list of the regions needing recycling points can be generated after applying the optimization model and this deter-mines the order of the installation of supplementary recycling points. That is to say, the ﬁrst region on the list has the mini-mum number of recycling points per capita among all the possible regions. To avoid the integrated system assigning supplementary recycling points to regions with very few people, a parameter setting identifying the minimum population of a region is also pro-vided. After the application of the optimization model, resolved regions have to pass through this population check prior to being recorded on the priority list. In addition, the number of regions on the list should be deﬁned by the user; its default value in the inte-grated system is equal to the number of allowable supplementary recycling points (ASRPs), as described in the next section.
618 H.-Y. Lin et al. / Resources, Conservation and Recycling 54 (2010) 615–622
2.7. Locations of supplementary recycling points
To assign the supplementary recycling points, the user has ﬁrst to deﬁne the number of ASRP, which may depend on budget,
Fig. 3. Illustrations of statistic functions of the integrated system. (a) Trend
anal-ysis: PET bottle recycled in Taichung City; (b) district analanal-ysis: quantities of paper recycled among different districts of Taichung City; (c) category analysis: ratios of four recyclable material among four different districts.
Fig. 4. Steps in the assignment of supplementary recycling collection points.
recycling containers in stock or the number of regions requir-ing recyclrequir-ing points. In addition, to ensure cost-effectiveness, a parameter called “minimal acceptable level” (MAL), essentially the minimal acceptable PL value of a region, is also provided. Only regions with lower PL values than the MAL are considered in need of supplementary recycling points. If the PL values of all regions are higher than the MAL, the integrated system will stop assigning the supplementary recycling points and report the remaining number of ASRP to the user. The default value of the MAL is the average value of PL in the entire study area.
Fig. 4presents the steps in the integrated system used to deter-mine the locations of supplementary recycling points. A region in the priority list will be selected in series each time. If a region’s PL value is less than the MAL and the number of the ASRP is still posi-tive, a supplementary recycling point will be allocated to the region and the number of ASRP will be reduced by one. In general, this pro-cedure will not be terminated until each region in the priority list is checked. However, if the number of ASRP is less than the number of regions in the priority list, some regions with higher PL values will not gain any supplementary recycling points. On the contrary, if the number of ASRP is more than the number of regions on the prior-ity list and if the MAL is greater than all PL values of those regions on the priority list, each region on the list will be assigned a sup-plementary recycling point. The integrated system will resolve the modiﬁed problem after supplementary recycling points are added in order to ﬁnd other regions requiring supplementary recycling points.
Essentially, the integrated system will automatically allocate a supplementary recycling point in the center of a region. The decision makers can manually change the location of the supple-mentary recycling point in the region. To realize the improvement generated by supplementary recycling points, an improvement indicator is deﬁned as the difference between the original and the subsequent minimal recycling points per capita of a region. After all the locations of supplementary recycling points have been con-ﬁrmed by the user, the integrated system will evaluate the value of the improvement indicator. Finally, a report containing supple-mentary recycling points will be generated with the location of
each new recycling point and the expected improvement after the installation of the new supplementary recycling points.
3. Case study
In order to demonstrate the applicability of the integrated sys-tem, a case study is presented: Taichung City, the third largest metropolis in Taiwan with an area of approximately 163 square kilometers and a population of more than one million inhabitants. In 2003, the total amount of recyclable materials collected was around 88,000 tonnes, about 33% of the total MSW (EPARC, 2006) from Taichung City. There were 1573 private recycling points in Taichung City collecting different recyclable materials. The proce-dure to illustrate the integrated system can be divided into ﬁve steps. First, the recyclable materials and the districts which require more recycling points are speciﬁed. Second, the parameters for the application of the integrated system have to be determined. Third, these selected districts are analyzed to ﬁnd regions in pressing need of recycling points. Thereafter, the locations of recycling points in these regions are provided and veriﬁed by the user. The ﬁnal step is to repeat the third and fourth steps if needed. For illustration, the recycling collection points for battery are analyzed. The steps in applying the integrated system to the recycling collection points of batteries are discussed as follows.
3.1. Step 1: ﬁnding the district requiring extra recycling points The total amount of batteries sold in Taiwan in 2005 was about 9946 tonnes while the recycled amount was 2177 tonnes in the same year, that is, 21.89% of the total sold (Chen et al., 2008). Estimating the amount of batteries consumed in Taichung City based on population ratio, the batteries consumed in Taichung City would be 452.7 tonnes, while the recycled amount was about 198 tonnes (EPARC, 2006). These waste batteries, including dry-cell batteries, cordless phone batteries, camcorder batteries and button batteries, usually contain high concentrations of heavy metals which are hazardous to public health and the environ-ment. To improve the collection ratio of batteries in Taichung City, the statistical charts for different districts are examined via
the integrated system.Fig. 5presents the battery recycling col-lection point information of different districts in Taichung City, including the total number of recycling collection points, pop-ulations and average recycling points per thousand capita in each district. It shows that Beitun and North Districts have the lowest recycling collection points per capita. Beitun and North Districts have an area of approximately 70 square kilometers and a population of about 350,000 inhabitants. The average recy-cling points for batteries in these two districts are 0.77 and 0.92 per thousand capita, respectively, signiﬁcantly less than 1.31 of the average value of all of Taichung City. Therefore, Beitun and North Districts have been selected for further analysis.Fig. 6 illus-trates the population distributions and the recycling points of these districts. The darker grids represent denser populated area (Fig. 6(a)) and the recycling collection points are indicated by dots (Fig. 7(b)).
3.2. Step 2: deﬁning the parameters of the integrated system To apply the optimization model and the embedded program to ﬁnd the regions in need of recycling points, several parameters have to be determined in advance, as illustrated inFig. 7. A cell in the map-layer is deﬁned by 50 m× 50 m. Therefore, the minimal area of a region is deﬁned by 122,500 square meters (49 cells). The width/length ratio to ensure the compactness of a region is deﬁned by 0.7. With this setting, the length of a short side of a solved region is never less than 70% of the length of its long side. The number of ASRP and the number of solved regions per run are deﬁned by 260 and 20, respectively. The minimal population of a region to install a supplementary recycling point is 1000. The MAL of PL is deﬁned by 0.002, indicating that one recycling point is shared by 500 persons. 3.3. Step 3: applying the optimization model to resolve regions
After the application of the optimization model, the integrated system lists twenty regions which are in need of recycling points. These regions are then assigned a supplementary recycling point, as described in the next step.Fig. 8presents the regions which are addressed by the model after the ﬁrst run. All of the regions
620 H.-Y. Lin et al. / Resources, Conservation and Recycling 54 (2010) 615–622
Fig. 6. (a) Population distributions of Beitun and North Districts. (b) Recycling collection points of Beitun and North Districts.
Fig. 8. Solved regions identiﬁed after utilizing the integrated system.
are in the left area where most residents dwell. Therefore, it is recommended that the supplementary points be installed in these regions.
3.4. Step 4: determining the locations of the recycling points After the regions are addressed in the former step, the supple-mentary recycling points are then assigned to these regions.Fig. 9 illustrates the procedure used to add new recycling points. The loca-tion of the supplementary recycling point by default is in the center of the region, marked as a star, while other existing recycling points are represented by unﬁlled circles. The user can click the “modify the point” button to alter the default location of a supplementary recycling point inside a region, if so desired. Thereafter, the inte-grated system evaluates the minimal PL value among all the regions in the Beitun and North Districts after adding supplementary recy-cling points.
3.5. Step 5: repeating Steps 3 and 4 until the procedure is terminated
Steps 3 and 4 are repeated until the procedure is terminated. The conditions for the termination of the analysis include: the ASRP is spent out or all the regional PL values are better than the MAL, as
Fig. 9. Interface to modify supplementary recycling point inside a solved region.
Fig. 10. Results of each model analysis run.
presented inFig. 4. In the case of Taichung City, there is a total 13 runs of model analysis, and the procedure is not terminated until the ASRP is fully assigned.
4. Results and discussions
The objective of the integrated system is to assist the user in the installation of supplementary recycling depots. After ﬁnding the regions, the suggested recycling point locations are also identiﬁed. The modiﬁed locations of the new recycling points can be taken into consideration along with former recycling points in subsequent optimization analyses in order to ﬁnd regions in most need of recy-cling points. This interactive process should continue to improve the accessibility for residents in the study area. Since the objec-tive function value of the optimization model is the minimum PL value of any region in the study area, the objective function value will gradually improve. In another words, the amount of recycling points per thousand capita (PL value) of a region increases dur-ing the analysis process.Fig. 10presents the results of each run, including population sizes and the PL values of the existing regions in most need of recycling points, which supports the demonstra-tion aforemendemonstra-tioned. Obviously, the PL values are ﬁxed at zero until the seventh run (140 pts). The populations of these regions range from 6677 to 1002, and represent a monotonic decrease trend. In this stage, the objective function values (the minimum PL value of any region) were null, which indicates that no recycling point existed in these regions. The integrated system assigned a higher priority to regions with greater populations. After the seventh run, the objective function values improved after the assignment of the number of supplementary recycling points proportionally and ﬁnally achieved 0.367 points per thousand capita with 260 supple-mentary recycling points attained. By contrast, the population of the regions in this stage varied, unlike the trend of the former 140 points. In this stage (after 140 pts), the objective function values gradually improve and ﬁnally achieve a value of 0.367 points per thousand capita. The result also reveals that the recycling collection points in Beitun and the Northern Districts are very insufﬁcient. In order to avoid the situation in which a region with more than 1000 persons does not have a recycling collection point, 140 sup-plementary recycling collection points need to be installed. While further improvement of the PL values may depend on the budgets or other concerns of local authorities, the results establish a clear relationship between the number of ASRP and improved PL values.
This work presents an innovative optimization model and system designed to assist local authority seeking to install
supple-622 H.-Y. Lin et al. / Resources, Conservation and Recycling 54 (2010) 615–622
mentary recycling points while taking into consideration of private take-back collection points. The optimization model was developed in order to identify regions requiring additional recycling collection points. For local authority responsible for allocating supplementary recycling facilities, the integrated system offers a level of ﬂexibility which is superior to most typical methods of identifying exact loca-tions. Everyday contingencies, the unwillingness of landowners and the unsuitability for facility installation can adversely affect the viability of selecting ideal locations. The user-friendly GIS interface and embedded rapid response programs for the proposed optimiza-tion model can be operated effortlessly without requiring expertise and experience of optimization techniques; this encourages local authorities to employ the system and, thus, to improve the practi-cal use of the optimization model. Several parameters concerning installation strategies for different supplementary recycling collec-tion points are also provided, which can satisfy a wide range of user demands. Due to the swift response time of the embedded pro-gram, local authorities can utilize the integrated system to modify the supplementary recycling points plan periodically in order to accommodate variations in demand over time.
The authors would like to thank the National Science Council of Taiwan, the Republic of China, for ﬁnancially supporting this research under Contract No. NSC 94-2211-E-324-004.
Chaerul M, Tanaka M, Shekdar AV. Resolving complexities in healthcare waste man-agement: a goal programming approach. Waste Manage Res 2008;26:217–32. Chang NB, Wei YL. Siting recycling drop-off stations in urban area by genetic
algorithm-based fuzzy multiobjective nonlinear integer programming model-ing. Fuzzy Set Syst 2000;114:133–49.
Chang NB, Chang YH, Chen YL. Cost-effective and equitable workload operation in solid-waste management systems. J Environ Eng 1997;123:178–90.
Chang NB, Davila E, Dyson B, Brown R. Optimal design for sustainable development of a material recovery facility in a fast-growing urban setting. Waste Manage 2005;25:833–46.
Chang NB, Parvathinathan G, Breeden JB. Combining GIS with fuzzy multicriteria decision-making for landﬁll siting in a fast-growing urban region. J Environ Manage 2008;87:139–53.
Chen YY, Chang SL, Lin LT. The review and the evaluation of the Effectiveness for collection systems of Waste batteries and waste lamps (in Chinese), EPA-96-HA14-03-A134. Taiwan: Environmental Protection Agency; 2008. pp. 2-1. Clarke MJ, Maantay JA. Optimizing recycling in all of New York City’s
neighbor-hoods: using GIS to develop the REAP index for improved recycling education, awareness, and participation. Resour Conserv Recycl 2006;46:128–48. EPARC (Environmental Protection Agency, Republic of China). Yearbook of
envi-ronmental protection statistics (in Chinese). Taiwan: Envienvi-ronmental Protection Agency; 2006.
Gautam A, Kumar S. Strategic planning of recycling options by multi-objective programming in a GIS environment. Clean Technol Environ Policy 2005;7: 306–16.
González-Torre PL, Adenso-Díaz B. Inﬂuence of distance on the motivation and fre-quency of household recycling. Waste Manage 2005;25:15–23.
He L, Huang G, Zeng G, Lu H. Fuzzy inexact mixed-integer semiinﬁnite pro-gramming for municipal solid waste management planning. J Environ Eng 2008;134:572–81.
Kao JJ, Lin TI. Shortest service location model for planning waste pickup locations. J Air Waste Manage 2002;52:585–92.
Kao JJ, Chen WY, Lin HY, Guo SJ. Network expert geographic information system for landﬁll siting. J Comput Civil Eng 1996;10:307–17.
Koa AS, Chang NB. Optimal planning of co-ﬁring alternative fuels with coal in a power plant by grey nonlinear mixed integer programming model. J Environ Manage 2008;88:11–27.
Lin HY, Chen GH. Regional optimization model for locating supplemental recycling depots. Waste Manage 2009;29:1473–9.
Lin HY, Kao JJ. Grid-based heuristic method for multifactor landﬁll siting. J Comput Civil Eng 2005;19:369–76.
McDonald S, Ball R. Public participation in plastics recycling schemes. Resour Con-serv Recycl 1998;22:123–41.
Sahoo S, Kim S, Kim BI, Kraas B, Popov A. Routing optimization for waste manage-ment. Interfaces 2005;35:24–36.
Shih LH, Lin YT. Optimal routing for infectious waste collection. J Environ Eng 1999;125:479–84.
Tilman C, Sandhu R. A model recycling program for Alabama. Resour Conserv Recycl 1998;24:183–90.
Vaillancourt K, Waaub JP. Environmental site evaluation of waste management facil-ities embedded into EUGENE model: a multicriteria approach. Eur J Oper Res 2002;139:436–48.