Sutapa Samanta

Multi Depot Probabilistic Vehicle Routing Problems with a Time Window: Theory, Solution and Application

Vehicle Routing Problems (VRPs) are prevalent in all large pick up and delivery logistics systems and are critical to city logistics operations. Of notable significance are three key extensions to classical VRPs: (1) multi-depot scenario; (2) probabilistic demand; and (3) time-window constraints, which are considered simultaneously with VRPs in this paper. The issue then becomes a Multi Depot Probabilistic Vehicle Routing Problem with a Time Window (MDPVRPTW). The underlying complexities of MDPVRPTW are analyzed and a heuristic approach is presented to solve the problem. Genetic algorithms (GAs) are found to be capable of providing an efficient solution to the so-called MDPVRPTW. Within the GAs, two modification operators namely, crossover and mutation, are designed specially to solve the MDPVRPTW. Three numerical examples with 14, 25, and 51 nodes are presented to test the efficiency of the algorithm as the problem size grows. The proposed algorithms perform satisfactorily and the limiting case solutions are in agreement with the constraints. Additional work is needed to test the robustness and efficiency of the algorithms as the problem size grows.

Identifying Feasible Locations for Rail Transit Stations: Two-Stage Analytical Model

A public transportation system is a viable alternative to reducing traffic congestion and environmental pollution in urban areas. A metro, subway, or light rail system may be a viable commuting alternative connected with a coordinated service of feeder buses in urban and suburban neighborhoods. The decision to build a rail transit system is largely driven by available land and feasible sites for tracks and stations. Factors like ridership and public perception are considered in identifying suitable rail corridor and station locations. A two-stage analytical model is developed for identifying feasible rail transit station sites based on the real geographical and demographic data. The model uses a genetic algorithm (GA) for optimally locating the stations and works in parallel with a geographical information system (GIS). The model is applied in an example by using real GIS data, road network, and demographic information. The potential station sites are identified in the first stage, and the optimization using the GA is performed in the second stage by minimizing the total cost of locating the stations.

Applicability of Genetic and Ant Algorithms in Highway Alignment and Rail Transit Station Location Optimization

The emergence of artificial intelligence (AI)-based optimization heuristics like genetic and ant algorithms is useful in solving many complex transportation location optimization problems. The suitability of such algorithms depends on the nature of the problem to be solved. This study examines the suitability of genetic and ant algorithms in two distinct and complex transportation problems: (1) highway alignment optimization and (2) rail transit station location optimization. A comparative study of the two algorithms is presented in terms of the quality of results. In addition, Ant algorithms (AAs) have been modified to search in a global space for both problems, a significant departure from traditional AA application in local search problems. It is observed that for the two optimization problems both algorithms give almost similar solutions. However, the ant algorithm has the inherent limitation of being effective only in discrete search problems. When applied to continuous search spaces ant algorithm requires the space to be sufficiently discretized. On the other hand, genetic algorithms can be applied to both discrete and continues spaces with reasonable confidence. The application of AA in global search seems promising and opens up the possibility of its application in other complex optimization problems.

A Bilevel Model for Optimizing Station Locations along a Rail Transit Line

The population growth and increase in number of commuters in urban areas has given rise to the need to either build or extend a rail transit system, which requires the: (1) optimization of the alignment of a rail transit line; and (2) optimization of station locations along rail transit lines. The paper describes how the positions of the stations depend on many factors, such as the total cost for locating them, proximity from the residential neighborhoods, feasibility studies, and environmental, and political factors. A bilevel programming approach is proposed in this paper, which minimizes the total cost in solving the station location problem. At the lower level, the potential ridership generated from the major cities or central business districts (CBDs) are estimated by dividing the study area in optimum number of zones, to maximize the usage by the potential riders. At the upper level, the number and location of intermediate stations are determined by minimizing the sum of user, operator, and location costs. The primary components of costs are the costs associated with traveling to the station and in-vehicle travel time, land-acquisition cost (also known as right-of-way cost), and cost of operating the train, construction of stations and parking facilities. The total cost is minimized using a Genetic Algorithm (GA). A Geographic Information System (GIS) is used in order to work directly with maps of the proposed rail-line, existing road networks and transit lines, and land and property boundaries. The population and passenger distribution in the study area as well as the travel times are obtained using a GIS which is integrated to the GA, to obtain the best station locations. The model is applied to decide on the optimum positions of stations along a transit corridor of an artificial case study. The results indicate that one can optimally locate station locations with improved precision if GIS data with sufficient accuracy were available.