Multi-objective Approaches in Transit Stop Location Problem

An efficient bus network design is a difficult problem, and hence is usually considered as a series of sub-problems solved sequentially. The bus stop location problem (BSLP) is foremost in this series of problems. The BSLP in this research is formulated as a mixed integer problem (MIP) that consideres three important design sub-problems: stop location, demand allocation and service frequency. Given a set of candidate bus stops, the level of trip connectivity and origindestination pair-wise trip demands, the problem is to answer the following related questions. How many bus stops should be required and where should they be located? Which users/demands are to be assigned to each bus stop? What should be the service frequency for each bus line while being feasible in fleet requirements? The BSLP is formulated as multiobjective approaches which address the tradeoffs between improving accessibility to transit service with more stops and increasing the efficiency of the transit system with fewer stops. Transit access is measured as the physical proximity to transit service assuming that additional stops can provide greater access to the service and reduce walking distance to stops. Trip connectivity is measured using a transit performance measure namely Transit Opportunity Index (TOI) for quantifying the ease of reaching a destination from a given location. TOI is formulated by integrating transit accessibility (spatial and temporal coverage) with topological network connections and travel time (trip coverage) to quantify public transit opportunity. The fewer the number of stops along a line can result passengers’ faster travel to destinations with less dwell time at stops. General Algebraic Modeling System (GAMS) software is used for modeling this problem and MIP CBC (COIN-OR Branch and Cut) solver is used to solve the problem for optimal solutions. Limitations on computational efficiency for resulting optimal solutions are identified. A heuristic (Randomized Feature Selection Algorithm) algorithm is developed to speed computation and provide near optimal solutions. Computation results for TOI, the BSLP model solutions for both the algorithms (i.e., Branch and Cut algorithm, and Randomized Feature Selection algorithm) are given for a bus network in City of New Haven, CT. The computational results of the heuristic algorithm and the CBC solver are provided to show the efficiency of the proposed heuristic algorithm. The randomized feature selection algorithm is shown to be very efficient in terms of time compared with the known optimal solutions generated by CBC solver