High coverage point-to-point transit: Local vehicle routing problem with genetic algorithms

High Coverage Point-to-Point Transit (HCPPT) is a new design of alternative transportation, which involves a sufficient number of deployed small vehicles with advanced information supply schemes. This paper focuses on Genetic Algorithms (GA) to improve system performance with real-time re-optimization and compares GA with the existing insertion heuristics for local vehicle routing in HCPPT. Two genetic operation schemes, Best Feasible Position (BFP) and Random Feasible Position (RFP), are designed. Simulations are performed with different demand levels based on OCTA (Orange County Transportation Authority) trip demands. The results show that BFP significantly improves the local routing performance in terms of both system efficiency and productivity whereas RFP shows improvement only in system efficiency compared to the insertion heuristics. This study also provides results of computational performances as well as convergence performance. In terms of computational performance, both GA approaches show viability in real-time operations.