A Hybrid Parallel-Sequential Insertion Heuristics for Shared Autonomous Vehicle Routing: A Melbourne Case Study

With technological advancements, shared autonomous vehicles (SAVs) could emerge as a dominant mode of transportation in the future. The promise of SAVs in reshaping urban transportation is contingent upon developing effective and adaptive ride-matching algorithms. To take a step forward, this study develops a hybrid insertion heuristic, which uses parallel and sequential insertions to match the riders with the vehicles. This proposed algorithm offers a global perspective by finding the least cost request-vehicle pairs, developing a queue based on the costs, and serving them sequentially based on that queue. The study uses the MATSim model of the Melbourne Metropolitan Area to create the SAV demand and the network for Inner Melbourne to evaluate the strategy for the created demand. Results showcase significantly improved ride-matching efficiency compared to the base sequential algorithm. Results show that the hybrid algorithm outperforms the base sequential algorithm regarding service quality, average wait, travel times, and VKT and EKT per ride. The results show that the hybrid algorithm performs significantly better for smaller fleet sizes and slightly improves the parameters for larger fleet sizes.