Simulation-based assessment of operational ridesharing strategies for shared autonomous vehicles in large-scale networks

Dynamic ridesharing (DRS) is envisioned as a sustainable mobility solution, with shared automated vehicles (SAVs) expected to facilitate DRS operations in the near future. Existing DRS-SAV-related literature typically focuses either on optimising fleet operations or simulating SAVs employing heuristic DRS methods, which may lead to an incorrect estimation of their true impacts. In this study, we design a time-based framework that enables communication between the DRS service provider and a mesoscopic traffic simulator. We incorporate individual socio-demographic features to generate future demand for SAVs and conduct two sets of simulation experiments within a detailed, large-scale urban network. First, we compare advanced DRS operation strategies during the morning peak. Then, we test the most effective strategy over an extended simulation period to replicate full business day operations. The results indicate that optimisation-based matching combined with fairness-aware pricing outperforms heuristic strategies in improving ridesharing performance and alleviating network congestion. Furthermore, the extended simulations reveal that profit-oriented objectives, which are common among transport network companies, result in significantly more empty VKT. To encourage the adoption of ridesharing, transport authorities should consider penalising (e.g., taxing) solo trips or subsidising shared trips.