Multi-objective optimization model of autonomous minibus considering passenger arrival reliability and travel risk

The advancement of self-driving technologies facilitates the emergence of autonomous minibuses (ABs) in public transportation, which could provide flexible, reliable, and safe mobility services. This study develops an AB routing and scheduling model considering each passenger’s arrival reliability and travel risk. Firstly, to guarantee each passenger’s arrival on time, the arrival reliability (a predetermined threshold of on-time arrival probability of α ​= ​0.9) is included in the constraints. Secondly, three objectives, including system costs, greenhouse gas (GHG) emissions, and travel risk, are optimized in the model. To assess the travel risk of ABs, an enhanced method based on kernel density estimation (KDE) is proposed. Thirdly, an advanced multi-objective adaptive large neighborhood search algorithm (MOALNS) is designed to find the Pareto optimal set. Finally, experiments are conducted in Shanghai to validate model performance. Results show that it can decrease GHG emissions (−2.12%) and risk (−9.47%), while only increasing costs by 2.02%. Furthermore, the proposed arrival reliability constraint can improve an average of 14.70% of passengers to meet their arrival reliability requirement ( α ​= ​0.9). • A multi-objective optimization model of autonomous minibuses is designed, minimizing system costs, emissions, and risks. • Passengers' arrival reliability is constrained to be earlier than their latest arrival times with a high probability. • An improved kernel density estimation method is proposed to evaluate and quantify travel risks of autonomous minibuses. • The optimization model integrates AB scheduling, routing, ride-matching, and timetabling in a joint framework. • The proposed model mitigates the travel risk (−9.47%) and GHG emissions (−2.12%), compared with the cost-minimized model.