Collaborative Bus-Taxi Scheduling for Large Airports: Time-Space Model and an Improved Reinforcement Learning-Guided Genetic Algorithm

Large hub airports often handle large volumes of passengers, which leads to crowding and longtime waiting for ground transfer service. To enhance service quality and shorten passenger wait durations, we propose an Incentive-based Collaborative Bus-Taxi (ICBT) transportation mode. In this mode, passengers are first transported via buses to the transfer points in the city, followed by taxi services to their final destinations, aiming to achieve efficient door-to-door service. We develop a space-time network-based optimization model to minimize the total travel costs, by optimizing bus paths and passenger matching schemes with taxis and transfer points. Furthermore, considering the time-sensitive nature of airport transportation scheduling, an adaptive genetic algorithm guided by deep reinforcement learning is designed to provide high-quality solutions effectively. Experimental results demonstrate the superior performance of our transportation model and the effectiveness of the proposed algorithm, showing potential to improve service performance for airport passengers.