Autonomous Landing of eVTOL Vehicles for Advanced Air Mobility via Deep Reinforcement Learning

The development of Advance Air Mobility (AAM) is revolutionizing air transportation by introducing autonomous aircraft, particularly electric vertical takeoff and landing (eVTOL) vehicles, for efficient movement of people and cargo in urban and remote areas. The majority of aircraft operating in the AAM system are expected to be eVTOL vehicles, which will utilize vertiports as their base of operations. However, unlike traditional airports and general aviation aircraft, the placement of vertiports within densely populated urban areas poses distinct challenges in effectively managing the air traffic flow associated with these facilities. The complexity of this challenge is amplified as the frequency of scheduled landings and take-offs increases. This paper focuses on the critical aspect of autonomous landing for eVTOLs in the context of AAM. Specifically, we propose a deep reinforcement learning (DRL) approach to enhance the reliability and safety of eVTOL autonomous landing, considering various algorithms and scenarios. By leveraging and customizing DRL algorithms for the eVTOL landing task, the accuracy and robustness of the landing operations are enhanced. To evaluate the effectiveness of our approach, extensive simulations are conducted, and the performance of our algorithm is analyzed under different scenarios and environmental conditions. We utilize the AirSim simulation environment integrated with Unreal Engine (UE), providing a realistic virtual platform for testing and validating the performance of our algorithm. Extensive simulations and evaluations conducted within the AirSim-UE environment provide compelling evidence of the exceptional landing precision and efficiency achieved by the proposed method.