Urban Autonomous Driving of Emergency Vehicles with Reinforcement Learning

The development of autonomous driving technologies has primarily focused on passenger vehicles, leaving behind the potential benefits for emergency vehicles like fire trucks and ambulances. This paper investigates reinforcement learning techniques’ application in driving emergency vehicles in complex urban scenarios. We demonstrate our approach’s adaptability and efficacy in various driving challenges, vehicle types, and changing surroundings by employing advanced learning algorithms, such as Soft Actor-Critic, in the CARLA simulator. Our preliminary research highlights the possibilities of using reinforcement learning methods to improve self-driving features in emergency vehicles, emphasizing the importance of further research to tackle the unique problems of these emergency vehicles.