Collision Avoidance Control for Connected Drones in Air-Intersections : Extended Abstract

In the past decade, the interest in automatic drones has constantly led to increasing the number of drones and made air traffic management prohibitive, especially in urban environments. The safe operation and optimized navigation of between drones have become more and more concerning issues. Therefore, drones rely on performing advanced collision avoidance strategies to achieve safety flying in crowded urban airspace. In this article, we proposed a collision avoidance control model by using reinforcement learning which data gathered from connected drones. Generally speaking, the acceleration/deceleration of a drone after a collision has been detecting, would allow ego-drone to pass first or last. Therefore, the proposed connected-drones-based reinforcement learning model would be obtaining optimal action policy based on the different collision scenarios. In future work, we plan to make drones better navigate urban areas and test them in natural environments.