A low-cost vision system for online reciprocal collision avoidance with UAVs

Abstract

In this article, we propose a reciprocal collision avoidance system for autonomous drones, based on computer vision and using relative positioning in an indoor environment. This dynamic environment represents a demanding challenge, but it is crucial for any future existence of multiple drones operating in urban areas. We use commercial AR Drone 2.0 robots, which represent that our proposal is suitable for low-cost equipment. In our case, we attempt to achieve the collision avoidance of two drones that fly one towards the other and react online autonomously to signals received by their computer vision systems with a decentralized control strategy. We test this in four different experiments with demanding conditions. For this purpose, we get the camera signal of the onboard drones and tune their behavior to react smoothly and precisely. We report encouraging positive results and provide the code we use in the experiments for replication.

Julián Estévez

Artificial Intelligence and Robotics Professor

My research interests include distributed robotics, mobile computing and programmable matter.