A Generalization of the CHOMP Algorithm for UAV Collision-Free Trajectory Generation in Unknown Dynamic Environments
Jiayu Men, Jesús Requena Carrión · 2020
Unmanned Aerial Vehicles (UAVs) are being used in an increasing number of application areas. In many applications, UAVs are expected to operate in unknown environments with dynamic obstacles, hence the growing interest in autonomous navigation systems. In this work, we propose a method to generate collision-free trajectories in an unknown environment with dynamic obstacles, by extending the usual notion of collision cost penalty into Dynamic Obstacle Cost, calculated using Dynamic Cost Map. The Dynamic Cost Map allows to identify future potential collisions by modelling the movements of the dynamic obstacles. We compare the performances of our method with conventional CHOMP and A* with polynomial smoothing in a simulation environment. Our results show that the times of replanning are drastically reduced with a limited increase in the computing time associated to each replanning task.