Safe trajectory design for indoor drones using reinforcement-learning-based methods

Abstract

This paper proposes a design method for achieving safe trajectory of indoor drones. The trajectory design with a reinforcement-learning-based (RL) agent is facilitated, which can result in efficient and collision-free motion. The method is developed for motion in indoor area with moving mobile robots, and thus, the collision with these obstacles must be avoided. Through RL-based design the fast motion of the drones can be achieved, which must perform a mission between workstations in a manufacturing system. The effectiveness of the design process through a simulation example on a real laboratory environment is illustrated.