[ICROS] DQN-based Collision Avoidance using Low-cost Proximity Sensors

Abstract In this paper, we propose a deep Q-network (DQN)-based agent that successfully avoids collisions using only low-cost proximity sensors. One of the previous works on collision avoidance using deep reinforcement learning requires expensive sensors such as a LiDAR or a depth camera.

However, these expensive sensors were not only cost-competitive, but also led to the use of overqualified se.....