Mobile Robot Motion Planners Using Depth-Difference Images for Multiple Obstacles
Satoshi Hoshino, Yu Kubota · The Proceedings of JSME annual Conference on Robotics and Mechatronics (Robomec) · 2024
For autonomous navigation, mobile robots are required to avoid obstacles ahead. For obstacle avoidance, we have proposed motion planners based on deep neural networks, such as CNNs with an LSTM block. Depth-difference images generated from two depth images enabled a robot based on the motion planner to avoid a static or dynamic obstacle by planning different motions. However, a single obstacle was assumed as the target for collision avoidane. In this paper, therefore, we focus on motion planning for multiple obstacles. Since the robot is allowed to have depth images, the nearest obstacle is determined as the target. The motion planner is trained through imitation learning for a single obstacle. Nevertheless, the navigation experiments show that the robot based on the motion planner is able to move toward a destination autonomously while avoiding multiple obstacles.