Скачать с ютуб Collision Avoidance for Indoor Service Robots through Multimodal Deep Reinforcement Learning в хорошем качестве

Collision Avoidance for Indoor Service Robots through Multimodal Deep Reinforcement Learning

Abstract. In this paper, we propose an end-to-end approach to endowindoor service robots with the ability to avoid collisions using Deep Reinforcement Learning (DRL). The proposed method allows a controller toderive continuous velocity commands for an omnidirectional mobile robotusing depth images, laser measurements, and odometry based speed estimations. The controller is parameterized by a deep neural network,and trained using DDPG. To improve the limited perceptual range ofmost indoor robots, a method to exploit range measurements throughsensor integration and feature extraction is developed. Additionally, toalleviate the reality gap problem due to training in simulations, a simple processing pipeline for depth images is proposed.As a case studywe consider indoor collision avoidance using the Pepper robot. Throughsimulated testing we show that our approach is able to learn a proficient collision avoidance policy from scratch. Furthermore, we showempirically the generalization capabilities of the trained policy by testing it in challenging real-world environments. Videos showing the behavior of agents trained using the proposed method can be found at   • Collision Avoidance for Indoor Servic...