Zhenshan Bing

Research on sparse reward-based reinforcement learning aims to better address the application of reinforcement learning to real robot tasks, which is a major challenge and scientific problem in the field of reinforcement learning. How to improve the training efficiency of reinforcement learning with sparse rewards, to integrate sparse reward reinforcement learning with real robot scenarios, and to use sparse reward reinforcement learning to solve complex robot grasping tasks are all pressing issues in this field. This presentation will briefly introduce the basic concepts and background of reinforcement learning, and focus on the idea of replaying hindsight experience as the basis for introducing a series of relevant algorithms for improving the efficiency of sparse reward reinforcement learning for real robot scenario problems nowadays, and summarize the future research directions of sparse reward reinforcement learning.

Biography:

Dr. Bing is a post-doc researcher at the Human Brain Project SP10 Neurorobotics research group at TUM. He received his PhD degree in computer science from the Technical University of Munich under the supervision of Prof. Alois Knoll in June 2019.

His research investigates the bio-inspired robot which is controlled to achieve autonomous locomotion and self-adaptive behavior with spiking neural networks and reinforcement learning and its related applications.