Motion Planning and Control for Autonomous Robot Inspection

Abstract

Motion planning and control play crucial roles in enhancing the autonomy and dexterity of robots. This seminar presents a series of techniques for robot inspection tasks and then introduces three applications within this context. First, for quadrotors, we introduce a reinforcement learning motion planner designed for aggressive narrow gap traversing. The planned motion was successfully demonstrated on a real quadrotor. Second, the seminar presents planners for autonomous exploration, aiming to maximize information collection within unknown scenarios. This algorithm is extended to explore contact-rich scenes, enabling robots to navigate within cluttered environments. Third, the seminar introduces teleoperation that involves human control in the loop. The integration of virtual reality and haptic feedback enables bidirectional teleoperation that enhances the operator’s situational awareness. Lastly, the seminar discusses the prospective applications of these technologies to flight control applications, highlighting the potential for accomplishing complex flight tasks that were previously unattainable.

Speaker

Dr Chenxi Xiao earned his PhD degree from the School of Industrial Engineering at Purdue University (Purdue), advised by Prof. Juan Wachs. His doctoral research focused on enabling robots to comprehend the environments when visual sensing is limited or unavailable.  Before joining Purdue, he worked on reinforcement learning for flight control at The Hong Kong Polytechnic University, developed various embedded control systems at China Electronics Technology Group Corporation, and obtained his BS and MS degrees from the School of Automation at Northwestern Polytechnical University. Dr Xiao's research interests encompass robot motion planning and control, teleoperation, deep learning, and electronics.