Design and implementation of shared autonomy protocols for unmanned ground vehicles

点击翻译按钮获取中文摘要

Building autonomous robots has been a central objective of research in robotics. Despite substantial advancement of autonomous navigation system, existing solutions are still not fully trusted for safety-critical situations (e.g. operating in close proximity of sensitive equipment or humans), due to both the complexity of desired tasks and lack of safety guarantees. Neither purely manual manipulation nor fully autonomous control can lead to satisfactory performance for robotic systems. Consider the situation when a mobile robot equipped with robotics arm is used to read sensor gauges and manipulate valves in a petroleum plant. In this case, the robot might be required to move to a specific location where the gauge is invisible in the camera, and the arm might be manually manipulated since valves come in different shape and or that they might be in positions difficult to detect by the robot. On the other hand, purely manual teleoperation of the robot is cumbersome to the operators as they need to constantly monitor the position of the robot in the environment using various sensor feeds, e.g., multiple camera feeds and pointcloud data. Low communication quality between the robot and the operation center (e.g., the high latency of the network or possible dropouts) makes it impossible to safely teleoperate the robot in the presence of highly dynamic objects. This project aims to design, implement, and test control protocols for unmanned ground vehicles (UGVs) to achieve to shared autonomy with human operators while satisfying safe performance guarantees. More specifically, the shared autonomy protocols will help (1) perform safe and optimal autonomous control when receiving no human commands, (2) filter out or modify unsafe human operations that caused by human errors such as avoiding collisions with surrounding objects, and (3) minimally intervene human operations.