A Framework for Failure Management and Recovery for Remote Autonomous Task Planning and Execution

Advanced robotic systems capable of near-human level autonomy will be required for NASA to maintain critical, human-rated spaceflight hardware in cislunar space and beyond. Occasionally, these systems may experience task planning failures, or detectable faults while executing a planned procedure to accomplish a necessary task such as transferring cargo or engaging levers. In order for a ground operator to effectively diagnose, mitigate, and respond to these failures, the system must provide comprehensive tooling that allows a remote user to asynchronously analyze the situation, identify solutions, and manually adjust the task specification to command the in situ robot to recover and ultimately complete the objective.We propose a novel system for fault detection and reporting that will augment a remote operatorrsquo;s ability to diagnose and resolve failures from an in situ robot during real-time task planning and execution. For this effort, PickNik will develop software tools that enable a robotic system to report exacting details into the nature of a reported failure, as well as user interfaces that allow a remote operator to asynchronously assist in recovering the system to complete the desired task. We will demonstrate these features on previously developed, NASA relevant autonomous tasks for pressing buttons, opening hatches, and transferring cargo. The developed features will be made available in MoveIt Studio, PickNik#39;s flagship product for supervised autonomy. nbsp;