Task and Motion Planning for Space Operations, with Human-Assisted Recovery

Space operations have been increasingly turning to Persistent Assets; systems that require repeated visits and/or regular maintenance over extended periods of time. To make such operations achievable from a cost perspective and dependence on human crews, relying on autonomous vehicles and robots is a common solution. However, deploying successful systems long-running space operations requires a combination of a) improved autonomous system capabilities and b) the ability for remote human operators to intervene when progress towards completing a task is blocked. Additionally, it is important to minimize the frequency of human intervention due to the cost of crewed operations (especially in situ), as well as communication delays over long distances.We propose a system that introduces Task and Motion Planning into our existing MoveIt Studio tool. This allows remote operators to provide high-level task specifications in structured text, and the autonomous system will produce and execute a sequence of actions to meet that specification. To deal with unanticipated errors with both the world model used for planning and during task execution, we will add a mechanism to autonomously identify failures and request manual interventions from remote operators. Our proposed system will improve the autonomy capabilities of robotic systems in space, while also reducing the cognitive load on remote operators by elevating their role in human-machine teams to higher-level goal specifications and task monitoring. nbsp;