Virtual Explanation Reasoning Agent (VERA)

On-board autonomy will be critical to the success of manned deep-space missions. Deep space missions will not have the same level of contact with ground-based mission control, and so the astronauts themselves will have to deal with novel situations that were impossible to anticipate and prepare for prior to launch. This fact requires a successful autonomy solution to be able to improvise and synthesize novel solutions to novel problems from pre-existing procedures and available resources in absence of ground-based support. SoarTech, teamed with Tietronix, are developing the Virtual Explanation Reasoning Agent (VERA), an AI-based software system that provides automated procedure identification and generation to address novel situations for which no procedures have been written ahead of time. VERA captures expert knowledge of spacecraft systems and diagnostic procedures from detailed SysML spacecraft models to identify situation-specific diagnostic requirements and identify or author appropriate and viable diagnostic procedures. To improve procedure generation, VERA can learn from interactions with astronauts. On-board autonomy will be critical to the success of manned deep-space missions. Deep space missions will not have the same level of contact with ground-based mission control, and so the astronauts themselves will have to deal with novel situations that were impossible to anticipate and prepare for prior to launch. This requires a successful autonomy solution be able to improvise and synthesize novel solutions to novel problems from pre-existing procedures and available resources in absence of ground-based support. SoarTech is developing the Virtual Explanation Reasoning Agent (VERA), an AI-based software system that provides automated procedure identification and generation to address novel situations for which no procedures have been written ahead of time. VERA captures expert knowledge of spacecraft systems and diagnostic procedures from detailed SysML spacecraft models to identify situation-specific diagnostic requirements and identify or author appropriate and viable diagnostic procedures. To improve procedure generation, VERA can learn from interactions with astronauts. Objective 1: Design and prototype an initial spacecraft procedure generation capability Objective 2: Design and prototype an initial learning capability Objective 3: Design a failure mode modelling approach and prototype models and their outputs Objective 4: Evaluate VERA prototype
Benefits: For NASA, VERA would help astronauts in deep space missions when contact with Mission Command is infeasible or too slow (e.g., 40-minute round trip communication from Mars). Additionally, any fully autonomous probes or stations (e.g., Gateway when no one is on board) could benefit from VERA. For the DoD, similar situations arise, such as repairing components on a submarine or repairing vehicles while in combat when there is no time to call an expert. Commercially, we would pursue the austere environment repair and maintenance market, initially in defense, and then later in commercial spaces such as manufacturing and automotive.