Probabilistic Reasoning for Enhanced Course of Action Generation (PRECOG)

ABSTRACT:Autonomous operations in space, particularly with respect to health and resilience, are becoming increasingly important features of satellite platforms. Systems that respond to threat indicators from sensor fusion and external sources have to generate and evaluate potentially large sets of COAs based on incomplete or noisy data. Such systems need to reason under uncertainty and anticipate possible contingencies for plans that involve complex sequencing and sensor tasking. The Probabilistic Reasoning for Enhanced Course of Action Generation (PRECOG) system uses advanced probabilistic reasoning methods to represent complex COAs and their evolution under uncertainty and to optimize the choice of COAs against multiple objectives. The representation of COAs includes the ability to task sensors and allows execution of actions to depend on the occurrence of contingent events or the results of sensor fusion. PRECOG will advance the state of the art in decision making algorithms, enabling efficient COA selection on board a satellite.BENEFIT:Satellite autonomy, which enables satellites to take action in response to threats without having to wait for ground control, is a growing field. PRECOG will support satellite autonomy by efficiently selecting an optimal COA in response to a situation, while reasoning about and anticipating uncertainties and contingencies. In addition, PRECOG is highly relevant to unmanned autonomous vehicles in general, all of which could benefit from the capability to rapidly respond to situations without needing human control.