Multi-agent Planning and Scheduling Environment for Enhanced Spacecraft Autonomy

Spacecraft autonomy has the potential for effecting significant cost savings in mission operations by reducing the need for dedicated ground staff. In an autonomous operating mode, operators will communicate only highlevel goals and deadlines directly to the spacecraft. The spacecraft will then perform its own planning and scheduling, decomposing a goal into a set of sub-goals to be achieved with onboard subsystems andlor in cooperation with other spacecraft in the environment. In this paper, we present this distributed (or equivalently, multi-agent) approach to onboard planning and scheduling that helps a spacecraft function as an autonomous agent. Such an agent's domain knowledge of tasks and their components is manifested through a hierarchical language taking into account spacecraft operational aspects and resource constraints. The task decentralization problem is solved by the use of the hierarchical knowledge structures, and the resource optimization problem is addressed by its explicit representation within the model. The reasoning performed by an agent for the required planning and scheduling tasks is based on a constraint propagation paradigm. Schedule quality is enhanced by the introduction of agent cooperation. A limited-scope Java prototype is developed and demonstrated using space-based scenarios involving onboard sensors and a satellite constellation. We are specifically targeting our effort to enhance the planning and scheduling capability of NASA's proposed Remote Agent architecture.