Low-level motion control and high-level decision-making algorithms must be integrated to enable full automation of complex dynamic systems. This project investigates knowledge representation and inference mechanisms for end-to-end plan construction in domains, such as spacecraft flight, where resource consumption is dependent both on computational processes and physical characteristics of each vehicle and its environment. This is an important and difficult problem that has been somewhat neglected by the robotics and AI communities. To address this challenge, the project aims to develop a Task and Motion Planning Architecture (TAMPA) and apply it to two realistic problem domains: ground transport and spacecraft mission design. TAMPA is composed of task and path planners, resource scheduler, and plan executor, bridging the long-standing gap between cognitive (symbolic) strategic-level and physics-based (continuous) tactical-level reasoning algorithms. Single and formation spacecraft missions will be designed in terms of science data quality and quantity, observation costs, and tradeoffs between algorithm complexity and solution optimality.
The education component of this award cuts across K-12 and higher education. This project plans to pursue engineering and K-12 curriculum enhancements to encourage computationally-aware students at all levels, working toward a basic understanding analogous to that achieved for math and science. Outreach to K-12 and undergraduate researchers will use hands-on robotic experiments to draw students of all ages into a second century of air and space operations in which autonomy will enable missions previously considered impossible.
