Russell Knight, Gregg Rabideau, Barbara Engelhardt, and Steve Chien
In general, most approaches to robust autonomy with respect to planning and execution are focused on either providing models that allow for flexibility or providing techniques for changing models to improve performance. We take these techniques into consideration, but focus the majority of our work on robust autonomous planning and execution with imperfect models. We deal with this through a number of means: 1) providing fast replanning in the event of unforeseen events, 2) modeling execution context, and 3) adjusting the planning and scheduling system heuristics given a model of the uncertainty of the environment. We have demonstrated these techniques using the Continuous Activity Scheduling, Planning, Execution, and Replanning (CASPER) system.