Knowledge-Based System research group
Osnabrück University/DFKI Robotics Innovation Center, Germany
Title: Continual hybrid hierarchical planning for robot control
Abstract: Continual planning has been described as a framework for making plan-based robots life-long autonomous and blurring appropriately the classical distinction between a plan generation and a plan execution phase. Robot plans, as others would argue, have to be hybrid in that causal, spatial, temporal, and resource aspects of acting are inseparably intertwined in many robot tasks. Finally, hierarchical planning appears to make sense in robot planning for structuring the planning search space, and for communicating plans to humans and/or fellow robots.
The talk reports about an approach towards such continual hybrid hierarchical planning. It is based on the planner CHIMP, which heavily builds upon the Meta-CSP system for representing and reasoning about plans in terms of constraints.