Task Planner for Simultaneous Fulfillment of Operational, Geometric and Uncertainty-Reduction Goals

Abstract

Our ultimate goal in robot planning is to develop a planner which can create complete assembly plans given as input a high level description of assembly goals, geometric models of the components of the assembly, and a description of the capabilities of the work cell (including the robot and the sensory system). In this paper, we introduce SPAR, a planning system which reasons about high level operational goals, geometric goals and uncertainty-reduction goals in order to create assembly plans which consist of manipulations as well as sensory operations when appropriate. Operational planning is done using a nonlinear, constraint posting planner. Geometric planning is accomplished by constraining the execution of operations in the plan so that geometric goals are satisfied, or, if the geometric configuration of the world prevents this, by introducing new operations into the plan with the appropriate constraints. When the uncertainty in the world description exceeds that specified by the uncertainty-reduction goals, SPAR introduces either sensing operations or manipulations to reduce that uncertainty to acceptable levels. If SPAR cannot find a way to sufficiently reduce uncertainties, it does not abandon the plan. Instead, it augments the plan with sensing operations to be used to verify the execution of the action, and, when possible, posts possible error recovery plans, although at this point, the verification operations and recovery plans are predefined.