Towards a Hybrid-ASP Planning Approach With Adjoint Observation for Incomplete Task-Relevant Information

Abstract

In the real world, robot task plans may easily become invalid due to unexpected state dynamics, preventing the robot from accessing the complete task-relevant information. The possible occurrence of information incompleteness during robot plan execution expects the robot to sense the environment and obtain the missing information actively. A set of planning and reasoning approaches has focused on robotic tasks in incomplete domains. However, the sensing action operators are generally abstracted as passively receiving sensory information and formulating a finite set of observations. The strict constraint of solely sensing without actively acting on the environment greatly reduces the chances of revealing more hidden information. This paper proposes a novel observation plan that couples an atomic sensing action with parallel actuating effects, and an adjoint observation scheme for actively run-time information gathering. To concretely realize the adjoint observation scheme, we present a Hybrid-ASP planning approach that integrates a deterministic task planner for generating an initially valid task plan for task achievement and a logic-based Answer Set Planner to make observation plans. Compared with the baseline planning approaches available, the experimental results have validated improved planning efficiency and plan effectiveness of information gathering from the Hybrid-ASP approach.