Experience-Based Planning Domains: an Integrated Learning and Deliberation Approach for Intelligent Robots

摘要

Deliberation and learning are required to endow a robot with the capabilities for acquiring knowledge, performing a variety of tasks and interactions, and adapting to open-ended environments. This paper presents the notion of experience-based planning domains (EBPDs) for task level learning and planning in robotics. EBPDs provide methods for a robot to: (i) obtain robot activity experiences from the robot's performance in a dynamic environment; (ii) conceptualize each experience producing an activity schema; and (iii) exploit the learned activity schemata to make plans in similar situations. Experiences are episodic descriptions of plan-based robot activities including environment perception, sequences of applied actions and achieved tasks. The conceptualization approach integrates different techniques including deductive generalization, abstraction, goal inference and feature extraction. A high-level task planner was developed to find a solution for a task by following an activity schema. The proposed approach is illustrated and evaluated in a restaurant environment where a service robot learns how to carry out complex tasks.