Gathering and Conceptualizing Plan-Based Robot Activity Experiences

Abstract

Learning from experiences is an effective approach to enhance robot's competence. This paper focuses on developing capabilities for a robot to obtain robot activity experiences and conceptualize the experiences as plan schemata, which are used as heuristics for the robot to make plans in similar situations. The plan-based robot activity experiences are obtained through human-robot interactions where a teaching action from a command-line user interface triggers recording of an experience. To represent human-robot interaction activities, ontologies for experiences and user instructions are integrated into a robot ontology. Recorded experiences are episodic descriptions of the robot's activities including relevant perceptions of the environment, the goals pursued, successes, and failures. Since the amount of experience data is large, a graph simplification algorithm based on ego networks is investigated to filter out irrelevant information in an experience. Finally, an approach to robot activity conceptualization based on deductive generalization and abstraction is presented. The proposed system was demonstrated in a scenario where a PR2 robot is taught how to "serve a coffee" to a guest, in the EU project RACE.

A semantic-centered cloud control framework for autonomous unmanned system

“…Miguelanez et al [14] first introduced semantic knowledge information into mission planning, which improved the autonomy level of autonomous underwater unmanned vehicles. Ekvall et al and Kragic [15,16] proposed a task-level planning system based on experiential activity schema, which realized task planning by configuring task goals and constraints. Mokhtari et al [17] proposesd a conceptional method for autonomous robots that conceptualized task experience into the activity mode and applied it to task planning.…”

Learning the Scope of Applicability for Task Planning Knowledge in Experience-Based Planning Domains

“…Experiences are collected through human-robot interaction and instruction-based teaching. We previously presented methods and approaches of instructing and teaching a robot how to achieve a task as well as extracting and recording experiences [22,23]. Experience extraction is beyond the scope of this paper.…”

“…Planning is a hierarchical problem solver consisting of an abstract and a concrete planner which applies learned activity schemata for problem solving. In previous work, algorithms have been developed for experience extraction [22,23], activity schema learning and planning [23,25,26].…”

“…An approach for teaching a robot to achieve a task and extracting experiences has been presented in[Mokhtari et al 2016a;2016b].…”

The RACE Project

“…For more robust and flexible future robot task plans, an approach was developed to support the extraction and conceptualization of robot activity experiences [12]. After applying temporal segmentation heuristics, the experience data (a set of occurrences) is filtered using a graph simplification method based on ego networks [15].…”