Integration of a multi-agent system into a robotic framework : a case study of a cooperative fault diagnosis application

Abstract

Programming multi-robot autonomous systems can be extremely complex without appropriate software development techniques to abstract hardware faults, as well as can be hard to deal with the complexity of software required the coordinated autonomous behavior. Real environments are dynamic and unexpected events may occur, leading a robot to unforeseen situations or even fault situations. This work presents a method of integration of Jason multi-agent system into ROS robotic framework. Through this integration, can be easier to describe complex missions by using Jason agent language and its resources, as well as abstracting hardware details from the decision-taken process. Moreover, software modules related to the hardware control and modules which have a high CPU cost are separated from the planning and decision-taken process in software layers, allowing plan and software modules reuse in different missions and robots. Through this integration, Jason resources such as plans reconsideration and contingency plans can be used in a way where they can enable the robot to reconsider its actions and strategies in order to reach its goals or to take actions to deal with unforeseen situations due the environment unpredictability or even some robot hardware fault. The presented integration method also allows the cooperation between multiple robots through a standardized language of communication between agents. The proposed method is validated by a case study applied in real robots where a robot can detect a fault in its hardware and diagnose it through the help of another robot, in a highly abstract method of cooperative diagnosis.