Multi-agent Software Architecture for Autonomous Robots: a Practical Approach

Received: 11 October 2010 Abstract Accepted: 25 November 2010 This paper analyses the issues of designing software for an autonomous robot, which is perceived as a real-time, multi-stage data processing system. We focus on a functional analysis of the data-driven world-modelling and motion planning tasks performed by the robot, and then we develop a multi-agent software architecture, which guarantees flexible use of sensors and world representations. We demonstrate use of this architecture on two practical examples: an autonomous indoor mobile robot, which can be applied for janitorial services or as an AGV in flexible manufacturing, and on a walking robot, which is intended for reconnaissance and search missions, e.g. in hostile or polluted industrial environments.