Robot Position Local Map Light Stripe Odometry Robot Position Robot Position Light Stripe Robot Position Robot Position Movement ‘ RALPH ’ Robot Camera Position

This paper discusses a parallel architecture designed for autonomous mobile robot navigation tasks. Parallelism is used to enhance system speed and developmental exibility. The system-called MADRON-consists of ve independent parallel agents sharing common data and control information. They are a communication agent for interfacing the system with a `Ralph' class autonomous mobile robot, a vision agent for performing robot vision functions, an environmental mod-eling agent (which generates an incremental global environment map), a localisation agent and a control agent for higher level machine intelligence. This paper further concentrates on the localisation problem of robot navigation and presents a new localisation technique based on fusing vision and odometry information to provide robot positioning in a priori unknown environments. The method uses light-stripe ranging to extract geometric features from the robot's environment and compares multiple views to determine position. The fusion of odometry information increases the accuracy of the method.