Distributed architecture for control and path planning of autonomous vehicles

This paper presents a new distributed architecture designed and built in or der to execute, among other functions, on-board movement control and path p lanning of a mobile robot by using Digital Signal Processors (DSP) and Neur onChip(R) Microcontrollers. Different functions have been asigned to indepe ndent processor modules interconected through a Serial Communications Link that uses the LonWorks(R) Protocol. Each module acts as a 'node' in a data network, were commands and control data flow as short messages increasing t he system flexibility. High level tasks can be developed by DSP nodes, whil e low level and communication tasks are carried out by the NeuronChip. An a pplication to Path-planning with Obstacle Avoidance using this architecture is described in detail: a DSP node, using appropriate sensory information, creates a model of the environment detecting obstacles, then it computes a path to follow and executes a full tracking algorithm, including position estimation and velocities commands generation; two NeuronChip nodes are res ponsible of motor control and state feedback, accepting velocities commands , computing a digital closed loop algorithm and giving back sensed velociti es and encoders sampling. (C) 2001 Elsevier Science B.V. All rights reserve d.