Cellular Automata and non-static image processing for embodied robot systems on a massively parallel processor array

Abstract

A massively parallel processor array which combines image sensing and processing is utilized for the implementation of a simple Cellular Automaton. This automaton is essential part of an image processing task supporting object detection in real-time for an autonomous robot system. Experiments are presented, which demonstrate that objects will be detected only if they move below a specific velocity in the visual scene. Based on these experiments we will discuss the role of configurations changes if a Cellular Automaton is seen as a parallel processing computer. This leads us to the conclusion that if CA are performing non-static data processing tasks they might be better approached as sensor-driven parameterized dynamical system rather than as parallel computers operating on initial configurations only.