On using deterministic models to design agent-based, robotic systems

Abstract

The ever growing complexity of multi-robot systems and of robot themselves result in an increasing degree of uncertainty and unpredictability. In this work, the multi-agent approach to design such systems is adapted so behaviors of agents are described in terms of abstract state machines. By constraining agent behaviors to deterministic machines, verification of resulting systems is expected to be more effective and, as a by-product, implementation of physical agents is more straightforward. In this paper, the design framework is described as well as some preliminary tests to prove the viability of this approach.