Distributed Agents for Multi-Rover Autonomy

This paper describes the results of an evolution activity of a multi agent system (MAS) used in a Mars Exploration mission demonstrator. The objective of the activity was to incorporate and evaluate different agent-related assets, including the commanding of the system using high level goals and service petitions rather than low level commands, the use of a common ontology to ease the communication among the agents and their reasoning, the distribution of functionality such as the on-board planning, and the design of the agents behaviors using production rules. Although the implementation and demonstrator are focused towards space applications, the focus of the paper will be on the assets/lessons learned in designing and prototyping the MAS. The system has been implemented in JADE following the FIPA Standards.