(9) Design of Multi-Agent type Robot Language/System

This paper outlines a parallel logic programming approach to implementing cooperative tasks for multiple robots. In this approach, an extended Horn clause logic can be used as a specification language for explicitly describing negotiation processes between robots, and the distinctive feature of the approach is supporting correctness and justification of multi-robot programming. This language provides concurrency and synchronization based on a parallel logic language KL1. We incorporate KL1 into several facilities in which multi-agent task planning is employed. First, we incorporate look ahead facility for handling emergent situation; thus reactive planning can be implemented within a logical framework. Second, scene analysis is integrated within multi-agent robot programming environment where dynamics in real-world can be recognized using stream-oriented computation. After recognizing a "change" of the world, new tasks are produced and performed by negotiating robotic agents. Third, inductive learning is introduced to generate control rules that are used for solving conflict resolution and deadlock detection on cooperative tasks. These rules effectively selected robot actions for improving multi-robot performance. We illustrate these facilities through typical cooperative tasks such as picking and placing distributed blocks.