Robotic task and motion planning using distributed knowledge bases

The authors describe research into the construction of a control system using distributed knowledge bases which is able to perform generalised task and motion planning for robotic systems. Distributed knowledge bases offer a solution to the information bottleneck caused by the central data structures in blackboard systems. The distributed knowledge base system was designed to improve on the response time of a previous distributed blackboard system developed by the authors (R.S. Kay and R. Marriott, 1994). The knowledge bases both request and produce information. Any information useful to other parts of the system is distributed through their individual communications managers. The major aspects of the distributed knowledge base system are discussed including the control system, design considerations, and suitability for real time control. Future directions are discussed, including the authors´ plan to test the system using a fish packing scheme developed for the distributed blackboard system