A massively parallel AI-based planning approach to process-control

Process-control systems form a critical component of embedded computer systems. Planning for such systems has always been looked at from the point of view of the optimization of decentralized control systems. The authors present a different approach for achieving high-performance planning for such control systems. They reduce a class of process-control systems into a general path planning problem commonly encountered in the field of AI and robotics. The path-planning problem is then solved by using an unconventional decentralized approach. This approach is based on a cellular automaton nearest neighbor wavefront propagation computing model. The massively parallel, decentralized algorithm for trajectory determination is discussed. The issue of mapping a process-control problem in its real-world space to a state space suitable for algorithmic and parallel computation is discussed. Details on the interface between the massively parallel system and the process-control system that is operating in the real world are presented. An example shows the versatility and applicability of this approach in solving control problems