Intelligent control and asset management: An event-based control road map

My students and I have participated in a seven-year effort to produce an advanced supervisory control system called ICAM, a system for Intelligent Control and Asset Management, with specific focus on petroleum industry applications. By design, however, ICAM was devised to be widely applicable in a variety of automation and manufacturing arenas. We adopted a multi-agent architecture to implement ICAM, and divided ICAM's functionality among the various agents to execute different specific tasks. Most agents were, by nature, computationally intensive; these were implemented as MATLAB^® routines. The Supervisor (or “master agent”), on the other hand, was meant to incorporate the “intelligence” of ICAM, so it was created in the G2 expert system shell. We also assumed that a wireless sensor and actuator network (WSAN) would be incorporated in the system under control, including links comprising control loops. Several important lessons were learned in the course of this project. First, we found it easier and more efficient to distribute much of the intelligence in the agents rather than the Supervisor, and secondly, that implementing the Supervisor as an expert system in G2 made the operation of ICAM very slow and cumbersome, due to the excessive overhead involved. We have concluded that replacing the expert system Supervisor with a top-level event-based controller will make ICAM considerably more effective, efficient, and easier to extend and maintain.