The primary goal of this work is to demonstrate the realistic potential of advanced adaptive critic designs -- a general approach to intelligent control which may be crucial to understanding and replicating biological intelligence. There exists a domain of problems which require large-scale decentralized control of complex non-linear dynamics. The use of hydrogen as both fuel and coolant in hypersonic vehicles, such as the National Aerospace Plane (NASP), presents such a problem. The NASP will use hydrogen fuel to cool the engine and remove the excessive heat on the vehicle surface caused by high energy air flow during hypersonic flight. In the thermal management system, the hydrogen flow patterns and routing schemes for cooling such a vehicle incorporate multiple levels of tiered distribution. However, there exist severe weight penalties in a large scale system which requires and abundance of fuel and high pressure valves necessary to cool the most extreme temperatures which appear only in local areas or "hot spot". Unless coolant can be efficiently and actively controlled to cool these local areas, the weight penalties could prevent orbital capability. Since flow is regulated through the valved adjustment of inlet and exit pressures for multiple sets of panels connected in parallel and/or series, and decentralized control approach will provide a method to globally optimize coolant to a large-scale distributed system of panels and valves. A decentralized control architecture will be developed and tested on a representative small scale problem to globally optimize the arm trajectory of a three link manipulator. Future work will address issues involved in the hierarchical and heterarchical scaling of this decentalized control architecture and the architecture will be evaluated using the NASP thermal management system simulation.
