The goal of this research is to develop a system controller based upon neural computing techniques that can be efficiently implemented using currently available digital VLSI technology. The controller will be capable of continuously adapting to changes in its environment by using a learning algorithm and associative memory as the basis for its operation. This research explores the use of digital associative memory coupled with fuzzy logic matching techniques in the design of such an adaptive associative processor. In contrast to a conventional computer, the associative processor will not operate in a deterministic fashion. Instead, its operation will be probabilistic and based upon a given set of heuristics. By using a closest-fit data matching unit, it will attempt to find a relatively optimal solution via associative and iterative techniques. A scoring unit will monitor its operation in order to provide feedback to the system. Although we anticipate implementing this processor using digital VLSI technology, investigation using computer simulations is a first step. A biped walker paradigm is used as a tool for developing the associative processor and learning algorithms. As a result of this investigation, we intend to build such a biped walker robot controlled by the VLSI associative processor.
