Abstract

The Constraint Satisfaction Problem (CSP) has been a useful model for various industrial and engineering applications. These include image processing and pattern recognition, VLSI engineering, robotics manipulation, and computer hardware design automation. In this paper, we give a novel AI architecture for discrete relaxation that effectively prunes a backtracking search tree in CSP. This algorithm has been implemented on a fine-grained, massively parallel hardware computer architecture. For practical application problems, many orders of magnitude of efficiency improvement can be reached on this hardware architecture. This enables real-time processing of a large class of practical problems.