A modified algorithm for the quadratic assignment problem using chaotic-neuro-dynamics for VLSI implementation

The quadratic assignment problem (QAP) is one of the NP-hard combinatorial optimization problems, which is very difficult to solve [1]. The tabu search technique, one of the various heuristic methods for the QAP, has been implemented in a neural network form with chaotic dynamics [2]-[4]. In order to achieve a high-speed massively parallel solution of the QAP, a mixed analog/digital integrated circuit implementation of the system is mandatory. However, the algorithm in [4] cannot be implemented with IC technology as it is. Therefore, the algorithm in [4] is modified to be suitable for the analog IC implementation in this paper. The performance of the modified algorithm is investigated with numerical simulations. The circuit characteristics obtained from a prototype chip is extensively used in the simulations.