Optimal-weights sensors-measurement fusion using genetic algorithms

Although there are many simulated Genetic Algorithms (GAs) in applications, less research has been directed toward their practical hardware implementations. In this paper we present a GA for optimum sensors-measurement fusion and characteristic weights. A multilevel verification of the GA is performed via the Mentor Graphics Design Architect (DA) and ModelSim CAD tools. In particular the design of efficient universal multipliers, dividers, and their integrated circuits is addressed. Effective mutation and crossover approach has been implemented in the GA system operation. It requires 960 clock cycles for complete iteration of 64 chromosomes, each with 3 genes of two binary-bits. This requires only 12 /spl mu/sec when implemented in the 0.25 /spl mu/m CMOS technology. The GA system is developed for a preprocessing unit to select optimal weights from real-time sensors measurement, and for fused measurements as in an electronic nose, integrated accelerometer systems, and for performance enhancement of recurrent dynamic neural networks in noisy environments. The proposed approach, simulation results, and possible experimental results are presented.