Path Cost Optimization Using Genetic Algorithm with Supervised Crossover Operator

Path cost optimization is essential for maneuvering vehicles in a cost effective way. The term cost can be interpreted as fuel consumption, path visibility, probability of being detected, probability of being attacked or a combination of the above. Exact algorithms such as linear programming and dynamic programming can always provide globally optimum solution to such a problem. However, as the size and dimension of the search space increases, computational complexities of these algorithms rise drastically. Meta-heuristic algorithms such as evolutionary algorithms and genetic algorithms can provide optimum to sub-optimum solutions to large scale path cost optimization problems. Though meta-heuristic algorithms are capable to solve large scale path cost optimization problems, precautions are needed to avoid premature convergence to suboptimum solutions. In this paper, a genetic-based path cost optimization algorithm is proposed. The generic crossover operator in genetic algorithms is replaced by a supervised crossover operator which is based on the operation of dynamic programming. Simulation results show that the proposed crossover operator can greatly improve the convergence rate and solution quality of genetic algorithms.