Research Of Aircraft Route Planning And Evaluation Algorithm

As one of the critical components of Mission Planning System, Route Planning is an interdisciplinary research which has developed with the developing of information obtaining and information processing. Because the circumstance of aircraft is very complicated, and restricting conditions are excessive, Route Planning algorithm becomes the most challenging problem in Mission Planning System. Route evaluation is a multi-factor decision making problem. It takes all related factors into condition, and produces an objective and accurate evaluation to the route.Existing route planning methods include optimization planning algorithm, deterministic planning algorithm, and random planning algorithm. These algorithms have some defects such as combination exploding danger and convergence time uncertainty. Multi-factor decision-making algorithms include SAW, AHP, TOPSIS, and so on. The core of these algorithms is judgment matrix, and they are obviously affected by subjectivity of decision maker. This paper improved adaptability by adjusting the current algorithm, and established a model to evaluate the routes by introducing artificial neural network.First of all, according to the problem of route, an arithmetic model was established, which described the problem in four aspects as space model, route expression, restricting conditions and planning algorithms.Next, this paper studied the planning method of overall aircraft route, and worked over the principle and character of A* algorithm, and three-dimensional sparse A* algorithm. By analysis of time complexity, it found the bottleneck of the algorithm, and brought up an improved method of weighted estimating function. Experiments indicated that the new method can greatly speed up the algorithm.Finally, this paper applied the artificial neural network to route evaluation. It established route evaluation indicator system, presented normalization method and built a model for the problem according to BP Networks. Then it analyzed the learning method LMBP which was founded on numerical optimization. The learning method has fast convergence, so it adapts to the circumstance when weighting value of networks is small. Experiments show that the established model can correctly evaluate the routes.