Modular Soft Computing Approach For Aircraft Carrier Landing Trajectory Prediction

A modular learning design for classifying aircraft flight data in time-series prediction is proposed in this paper. This is part of the decision support system to assist landing signal officers in guiding aircraft to land on aircraft carriers. NeuroFuzzy systems are used to emulate the flight patterns for future real-time flight prediction. To improve the learning efficiency, a two stage modular learning design is proposed. The data to be learned is first decomposed into categories in accordance to their physical structure. Each module of data is presented to a different NeuroFuzzy system for learning purpose. Individually trained modules are modeled as genetic chromosomes. Genetic algorithm is used to produce a chromosome module that represents a generalization of all the trained modules. As compared with the non-modular approach, the modular approach offers comparable prediction performance with significantly lower overall computation time. We show that the reduction in computation time with the modular approach is exponential as the problem size increases. Navy aircraft data were used to validate the effectiveness of the modular design and the result is consistent and promising.