Radar Cross Section Modelling Using Genetic Algorithms


In the design of new, more sophisticated missile systems, simulations need to be realistic and fast. Realistic target models are just as important as realistic models of the missile, but have often been overlooked in the past. Existing methods for creating realistic target models require considerable computational resources. This thesis addresses the problem of using limited resources to create realistic target models for simulating engagements with radar guided homing missiles. A multiple genetic algorithm approach is presented for converting inverse synthetic aperture radar images of targets into scatterer models. The models produced are high fidelity and fast to process. Results are given that demonstrate the generation of a model from real data using a desktop computer. Realistic models are used to investigate the effects of target fidelity on the missile performance. The results of the investigation allow the model complexity to be traded against the fidelity of the representation to optimise simulation speed. Finally, a realistic target model is used in a feasibility study to investigate the potential use of glint for target manoeuvre detection. Target glint is considered as noise in conventional missile systems and filtered to reduce its effects on the tracking performance. The use of glint for target manoeuvre detection would provide a cheap and novel alternative to the optical techniques currently being developed. The feasibility study has shown that target manoeuvre detection using glint may be as fast as optical techniques and very reliable.