Multi-sensor data fusion for helicopter guidance using neuro-fuzzy estimation algorithms

The main objective of this paper is to present some algorithms to fuse information about obstacles, whose dynamics are a-priori unknown, in a helicopter's environment, provided by multiple spatially separate sensors. The fused information can then be used to help helicopters locate obstacles in hazardous conditions so that it can avoid them. Obstacle track estimation has been commonly carried out using the Kalman filter (KF), a linear estimator, or one of its variations. The extended Kalman filter, one such variation designed for use on non-linear problems, produces the best linear approximation to the object track. However certain assumptions made in the derivation of these algorithms render them suboptimal for aerial obstacle track estimation. Work produced by the University of Southampton has highlighted a link between fuzzy networks and associative memory neural networks. This link is important as it allows new learning rules to be developed for training fuzzy rules, and the conditions under which convergence can be proved to be derived. This paper explores methods for the fusion of estimates using these neurofuzzy models, and also addresses some of the weaknesses of the Kalman filter approximation introduced by the assumptions made in its derivation.