Algorithms for fusion of multiple sensors having unknown error distributions
Nageswara S. V. Rao · University of North Texas Digital Library (University of North Texas) · 1997
The authors presented recent results on a general sensor fusion problem, where the underlying sensor error distributions are not known, but a sample is available. They presented a general method for obtaining a fusion rule based on scale-sensitive dimension of the function class. Two computationally viable methods are reviewed based on the Nadaraya-Watson estimator, and the finite dimensional vector spaces. Several computational issues of the fusion rule estimation are open problems. It would be interesting to obtain necessary and sufficient conditions under which polynomial-time algorithms can be used to solve the fusion rule estimation problem under the criterion. Also, conditions under which the composite system is significantly better than best sensor would be extremely useful. Finally, lower bound estimates for various sample sizes will be very important in judging the optimality of sample size estimates.