Basic-evolutive algorithms for neural networks architecture configuration and training [spacecraft control]

This paper presents a procedure for optimising a neural network architecture used in a system for spacecraft attitude and position determination. The procedure establishes the neural network structure and the training algorithm. A new version of Basic-Evolutive algorithm is presented, Basic-Evolutive 1 and Basic-Evolutive 2 algorithms are capable of setting the appropriate dimension of the neural network and the adequate weights interconnecting the neurons. The results produced by both versions are tested with a very wide set of different spacecraft manoeuvre simulations. The algorithm performance is contrasted with backpropagation training algorithm performances. The capability of the resulting neural network architecture for generalising is also verified.

[1]  Andrew Luk,et al.  Pruning via number of links and neuron activities , 1993, 1993 IEEE International Symposium on Circuits and Systems.

[2]  Yih-Fang Huang,et al.  Bounds on the number of hidden neurons in multilayer perceptrons , 1991, IEEE Trans. Neural Networks.

[3]  S. P. McGlynn,et al.  The Geometry of Genetics , 1989 .

[4]  Alberto Tesi,et al.  On the Problem of Local Minima in Backpropagation , 1992, IEEE Trans. Pattern Anal. Mach. Intell..

[5]  J. Seijas,et al.  Evolutive algorithms pioneering neural networks training for artificial vision applications , 1994, Proceedings of 1994 IEEE International Conference on Neural Networks (ICNN'94).

[6]  David B. Fogel,et al.  An introduction to simulated evolutionary optimization , 1994, IEEE Trans. Neural Networks.

[7]  D. E. Goldberg,et al.  Genetic Algorithms in Search , 1989 .

[8]  John J. Shynk,et al.  Convergence properties and stationary points of a perceptron learning algorithm , 1990 .