Attitude Feedforward Neural Controller in Quaternion Algebra

L. Fortuna,G. Muscato,M. G. Xibilia
DOI: https://doi.org/10.1080/10798587.1999.10750760
1999-01-01
Abstract:In the paper, in order to deal with the attitude control problem of a rigid body in a 3-D space, a new control strategy in hypercomplex algebra is developed. The proposed approach is based on two parallel controllers derived in quaternion algebra. The first one is a feedback controller of PD type, while the second is a feed-forward controller implemented by means of an hypercomplex multilayer perceptron (HMLP) neural network. Quaternion algebra allows to simplify the computational complexity of the controllers and leads to a more efficient learning algorithm for the neural network. Several simulations and comparisons with other control strategies show the suitability of the proposed approach.
automation & control systems,computer science, artificial intelligence
What problem does this paper attempt to address?