Optimal controller design for finite word length implementation using genetic learning algorithm

In this paper, a linear quadratic Gaussian (LQG) controller with genetic learning algorithm (GLA) is proposed to tackle the numerical errors due to the conversions of the A/D and D/A converters in a digital computer. This scheme can be directly used for the design of the ideal LQG and also is optimal in the presence of the numerical errors due to the finite word length. By converting the stochastic problem to a deterministic game theoretic one, we find the estimation states using GLA and controller can minimize a suitable performance measure. The GLA, via reproduction, crossover, and mutation procedures, is used to tackle the signals from ADC to reduce the numerical errors and to obtain their optimal values