Recursive parameter estimation algorithms for nonlinear output error model with quantized observations

In this study, the parameter estimation for a nonlinear system that consists of dynamic nonlinear output error model whose parameters are unknown and a static known arbitrary quantization function is considered. The difficulty occurs when the output of the unknown output-error model is unmeasurable and the system output is quantized. Two estimation algorithms are proposed to deal with the problem. The first one introduces a weighted innovation into the recursive algorithm that different weights yield different accuracies. The second algorithm is based on the recursive prediction error algorithm that an improved variable gradient prediction error algorithm is proposed. It can be seen from the simulation example that the proposed algorithm is more capable of tracking if the appropriate weighting factor is selected and has a higher estimation accuracy compared to the conventional algorithms.