Joint State and Unknown Input Estimation for a Class of Artificial Neural Networks With Sensor Resolution: An...

Area of Science:

• Control Systems Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Real-world sensor resolution (SR) impacts system accuracy.
• Encoding-decoding mechanisms are crucial for communication networks with limited bandwidth.
• Accurate state and unknown input (SUI) estimation is vital for artificial neural networks (ANNs).

Purpose of the Study:

• To develop a set-membership estimation algorithm for ANNs.
• To address joint SUI estimation under SR and encoding-decoding constraints.
• To achieve accurate ANN state estimation unaffected by unknown inputs.

Main Methods:

• Derivation of a sufficient condition for ellipsoidal constraint on estimation error.
• Formulation and solution of an optimization problem for estimator gain design.
• Development of a set-membership estimation algorithm accounting for SR and encoding-decoding.

Main Results:

• An ellipsoidal constraint on the estimation error is guaranteed.
• Optimal estimator gains are designed to minimize the ellipsoidal constraint.
• The proposed algorithm provides accurate joint SUI estimation for ANNs.

Conclusions:

• The developed algorithm effectively performs joint SUI estimation for ANNs.
• The method is robust to sensor resolution limitations and encoding-decoding mechanisms.
• Validation through an example confirms the scheme's practical applicability.