Adaptive Learning Control of Uncertain Systems via Weight and Intrinsic Plasticity-Based Neural Networks.

Area of Science:

• Robotics
• Control Systems Engineering
• Computational Neuroscience

Background:

• Artificial neural networks (ANNs) are widely used for nonlinear system control.
• Traditional ANN control often overlooks neuronal intrinsic plasticity (IP) and assumes bounded inputs without theoretical basis.
• These overlooked factors complicate ANN-based function approximation and control design.

Purpose of the Study:

• To propose a novel intrinsic plasticity neural network (IP-NN)-associated control method for nonlinear systems.
• To address the challenges posed by IP mechanisms and input boundedness in ANN control.
• To develop a theoretically sound and practically verifiable control strategy.

Main Methods:

• Developed an alternative mathematical model for ANNs incorporating both weight and IP plasticity, inspired by biological nervous systems.
• Utilized the mean value theorem to convert nonaffine functions with IP parameters into an approximately affine form.
• Employed virtual parameter estimation to bypass the need for detailed IP parameters.
• Proposed a barrier certificate based on Lyapunov stability theory to ensure bounded NN inputs.

Main Results:

• The proposed method effectively incorporates neuronal intrinsic plasticity into ANN control design.
• The mean value theorem and virtual parameter estimation allow for practical control implementation without explicit IP parameters.
• The barrier certificate guarantees the boundedness of NN inputs, ensuring system stability.
• Experimental validation on a 3-DoF robotic manipulator demonstrated the benefits of the IP-NN-based control.

Conclusions:

• The novel IP-NN control method offers a robust approach for nonlinear systems.
• The integration of IP and input boundedness enhances the reliability and performance of ANN-based controllers.
• This work provides a theoretical framework and practical validation for advanced ANN control strategies.