Dual adaptive dynamic control of mobile robots using neural networks

Area of Science:

• Robotics
• Control Systems
• Artificial Intelligence

Background:

• Nonholonomic mobile robots present complex control challenges due to their nonlinear dynamics.
• Accurate dynamic models are often unavailable, necessitating adaptive control strategies.
• Existing adaptive techniques may rely on simplifying assumptions like certainty equivalence.

Purpose of the Study:

• To propose two novel dual adaptive neural control schemes for discrete-time nonholonomic mobile robots.
• To address unknown nonlinear dynamic functions and estimate network parameters in real-time.
• To improve trajectory-tracking performance by accounting for parameter uncertainties.

Main Methods:

• Utilizing Gaussian radial basis function and sigmoidal multilayer perceptron neural networks for function approximation.
• Implementing stochastic real-time estimation of unknown neural network parameters without offline training.
• Developing dual control laws that explicitly consider uncertainty in parameter estimates, deviating from the certainty equivalence principle.

Main Results:

• Demonstrated significant improvements in trajectory-tracking performance for a differentially driven wheeled mobile robot.
• Validated the effectiveness of the proposed stochastic controllers under plant uncertainty and unmodeled dynamics.
• Utilized Monte Carlo simulations and statistical hypothesis testing for rigorous evaluation.

Conclusions:

• The proposed dual adaptive neural control schemes offer a robust solution for controlling nonholonomic mobile robots with unknown dynamics.
• The novel approach enhances tracking accuracy by directly addressing parameter estimation uncertainty.
• This work advances adaptive control methodologies in robotics, particularly for trajectory tracking applications.