Adaptive optimal output regulation for wheel-legged robot Ollie: A data-driven approach

Area of Science:

• Robotics
• Control Systems Engineering
• Artificial Intelligence

Background:

• Robot dynamics are susceptible to internal and external factors, leading to performance degradation.
• Model uncertainties and disturbances pose significant challenges for robot control systems.
• Existing controllers may struggle with real-world operational variations.

Purpose of the Study:

• To design and evaluate an adaptive optimal output regulation (AOOR)-based controller for the wheel-legged robot Ollie.
• To address model uncertainties and external disturbances in a data-driven manner.
• To enhance the robot's stability and control performance.

Main Methods:

• Development of a data-driven adaptive optimal output regulation (AOOR) controller.
• Online training using a small dataset for controller optimization.
• Experimental validation focusing on static balancing and dynamic object balancing tasks.
• Testing under various working conditions to assess robustness.

Main Results:

• The AOOR controller successfully stabilized the Ollie robot within small displacements.
• The controller demonstrated optimal control performance through online training.
• The robot achieved enhanced stability, including balancing a rolling bottle, which the initial controller could not.
• Experiments confirmed the controller's effectiveness and robustness against uncertainties and disturbances.

Conclusions:

• The AOOR-based controller offers a robust and effective solution for stabilizing the Ollie robot.
• This data-driven approach successfully mitigates the impact of model uncertainties and external disturbances.
• The controller significantly enhances the robot's balancing capabilities, even with added dynamic loads.