Multiple Observer Adaptive Fusion for Uncertainty Estimation and Its Application to Wheel Velocity Systems.

Area of Science:

• Control Systems Engineering
• Robotics
• Signal Processing

Background:

• Real-world uncertainty estimation faces challenges from peaking phenomena and measurement noise.
• Existing methods struggle to effectively mitigate these complexities, impacting dynamic system performance.
• Accurate uncertainty estimation is crucial for robust control and reliable operation of dynamic systems.

Purpose of the Study:

• To introduce a novel scheme for practical uncertainty estimation that addresses peaking dynamics and measurement noise.
• To enhance the overall dynamic behavior and robustness of uncertainty estimation frameworks.
• To validate the proposed method through theoretical analysis, simulations, and experimental testing.

Main Methods:

• Development of a fusion estimation framework utilizing multiple extended state observers (ESOs) for lumped uncertainties.
• Implementation of a low-frequency adaptive parameter learning technique for optimal fusion approximation.
• Integration of cascading filters within the adaptive fusion framework to improve noise rejection.

Main Results:

• The adaptive fusion estimation effectively attenuates transient peaks in uncertainty estimation.
• The proposed method achieves fast convergence and high accuracy, particularly under high-gain scheduling of ESOs.
• Enhanced robustness against measurement noises is demonstrated through theoretical analysis and experimental validation.

Conclusions:

• The novel adaptive fusion estimation scheme provides a practical solution for uncertainty estimation challenges.
• The method significantly improves dynamic behavior, peak rejection, and noise robustness in complex systems.
• Validated on a mobile robot's wheel velocity system, demonstrating its feasibility and effectiveness.