Random Weighting, Strong Tracking, and Unscented Kalman Filter for Soft Tissue Characterization

Area of Science:

• Biomedical Engineering
• Robotics
• Control Systems

Background:

• Accurate soft tissue characterization is crucial for medical simulations and robotic surgery.
• Existing nonlinear filtering methods, like the unscented Kalman filter, suffer performance degradation due to contact model errors.
• Online estimation of tissue properties requires robust algorithms that can handle uncertainties.

Purpose of the Study:

• To develop a new nonlinear filtering method for online soft tissue characterization that overcomes the limitations of existing filters regarding contact model errors.
• To enhance the accuracy and robustness of soft tissue parameter estimation in robotic indentation systems.
• To validate the proposed method through simulations and experimental studies.

Main Methods:

• A novel nonlinear filtering approach based on the Hunt-Crossley model is proposed.
• Mahalanobis distance is utilized to detect contact model errors.
• A scaling factor is incorporated into the predicted state covariance to compensate for identified model errors, determined via innovation orthogonality.
• Random weighting is employed to improve innovation covariance estimation accuracy, avoiding Jacobian matrix computation.

Main Results:

• The proposed method effectively identifies and compensates for contact model errors.
• The nonlinear filtering technique demonstrates improved estimation accuracy for soft tissue parameters.
• Validation through a master-slave robotic indentation system confirms the method's efficacy in simulation and experiments.

Conclusions:

• The developed nonlinear filtering method provides a robust solution for online soft tissue characterization, even with contact model uncertainties.
• The approach enhances the reliability of parameter estimation in robotic-assisted medical procedures.
• This work contributes to advancing the field of soft robotics and biomedical modeling.