Online model estimation and haptic characterization for robotic-assisted minimally invasive surgery

Shin, J 2017, Online model estimation and haptic characterization for robotic-assisted minimally invasive surgery, Doctor of Philosophy (PhD), Aerospace, Mechanical and Manufacturing Engineering, RMIT University.

Document type: Thesis
Collection: Theses

Attached Files
Name Description MIMEType Size
Shin.pdf Thesis application/pdf 1.78MB
Title Online model estimation and haptic characterization for robotic-assisted minimally invasive surgery
Author(s) Shin, J
Year 2017
Abstract Online soft tissue characterization is important for robotic-assisted minimally invasive surgery (RAMIS) to achieve a precise and stable robotic control with haptic feedback. The traditional linear regression method (i.e. the recursive least square (RLS) method) is inappropriate to handle nonlinear Hunt-Crossley (H-C) model since its linearization process involves unacceptable errors. This thesis presents a new nonlinear estimation method for online soft tissue characterization. To deal with nonlinear and dynamic conditions involved in soft tissue characterization, the approach expands the nonlinearity and dynamics of the H-C model by treating parameter p as an independent variable. Based on this, an unscented Kalman filter (UKF) was adapted for online nonlinear soft tissue characterization. A comparison analysis of the UKF and RLS methods was conducted to validate the performance of the UKF-based method.

The UKF-based method suffers from two major problems. The first one is that it requires prior noise statistics of the corresponding system to be precisely known. However, due to uncertainties in the dynamic environment of RAMIS, it is difficult to accurately describe noise characteristics. This leads to biased or even divergent UKF solutions. Therefore, in order to attain accurate estimation results from the UKF-based approach, it is necessary to estimate noise statistics online to restrain the disturbance of noise uncertainty. Secondly, the UKF performance depends on the pre-defined system and measurement models. If the models involve stochastic errors, the UKF-based solution will be unstable. In fact, the measurement model’s accuracy can be guaranteed by using high-precision measurement equipment together with a high volume of available measurement data. On the other hand, the system model is more often involved with the inaccuracy problem. In RAMIS, the system model is a theoretical approximation of the physical contact between robotic tool and biological soft tissue. The approximation is intended to fulfil the requirement of real-time performance in RAMIS. Therefore, it is essential to improve the UKF performance in the presence of system model (the contact model) uncertainty.

To address the UKF problem for inaccurate noise statistics, this thesis further presents a new recursive adaptive UKF (RAUKF) method for online nonlinear soft tissue characterization. It was developed, based on the H-C model, to estimate system noise statistics in real-time with windowing approximation. The method was developed under the condition that system noises are of small variation. In order to account for the inherent relationship between the current and previous states of soft tissue deformation involved in RAMIS, a recursive formulation was further constructed by introducing a fading scaling factor. This factor was further modified to accommodate noise statistics of a large variation, which may be caused by rupture events or geometric discontinuities in RAMIS. Simulations and comparison analyses verified the performance of the proposed RAUKF.

The second UKF limitation regarding the requirement of the accurate system model was also addressed. A random weighting strong tracking unscented Kalman filter (RWSTUKF) was developed based on the Hunt-Crossley model for online nonlinear soft tissue characterization. This RWSTUKF overcomes the problem of performance degradation in the UKF due to system model errors. It adopts a scaling factor in the predicted state covariance to compensate the inaccuracy of the system model. This scaling factor was derived by combining the orthogonality principle with the random weighting concept to prevent the cumbersome computation from Jacobian matrix and offer the reliable estimation for innovation covariances. Simulation and comparison analyses demonstrated that the proposed RWSTUKF can characterise soft tissue parameters in the presence of system model error for RAMIS in on online mode.

Using the proposed methods, a master-slave robotic system has been developed with a nonlinear state observer for soft tissue characterization. Robotic indentation and needle insertion tests conducted to evaluate performances of the proposed methods. Further, a rupture detection approach was established based on the RWSTUKF. It was also integrated into the master-slave robotic system to detect rupture events occurred during needle insertion. The experiment results demonstrated that the RWSTUKF outperforms RLS, UKF and RAUKF for soft tissue characterization.
Degree Doctor of Philosophy (PhD)
Institution RMIT University
School, Department or Centre Aerospace, Mechanical and Manufacturing Engineering
Subjects Biomechanical Engineering
Mechanical Engineering not elsewhere classified
Numerical Analysis
Numerical Modelling and Mechanical Characterisation
Keyword(s) Soft tissues characterization
Hunt-Crossley model
Unscented Kalman filter
Adaptive Kalman filter
Robotic-assisted minimally invasive surgery
Needle insertion robot
Version Filter Type
Access Statistics: 177 Abstract Views, 102 File Downloads  -  Detailed Statistics
Created: Mon, 07 Aug 2017, 09:20:43 EST by Denise Paciocco
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us