Learning Control Policies for Tooth Extraction

Learning Control Policies for Tooth Extraction: An Application of Programming by Demonstration

Zhang, X.

Kober, J. (mentor)

Mechanical, Maritime and Materials Engineering

Delft Center for Systems and Control

System and Control

2016-11-25

: Tooth extraction is one of the most common procedures in oral surgery. However, it is complicated as well as not fully understood. Different teeth have different structures and strengths. Even teeth of the same type differ depending on ages, genders, races, and other reasons affecting its integrity. During the procedure, the properties of the reaction force from the tooth is changing due to broken fibers and the operations are different between individuals. The wrench and pose in the human demonstrations are collected by a recently introduced measurement setup. The purpose of the analysis is to obtain a good model for the reaction force and the effective control policy that can extract the tooth without complications such as fractures of the jaw bone. We introduce a framework called Interactive Analysis to deal with the uncertain information in the demonstration. The framework consists of two parts, the application part and the analysis part. The model from the analysis part is used in the application for a tooth simulation algorithm. This algorithm can be applied on a robotic setup for the education and training. As for the analysis, we propose the Factor-Controlled Experiment where the uncertain information are translated into the hypotheses and are tested on a robot reproduction system. Since the robot can reproduce the same trajectory over and over again, it can keep the components in the control policies unchanged and test the hypothesis. Both components are implemented by Programming by Demonstration (PbD) which is a machine learning algorithm that encodes the control policy from the human demonstrations. The simulation is for the scenario where a robot extracts a tooth. Those two components in the framework can render an output that is similar to the demonstrations. The Factor-Controlled Experiment is performed based on the hypothesis that we derived for the plastic jaws by a robot. We found out in the experiment that the vertical force is important for loosening the tooth and settled some uncertainties about the reaction force of the tooth. However, all of the experiments are carried out on plastic jaws. We need to perform the experiment on more representative jaws in the future.

Tooth Extraction
Programming by Demonstration
Gaussian Mixture Model
Interactive Analysis
Hypothesis Testing

http://resolver.tudelft.nl/uuid:e19c03ef-b212-431c-8511-24446aa48fa1

2021-11-25

Student theses

master thesis

(c) 2016 Zhang, X.