- 261 views
- 1328 downloads
Towards Prosthetic Arms as Wearable Intelligent Robots
-
- Author / Creator
- Sherstan, Craig
-
The control of powered prosthetic arms has been researched for over 50 years, yet prosthetic control remains an open problem, not just from a research perspective, but from a clinical perspective as well. Significant advances have been made in the manufacture of highly functional prosthetic limbs, yet the control of such limbs remains largely impractical. The core issue is that there is a significant mismatch between the number of functions available in modern powered prosthetic arms and the number of functions an amputee can actively attend to at any given moment.
One approach to addressing this mismatch is the idea of treating the arm as an intelligent, goal-seeking agent - such an agent can learn from its experience and adapt its actions to improve its ability to accomplish a goal. It is hypothesized that such intelligent agents will be able to compensate for the existing limitations in the communication bandwidth between a powered prosthetic arm and an amputee. The work of this thesis looks at several steps towards building such agency into a prosthetic arm, including pattern recognition methods, compound predictions, and collaborative control between the arm and the user. Essentially, this body of work looks at ways of understanding the user's desires, as measured in various ways, such as desired movements, or expected future joint angles, and controlling the arm so as to achieve those desires.
The first contribution of this thesis is the identification of a scenario under which current pattern recognition approaches to prosthetic control do not generalize well. The second contribution is the demonstration that it is possible to layer predictors, known as general value functions, and that such layering can improve feature representation and predictive power. Finally, this thesis demonstrates a method for improving the control of a prosthetic arm using a collaborative control method that learns predictions of user behavior which are then used to assist in controlling the arm.
In the long term, the methods and philosophy to prosthetic control explored in this thesis may greatly improve an amputee's ability to control their prosthesis. Further, this approach may be extended to other domains of human-machine interaction where there is a mismatch between the number of functions in a system and the user's ability to attend to those functions, such as smart phones, computers and teleoperated robots.
-
- Graduation date
- Fall 2015
-
- Type of Item
- Thesis
-
- Degree
- Master of Science
-
- License
- This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.