ECE Theses and Dissertations

Permanent URI for this collection

https://hdl.handle.net/1911/8316

Browse

Browsing ECE Theses and Dissertations by Subject "Aerospace engineering"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Issues in myoelectric teleoperation of complex artificial hands
    (1995) Farry, Kristin Ann; Walker, Ian D.
    This dissertation introduces a novel method of teleoperation of complex anthropomorphic robotic hands: converting the myoelectric signal generated by an operator's muscles during movement into robot commands replicating the motion. This teleoperation scenario is, in a sense, the limiting case of myoelectric prosthetic hand control. This project contributes to implementation of a practical myoelectric teleoperation system and improved prosthetic hand control by analyzing the myoelectric spectrum's variation during thumb motions. The investigation applies a new spectral estimation approach, Thomson's multiple window method (MWM), to the myoelectric signal. The MWM estimate has much lower bias and variance than traditional periodogram estimates, making it a better candidate to compute motion classification features. The MWM is also less sensitive to motion artifact than autoregressive methods. Extending Thomson's MWM into a time-frequency analysis tool analogous to the short-time Fourier transform, here called the short-time Thomson transform, shows that the myoelectric signal may be more stationary than previously thought. This project includes development of a unique myoelectric data collection system (MDCS) and a myoelectric teleoperation demonstration system (MTDS). The MDCS allows simultaneous measurement of 16 hand joint motions and 8 myoelectric signals. This capability enables close alignment of myoelectric signatures in time based on the hand motions and a search for motion-specific temporal characteristics in the myoelectric signal. While this study yields little evidence of motion-specific temporal consistency, it shows promising motion-specific spectral consistency. Spectral analysis proves less sensitive to alignment uncertainties than temporal analysis. An evaluation of five techniques for finding a motion's starting point in the myoelectric signal, a major implementation concern, suggests that we not pursue alignment-sensitive myoelectric control algorithms. Finally, the MTDS is used to demonstrate myoelectric control of chuck and key grasp motions of NASA/JSC's Utah/MIT Dextrous Hand, realtime, with 90% accuracy. The demonstration uses the time-varying myoelectric spectrum estimated with short-time Fourier transforms; however, this project lays the foundation for using the superior short-time Thomson transforms in this application.
  • Thumbnail Image
    Item
    Using reinforcement learning to control advanced life support systems
    (2005) Klein, Theresa J.; Subramanian, Devika
    This thesis deals with the application of reinforcement learning techniques to the control of a closed life support system simulator, such as could be used on a long duration space mission. We apply reinforcement learning to two different aspects of the simulator, control of recycling subsystems, and control of crop planting schedules. Comparisons are made between distributed and centralized controllers, generalized and non-generalized RL, and between different approaches to the construction of the state table and the design of reward functions. Distributed controllers prove to be superior to centralized controllers both in terms of speed and performance of the controller. Generalization helps to speed convergence, but the performance of the policy derived is dependent on the shape of the reward function.