Publication Date
8-2023
Date of Final Oral Examination (Defense)
May 2023
Type of Culminating Activity
Dissertation
Degree Title
Doctor of Philosophy in Electrical and Computer Engineering
Department
Electrical and Computer Engineering
Major Advisor
Aykut C. Satici, Ph.D.
Advisor
John Chiasson, Ph.D.
Advisor
Kurtis Cantley, Ph.D.
Advisor
Hao Chen, Ph.D.
Advisor
Arash Komaee, Ph.D.
Abstract
We provide several data-driven control design frameworks for contact-rich robotic systems. These systems exhibit continuous state flows and discrete state transitions, which are governed by distinct equations of motion. Hence, it is difficult to design a single policy that can control the system in all modes. Typically, hybrid systems are controlled by multi-modal policies, each manually triggered based on observed states. However, as the number of potential contacts increase, the number of policies can grow exponentially and the control-switching scheme becomes too complicated to parameterize. To address this issue, we design contact-aware data-driven controllers given by deep-net mixture of experts. This architecture automatically finds a switching-control scheme that can achieve the desired overall performance of the system, and a gating network, which determines the region of validity of each expert, based on the observed states.
Additionally, we address the adverse effects of model uncertainties in the control of contact-rich robots. Lack of accurate environmental models can misrepresent the effects of contact forces on the system. Policies designed from such models can lead to poor performance or even instability. In particular, we demonstrate the effects of system parameter uncertainties and measurement errors on the overall performance of the system. Then, we design data-driven stochastic controllers that combine the stability properties of passivity-based control with the robustness properties of Bayesian learning.
DOI
https://doi.org/10.18122/td.2091.boisestate
Recommended Citation
Ashenafi, Nardos Ayele, "Robust Control of Contact-Rich Robots via Neural Bayesian Inference" (2023). Boise State University Theses and Dissertations. 2091.
https://doi.org/10.18122/td.2091.boisestate