Record Details
Field | Value |
---|---|
Title | Optimal control of an underactuated bipedal robot |
Names |
Jones, Mikhail S.
(creator) Hurst, Jonathan W. (advisor) |
Date Issued | 2014-06-06 (iso8601) |
Note | Graduation date: 2015 |
Abstract | Bipedal robots represent a unique class of control problems that combine many of the most difficult elements of nonlinear control. These robots are typically designed to be mobile and as such have limited energy and actuator authority making efficiency a prime concern. Unlike wheeled robots, legged robots must transition between different equations of motion as legs make and break contact with the ground, aggravating the complexities of hybrid dynamics. Furthermore, small feet and series elastic actuation of joints leads to an underactuated system, causing traditional nonlinear control methods to struggle. This thesis describes and validates a general framework for efficiently controlling systems with the properties most problematic in bipedal robots, i.e., nonlinearity, hybrid dynamics, and underactuation. This process is two-step in that, first, an optimal trajectory is generated using direct collocation trajectory optimization, and second, the feasible trajectory is stabilized using a time varying linear quadratic regulator. We demonstrate this process on a number of toy problems including the triple pendulum on a cart and a reduced order template for a running robot. This approach is then applied to the bipedal robot ATRIAS in simulation in order to achieve efficient, dynamic, and robust locomotion behaviors. |
Genre | Thesis/Dissertation |
Topic | ATRIAS |
Identifier | http://hdl.handle.net/1957/50690 |