Rabbit: A Testbed for Advanced Control Theory. Chevakkereau, Abba, Aoustin, Plestan, Westervelt, Canudas-de-Wit, Grizzle. IEEE Control Systems Magazine. | Ari Weinstein's Research

Rabbit: A Testbed for Advanced Control Theory. Chevakkereau, Abba, Aoustin, Plestan, Westervelt, Canudas-de-Wit, Grizzle. IEEE Control Systems Magazine.

This is about an actual bipedal robot (basically they hips and below) for doing research on locomotion

Motivations are to get robots to go in to dangerous locations, prosthesis development

Challenges are high DOF, intermittent contacts, underactuation

“Most control strategies are built around traj tracking, with the trajs generated either off-line during a path-planning phase, or on line through a high level motion planner. Bipedal robots should clearly be an exception to this as stability is an overriding concern. The combination of highly interactive dynamics, intermittent ground contact, and underactuation makes the planning of asymptotically stabilizable, dynamic motions extremely difficult.”

Pace of theory development has not kept up with the robots themselves. Many methods heavily utilize heuristics (such as zero moment point/ZMP principle), which oversimplifies the dynamics to a set of steady state equations, which turns stable motion to frozen dynamics

This means that only slow motions can be accomplished with this method. Any dynamic motions are not allowed, even fast walking.

The robot is underactuated (no feet), and is described in terms of hybrid dynamics (floor contacts)

Has one degree of underactuation – the difference between #of DOFs and #of independent actuators

ZMP heuristic isn’t applicable on this robot because it doesn’t have feet

It is common to model impact with the floor as ocurring with compliant surface to smooth the dynamics. The problem is that in reality the surface is noncompliant, its difficult to accurately measure what the compliance would be, and if the compliance is too low, integration problems occur with the differential equations

Here they assume contact occurs instantaneously, and that the resulting change in velocities is also instantaneous. There is no change in position when this occurs

Skipping the rest (most of it)

What a nightmare! (not that the paper is poorly written, it just feels like such a complicated way to deal with the problem)

Privacy & Cookies: This site uses cookies. By continuing to use this website, you agree to their use.
To find out more, including how to control cookies, see here:
Cookie Policy