Human walking is a complex dynamic process, involving moving through positions which are not themselves stable to create an overall periodic movement that is stable. The arrangements of our joints, muscles and tendons give us the ability to do more than just control the position of our limbs: we can store energy, absorb shock forces, and adjust the stiffness and damping of our joints. Importantly, through the interplay of these mechanisms and neural control we are able to produce walking behaviour which is remarkably efficient, taking full advantage of the natural dynamics of our bodies. It is very difficult to deconstruct this behaviour to discover, for example, how the stiffness of our joints changes as we walk, or as we change the speed at which we are walking.
In the field of robotics bipedal locomotion as been a hot topic for some time. Typically constructed robots used only rigid joints, and were largely incapable of taking advantage of their dynamics in any way.
There was a movement to create passive dynamic walkers, which were mechanically designed with compliance and no (or very little) actuation to produce a walking gait, however these robots can only walk at one speed, and are incapable of even standing still. The introduction of compliance into a fully powered walker gives more options for energy storage and shock absorption, but is not adaptable to take advantage of the dynamics during different movements, and can never be made rigid without driving the motors to oppose induced changes in joint angles.
Only recently have researchers begun to look at bipedal robots which are capable of varying their joint impedance, and so far no robust platforms mimicking human dynamics have been created. Furthermore there have been no bipedal robots at all that are capable of varying joint stiffness and damping independently.
The creation of such a robot, which would be capable of exhibiting efficient multimodal locomotion, would provide an excellent platform not only to further the field of bipedal robotics, but to investigate the role of stiffness and damping in bipedal locomotion. We can explore control strategies that exploit these parameters, using optimal control methods with cost functions based around efficiency, stability, speed etc. and neurologically inspired control architectures combining rhythmical pattern generators with reflexive feedback. On a robot we can vary control parameters in ways that would be impossible in humans, and attempt to gather suppositions about the underlying dynamics of human walking. Furthermore the work could see practical applications (in addition to pure robotics) in prosthesis, control of prosthesis, and rehabilitation robotics.
This project therefore aims to create a robust bipedal robotic platform, the first in the world which will allow the independent control of joint stiffness and damping. With this platform strategies for control will be investigated, and it is hoped that some underlying principles can be generalised for bipedal locomotion in order to provide feedback into the understanding of human walking.
Year one sees the construction of the first iteration of the robotic platform and rudimentary control research. Years two and three will concentrate on developing control strategies for efficient, robust, adaptive locomotion. The hardware may be revised in year three as necessary.
Related Publications and Presentations
- Alexander M Enoch, Andrius Sutas, Shin ichiro Nakaoka, and Sethu Vijayakumar, “BLUE: A Bipedal Robot with Variable Stiffness and Damping”, IEEE-RAS International Conference on Humanoid Robots 2012, 2012.