TASP PhD Seminar – Adar Gaathon – “Development of a Robotic Leg Using Minimalistic Control Methods for Increased Robustness of a Running Robot”

Work towards PhD degree under the supervision of Assoc. Prof. Amir Degani

When: 21.2.2021 at 10:30

Where: zoom

Abstract: The superior ability of dynamic legged locomotion to traverse rough terrain comes with the cost of fragile stability. As opposed to the commonly used closed loop control schemes, simple control schemes that only use a few basic sensors and no feedback, improve the stability of simple robots when applying a single controller. Controlling the leg angle during descent of our hopping monopod helps it keep its balance when traversing unforeseen rough terrain. Exploiting multiple controllers simultaneously, such as the free-leg length and stiffness, can further improve robustness but is often mechanically hard to implement. To overcome the mechanical complexity of designing and implementing multiple controllers, we investigate a curved leg shape that applies variable leg stiffness and free-leg length coupled with the controller, which is the leg angle during descent. We study the effect of a combination of parameters during the stance phase and show that, when traversing unknown rough terrain, it can increase robustness to perturbations in the initial horizontal velocity. We use the policy of keeping a relative height above terrain that was previously demonstrated to increase a hopper’s multistep stability. We further investigate the ability of a single leg shape to reach various relative heights with only a change in the leg angle controller. We exploit the fact that as various areas of the curved leg come in contact with the ground, they behave differently in applying physical parameters like stiffness and free-leg length. We propose a theoretical model to describe the controller coupling; present the process of obtaining the optimal coupled parameters and demonstrate its benefits by simulation. Our work also exemplifies the leg design process and validates the simulated results with experiments.

You can see the seminar Here