Questions on VH-MPC with Swing Foot Dynamics for Bipedal Walking Control


I have some questions regarding this paper after reading and taking notes. I wasn’t sure what topic category this would fall under so I just put General for now.

Question 1:
How is the contact switch determined when mentioned in Section II: Foot Location & Timing Adjustment? From what I understand, the contact switch is how the system knows that the DCM-dynamics has ended of the swing foot/leg?

But the real bolt robot hardware does not have any foot contact sensors, so how is this terminal state determined/verified?

Question 2:
Under Section III: Projected Swing Foot Dynamics part B: Constraints on the Swing Foot Actuation Force,
for the linear program to find the approximation of force boundaries on the swing foot, is there any physical way of interpreting roe and eta matrices?


I took a look at the referenced [24] paper, An Optimization-Based Locomotion Controller for Quadruped Robots Leveraging Cartesian Impedance Control. But I still don’t fully understand why the constraints are expressed this way and what they actually mean/encode. Is there further reading/material that I need to understand first?


Hi Roy,

Thanks for the questions. Here are the answers:
Question 1: The contact switch happens in the implementation only based on the planned time of contact and no feedback is used. However, it is possible to make the contact switch based on feedback by using estimated force force at the end-effector using measure current at the joints and then thresholding this force to detect contact.
Question 2: Yes, $\rho + \eta$ is simply the amount of force asked by the motion of the robot at the stance foot, assuming that the stance foot is stationary .Now if this force is within the friction cone constraint, this motion is dynamically feasible and our assumption is valid.

1 Like