1. Home ›
  2. Research ›
  3. Multi-body Dynamics and Identification

Multi-body Dynamics and Identification

Illustrations of a Dash humanoid robot performing sprint running, powerlifting, and carrying heavy objects while walking movements.

Overview

Advances in electric actuators technology have elevated robot’s physical capabilities closer to the performance level expected in real-world operations. While these powerful actuators have been highly adopted in quadrupeds, they have yet found the same level of adoption in human form factors. The gap can be attributed to the necessity to install these large-radius electric actuators in a parallel arrangement in order to facilitate load sharing between motors. However, the parallel actuation topology introduces additional complexities in the form of closed kinematics chain which provides a significant hurdle in dynamic simulation and identification.

One aspect of this research is to develop an efficient and accurate simulation method for dynamic systems with parallel actuation topologies. While existing methods have steep trade-offs between accuracy and computation cost, we look into approaches that are able to perform well on both metrics. We aim to do this by leveraging on Gauss principle of least constraint to generalize open-chain rigid body dynamic algorithms to cover more complex structures.

The second aspect of this research is to develop computational tools to precisely identify the actuator parameters exhibited in parallel actuation designs. In open-chain arrangements, each actuator can traditionally be lumped into the mass of the preceding link. However, this assumption breaks down in the closed-chain arrangement. To resolve this, we take a closer look into how the actuator arrangement affects the identifiability of the parameters and the experimental conditions necessary to perform the identification.

Our contribution will be implemented on a Dash humanoid and will be validated on challenging dynamics movements such as sprint running, powerlifting, and carrying heavy objects while walking.

Publications

M. Chignoli, N. Adrian, S. Kim, and P. M. Wensing. A propagation perspective on recursive forward dynamics for systems with kinematic loops, (submission under review) 2024. arXiv: 2311.13732.