Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

The NING Humanoid: The Concurrent Design and Development of a Dynamic and Agile Platform

Yan Ning, Song Liu, Taiwen Yang, Liang Zheng, Ling Shi

TL;DR

This work tackles the challenge of achieving highly dynamic humanoid motion by proposing a concurrent hardware-control co-design centered on centroidal dynamics and a real-time whole-body model predictive control framework. The NING humanoid integrates modular quasi-direct-drive actuators, a four-bar transmission, and inertia-focused layout to concentrate mass near the center of mass, enabling robust, dynamic tasks such as walking, push recovery, stair climbing, and simulated back-flips. The control pipeline combines an MPC-based locomotion predictor with a convex optimization-based whole-body controller solved in real time, providing stable and responsive locomotion across varied terrains. Overall, the design demonstrates that co-designing hardware and control around centroidal dynamics yields improved agility and robustness, with future work aimed at perception-based estimation and trajectory planning.

Abstract

The recent surge of interest in agile humanoid robots achieving dynamic tasks like jumping and flipping necessitates the concurrent design of a robot platform that combines exceptional hardware performance with effective control algorithms. This paper introduces the NING Humanoid, an agile and robust platform aimed at achieving human-like athletic capabilities. The NING humanoid features high-torque actuators, a resilient mechanical co-design based on the Centroidal dynamics, and a whole-body model predictive control (WB-MPC) framework. It stands at 1.1 meters tall and weighs 20 kg with 18 degrees of freedom (DOFs). It demonstrates impressive abilities such as walking, push recovery, and stair climbing at a high control bandwidth. Our presentation will encompass a hardware co-design, the control framework, as well as simulation and real-time experiments.

The NING Humanoid: The Concurrent Design and Development of a Dynamic and Agile Platform

TL;DR

This work tackles the challenge of achieving highly dynamic humanoid motion by proposing a concurrent hardware-control co-design centered on centroidal dynamics and a real-time whole-body model predictive control framework. The NING humanoid integrates modular quasi-direct-drive actuators, a four-bar transmission, and inertia-focused layout to concentrate mass near the center of mass, enabling robust, dynamic tasks such as walking, push recovery, stair climbing, and simulated back-flips. The control pipeline combines an MPC-based locomotion predictor with a convex optimization-based whole-body controller solved in real time, providing stable and responsive locomotion across varied terrains. Overall, the design demonstrates that co-designing hardware and control around centroidal dynamics yields improved agility and robustness, with future work aimed at perception-based estimation and trajectory planning.

Abstract

The recent surge of interest in agile humanoid robots achieving dynamic tasks like jumping and flipping necessitates the concurrent design of a robot platform that combines exceptional hardware performance with effective control algorithms. This paper introduces the NING Humanoid, an agile and robust platform aimed at achieving human-like athletic capabilities. The NING humanoid features high-torque actuators, a resilient mechanical co-design based on the Centroidal dynamics, and a whole-body model predictive control (WB-MPC) framework. It stands at 1.1 meters tall and weighs 20 kg with 18 degrees of freedom (DOFs). It demonstrates impressive abilities such as walking, push recovery, and stair climbing at a high control bandwidth. Our presentation will encompass a hardware co-design, the control framework, as well as simulation and real-time experiments.
Paper Structure (12 sections, 1 equation, 7 figures)

This paper contains 12 sections, 1 equation, 7 figures.

Table of Contents

  1. INTRODUCTION
  2. RELATED WORKS
  3. HARDWARE DESIGN
  4. Actuation and Mechanism
  5. Model-based Concurrent Design
  6. CONTROL FRAMEWORK
  7. SIMULATION & EXPERIMENTS
  8. CONCLUSION
  9. Specification of the Actuators
  10. Comparison between Inertia of Different Designs
  11. Illustrations of the Design and Experiments
  12. The demonstration

Figures (7)

  • Figure 1: The NING humanoid platform at experimental scenarios.
  • Figure 2: The design details of the NING humanoid, including robot size, transmission mechanism, axes and joints.
  • Figure 3: Illustrations of back-flipping in simulation and the results of different joint torques in each leg.
  • Figure 4: The specification each QDD actuator utilized in the NING humanoid
  • Figure 5: The comparison of CoM and momentum of inertia change based on different design principles.
  • ...and 2 more figures