Design and Control of a Bio-inspired Wheeled Bipedal Robot

Haizhou Zhao, Lei Yu, Siying Qin, Yurui Jin, Yuqing Chen

This paper presents a significant step forward in the design of Wheeled Bipedal Robots (WBRs), combining bio-inspired mechanical engineering with a sophisticated control framework to achieve human-like agility and robustness.

TL;DR

Researchers have developed a wheeled bipedal robot that mimics human squatting mechanics to balance motor loads and maximize lifting capacity. By utilizing a novel "Height-Variable Wheeled Linear Inverted Pendulum" (HV-wLIP) model and stability-guaranteed control, the robot can perform deep squats, navigate rough terrain, and recover from heavy impacts with remarkable agility.

What Problem Does This Paper Address?

Wheeled bipedal robots aim to combine the high-speed efficiency of wheels with the obstacle-clearing capabilities of legs. However, current designs face two major hurdles:

  1. Mechanical Inefficiency: Most serial-legged WBRs place the bulk of the load on the knee motors, leading to rapid overheating and limited carrying capacity, while the hip motors remain underutilized.
  2. Control Limitations: Standard simplified models (like the Wheeled Inverted Pendulum) assume a fixed body height. This makes them inaccurate during dynamic maneuvers like squatting or jumping, where the center of mass (CoM) height changes rapidly.

The authors tackle these issues by looking at how humans perform a barbell squat to distribute weight effectively across joints.

Methods

The authors propose a dual-pronged innovation in hardware and software:

  • Bio-inspired Mechanical Design: The robot’s leg segments (torso-to-CoM, thigh, and shank) are sized in a 3:5:4 ratio, mimicking human proportions. This specific geometry allows the robot to distribute torque evenly between the hip and knee joints (aiming for a 1:1 torque ratio), which prevents motor burnout and increases the maximum load the robot can lift.
  • HV-wLIP Model: They developed a new reduced-order model called the Height-Variable Wheeled Linear Inverted Pendulum. Unlike previous models, this one accounts for changing CoM height and wheel torques, providing a much more accurate mathematical "shorthand" for the robot's complex dynamics.
  • Stability-Guaranteed Control: The system uses a Control Lyapunov Function (CLF) integrated into a Whole-Body Controller (WBC). The CLF provides a mathematical guarantee of stability, while the WBC (formulated as a Quadratic Program) solves for the optimal motor torques in real-time (every 2ms) to satisfy balance, friction, and task requirements.

Datasets

As a robotics hardware and control paper, it does not use standard vision or language datasets. Instead, the authors evaluated their work through:

  • RaiSim Physics Simulator: Used for initial validation of the HV-wLIP model against standard models.
  • Custom Hardware Platform: A physical robot built with carbon fiber, 3D-printed parts, and high-torque brushless motors, controlled by an onboard Intel NUC.

Key Insights and Ideas

  • The "Squat" Advantage: By imitating human biomechanics, the robot can actively change its posture to control the "moment arm" of its joints. This allows it to "shift" the load between motors, much like a powerlifter adjusts their stance.
  • Decoupling Dynamics: The HV-wLIP model allows the robot to regulate its velocity by moving its legs independently of its upper body. This reduces the "excessive kinetic energy" that usually makes WBRs wobblier, leading to much faster and more stable stops.
  • Theoretical Stability in the Wild: By embedding the CLF stability condition directly into the optimization loop, the robot remains stable even when navigating unknown slopes or being kicked, without needing manual "tuning" for every specific scenario.

Key Results

  • Agile Braking: In simulations, the HV-wLIP controller stopped the robot in just 0.38m, compared to 0.60m for standard controllers—a 36% improvement in stopping distance.
  • Deep Squatting: The robot successfully performed human-like deep squats while maintaining a near-perfect 1:1 torque distribution between hip and knee motors.
  • Extreme Robustness: The physical robot recovered from a 10° lateral tilt in less than 0.8 seconds and remained upright after being heavily kicked (recovering its position within 2.2 seconds).
  • Terrain Handling: It successfully navigated steep, symmetric slopes in the wild, actively lowering its base height to keep its CoM stable.

Limitations and Future Work

  • Roll Control: The current mechanical design is passively stable in the roll direction (side-to-side), but the authors note that active roll tracking is currently limited by the mechanical structure.
  • Inertia Mismatch: There is a slight "sim-to-real" gap in torque ratios because the current model doesn't perfectly account for the mass of the lower leg (shank) segments.
  • Future Expansion: The authors plan to add robotic arms or a more complex torso to the platform to test its utility in real-world manipulation tasks, such as carrying and placing heavy objects.

Summary updated

Abstract

Wheeled bipedal robots (WBRs) have the capability to execute agile and versatile locomotion tasks. This paper focuses on improving the dynamic performance of WBRs through innovations in both hardware and software development. Inspired by the human barbell squat, a bionic mechanical design is proposed and implemented as shown in Fig. 1. It distributes the weight onto its hip and knee joints to improve the effectiveness of joint motors while maintaining a relatively large workspace of the base link. Meanwhile, a novel model-based controller is devised, synthesizing height-variable wheeled linear inverted pendulum (HV-wLIP) model, Control Lyapunov Function (CLF) and whole-body dynamics for theoretically guaranteed stability and efficient computation. Compared with other alternatives, as a more accurate approximation of the WBR dynamics, the HV-wLIP can enable more agile response and provide theory basis for WBR controller design. Experimental results demonstrate that the robot could perform human-like deep squat, and is capable of maintaining tracking CoM velocity while manipulating base states. Furthermore, it exhibited robustness against external disturbances and unknown terrains even in the wild.
BibTeX
@misc{zhao2023design,
  author = {Haizhou Zhao and Lei Yu and Siying Qin and Yurui Jin and Yuqing Chen},
  title = {{Design and Control of a Bio-inspired Wheeled Bipedal Robot}},
  year = {2023},
  eprint = {2308.13205},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2308.13205},
}
Figure