Nonlinear modelling and analysis of stable behaviour of robot gait control system based on image processing technology

Jin Wang, Hui Lin, Wenbing YangInt. J. Comput. Intell. Stud 2023 · DOI

Introduction

Achieving stable, human-like walking in bipedal robots remains a significant challenge in the field of humanoid robotics. Traditional gait control systems often struggle to balance accuracy with real-time processing speed. This research presents a robust nonlinear modeling approach that integrates advanced image processing with sliding mode control (SMC) to enhance the stability and responsiveness of bipedal gait.

The Challenge of Bipedal Stability

Most existing gait control systems rely on linear models. While computationally simpler, these models often fail to capture the complex, high-dimensional dynamics inherent in walking, especially when the robot encounters uneven terrain or external disturbances. Furthermore, robots utilizing visual inputs frequently encounter "noisy" data from CCD sensors, which can lead to erratic movements or significant system lag. The objective of this study is to bridge the gap between mathematical stability and real-time computational efficiency.

Methodology

The proposed system utilizes a multi-layered architecture designed to handle 19 degrees of freedom (DOF) without overloading the central processor.

Hardware Architecture

The system employs a "two-stage" hardware design. A high-performance DSP (TMS320VC5509A) serves as the primary controller, handling high-level logic, attitude calculations, and sensor fusion. This is supported by a dedicated steering gear control panel that manages pulse-width modulation (PWM) signals for the individual servos, ensuring precise joint movement.

Hardware system block diagram showing the DSP controller, sensors, and steering engine interface.

Image Processing and Control

To address sensor noise, the researchers implemented a bilateral image filtering method. Unlike standard filters that may blur critical edges, bilateral filtering considers both spatial proximity and brightness similarity. This allows the robot to eliminate sensor noise while preserving the sharp contours of obstacles or markers necessary for navigation.

The core of the stability analysis is a nonlinear sliding mode control (SMC) closed-loop system. By mapping image space coordinates to motion space velocities via a Jacobian matrix, the controller ensures the robot's state remains on a stable trajectory. The authors mathematically demonstrated that this error system state space possesses a globally asymptotically stable equilibrium point, guaranteeing that the robot will consistently work toward a stable gait regardless of its initial position.

Experimental Setup

The methodology was validated using a custom bipedal prototype. The robot stands 42 cm tall, weighs 2.64 kg, and features 19 degrees of freedom. Testing involved both simulation and physical experiments where the robot was tasked with accelerating to 1 m/s, maintaining speed, and stopping within specific timeframes.

The 19-DOF bipedal robot prototype used for gait stability testing.

Key Results

The proposed nonlinear method significantly outperformed traditional benchmarks across several critical metrics:

  • Control Accuracy: The system achieved a gait control accuracy of 95% to 100%, a substantial improvement over previous robust reliability methods which typically hover between 70% and 80%.
  • Reaction Time: In complexity tests, the system reacted in 0.194s to 0.202s, roughly twice as fast as competing algorithms (0.410s).
  • Efficiency: Total gait control time remained consistently under 28 seconds, compared to the 35-45 second range seen in other recent studies.

Conclusion

By embracing the nonlinear nature of bipedal movement and integrating image pre-processing directly into the control loop, this research provides a more resilient framework for humanoid navigation. Future work will focus on optimizing the image processing pipeline for extreme real-world environments and improving high-level trajectory planning to achieve more fluid, human-like motion.

Summary updated

BibTeX
@article{wang2023nonlinear,
  author = {Jin Wang and Hui Lin and Wenbing Yang},
  title = {{Nonlinear modelling and analysis of stable behaviour of robot gait control system based on image processing technology}},
  journal = {Int. J. Comput. Intell. Stud.},
  year = {2023},
  doi = {10.1504/ijcistudies.2023.137848},
}
Figure