Safety-based Speed Control of a Wheelchair using Robust Adaptive Model Predictive Control

Meng Yuan, Ye Wang, Lei Li, Tianyou Chai, Wei Tech Ang

Introduction

Electric wheelchairs are vital for mobility, yet their control systems often rely on raw joystick inputs that ignore physical constraints and safety boundaries. Sudden joystick movements can lead to high-acceleration maneuvers, potentially causing the wheelchair to tip or injuring the user. Furthermore, wheelchairs operate under significant uncertainty, including external disturbances like inclined ramps and parameter variations such as changes in user weight. A controller optimized for one user may become unsafe for another. This paper addresses these challenges by introducing a safety-first control system using Robust Adaptive Model Predictive Control (RAMPC).

Methodology

The proposed architecture employs a multi-layered approach to ensure stability and safety:

  • Linear Parameter Varying (LPV) Model: Captures the dynamics between motor voltage and wheelchair velocity/acceleration.
  • Set-Membership Parameter Estimation: Instead of a fixed parameter set, the system maintains a range of possible values (e.g., mass, friction) and updates them online. This allows the controller to adapt to different users automatically.
  • Admissible Reference Generation: This "safety filter" intercepts joystick inputs. If a user requests a speed or acceleration that violates safety limits, the system calculates and executes the closest safe alternative.
  • Tube-based Robust MPC: To handle disturbances, the controller plans a central trajectory while ensuring the wheelchair remains within a safe "tube" of operation, even under worst-case conditions.

Experimental Setup

The researchers performed system identification on a physical electric wheelchair equipped with Brushless DC motors and absolute encoders. Data was collected from an 80kg participant to identify motor parameters and nominal mass/damping matrices.

The experimental electric wheelchair platform used for system identification and parameter estimation.

The final evaluation utilized a high-fidelity simulation model to test the adaptive capabilities across varying user weights (60kg to 85kg).

Key Results

The RAMPC system was compared against a standard Proportional-Integral (PI) controller across two primary tasks:

  1. Ramp Navigation: On 5-degree inclines, the RAMPC maintained acceleration within the 2 m/s² safety limit. The PI controller, conversely, exceeded these tolerances and exhibited velocity oscillations.
  2. Sinusoidal Tracking and Safety Clipping: When tracking a fluctuating target speed that crossed safety boundaries, the RAMPC followed the reference when safe and strictly "clipped" the speed at the safety limit.
Comparison of the proposed RAMPC controller against a standard PI controller, demonstrating adherence to velocity bounds.

Quantitatively, the RAMPC achieved a maximum tracking error of 0.048 m/s, a significant improvement over the PI controller's 0.158 m/s.

Conclusion and Future Work

The RAMPC framework effectively prioritizes safety without sacrificing responsiveness, making it ideal for shared mobility scenarios like hospital wheelchair rentals. While the computational load of MPC is higher than simpler loops, the safety benefits are substantial. Future research will focus on integrating external obstacle avoidance (e.g., LiDAR) and conducting extensive real-world trials with human subjects to validate safety performance in edge-case scenarios.

Summary updated

BibTeX
@article{yuan2024safety,
  author = {Meng Yuan and Ye Wang and Lei Li and Tianyou Chai and Wei Tech Ang},
  title = {{Safety-based Speed Control of a Wheelchair using Robust Adaptive Model Predictive Control}},
  journal = {IEEE Trans. Cybernetics},
  year = {2024},
  doi = {10.1109/tcyb.2023.3309369},
}
Figure