Rollover Prevention for Mobile Robots With Control Barrier Functions: Differentiator-Based Adaptation and Projection-to-State Safety

Ersin Das, Aaron D. Ames, Joel W. Burdick

TL;DR

This paper introduces a novel safety framework for mobile robots that prevents rollovers on uneven terrain by combining control barrier functions (CBFs) with a differentiator-based approach. This method robustly handles time-varying system parameters and disturbances, ensuring safety even with noisy sensor data, and is experimentally validated on a tracked robot.

What Problem Does This Paper Address?

Ensuring the safety of autonomous mobile robots, especially when operating in challenging, real-world environments like steep slopes, is paramount. A critical failure mode for such robots is rollover. Traditional methods for preventing rollovers often rely on precise models or extensive sensor data, which can be impractical or fail in the face of real-world uncertainties. Specifically, rollover risk is often assessed using metrics like the Zero Moment Point (ZMP), but these metrics can depend on parameters that change over time and are measured with noise (e.g., gravity vector derivatives). This noise makes it difficult to guarantee that the robot stays within its safe operating region, leading to potential rollovers. The paper aims to bridge this gap by developing a safety-critical control strategy that is both rigorous and implementable, even with noisy, time-varying parameters.

Methods

The core of the paper's contribution lies in its novel approach to robustifying control barrier functions (CBFs) against time-varying parameters and disturbances.

  • Differentiator-Based Adaptation (DA-CBFs): The authors introduce a method to handle time-varying parameters that are crucial for safety conditions (like ZMP). Instead of directly using noisy measurements of these parameters, they employ a differentiator to estimate their time derivatives. This differentiator is designed to be Input-to-State Stable (ISS) with respect to noise, meaning that bounded noise inputs lead to bounded estimation errors. The CBF conditions are then adapted to incorporate the estimated parameters and the dynamics of the differentiator.
  • Time-Varying Projection-to-State Safety (tPSSf): To further enhance robustness against disturbances, the paper extends the concept of Projection-to-State Safety (PSSf). This new framework, tPSSf, accounts for time-varying bounds on the "projected disturbance" (the effect of system disturbances on the CBF's time derivative). This allows for a more dynamic and less conservative safety guarantee compared to static bounds.
  • CBF-Quadratic Program (CBF-QP) Safety Filter: The derived DA-CBFs are integrated into a CBF-QP framework. This safety filter modifies the output of a nominal (potentially unsafe) controller only when necessary to ensure that the DA-CBF conditions are met, thereby guaranteeing safety.

Datasets

Not discussed in the paper. The experiments utilize a physical tracked robot and its onboard sensors.

Key Insights and Ideas

  • The critical role of accurate parameter estimation: The paper highlights that noisy measurements of time-varying parameters (like those related to gravity and robot dynamics) are a major bottleneck for robust safety guarantees in mobile robots. Simply using static bounds on these parameters leads to overly conservative behavior or safety violations.
  • ISS differentiators as a robust building block: By employing ISS differentiators, the authors show that it's possible to obtain reliable estimates of time-varying parameters even in the presence of sensor noise. This is crucial for enabling dynamic safety conditions.
  • Time-varying safety bounds are more effective: The introduction of tPSSf allows for more adaptive safety margins. Instead of a fixed, conservative safety region, the safe set can dynamically adjust, leading to less restrictive trajectories while still maintaining safety.
  • DA-CBFs bridge the gap: The proposed DA-CBFs effectively integrate the ISS differentiator dynamics into the CBF framework, creating a safety condition that is robust to both parameter variations and external disturbances.

Key Results

The paper demonstrates the effectiveness of its proposed method through experimental validation on a tracked mobile robot operating on a steep slope (approximately 27°).

  • Successful Rollover Prevention: The experiments show that the proposed differentiator-based CBF approach successfully prevents rollovers, even when the nominal controller would lead to unsafe conditions.
  • Improved Performance over PSSf: Comparisons with a standard PSSf approach (using time-invariant bounds) reveal that the proposed method yields less conservative trajectories. While PSSf with larger bounds ensures safety, it leads to significantly restricted robot motion. The proposed method achieves safety with more efficient paths.
  • Robustness to Noise: The experiments highlight that a simple backward differentiator can lead to safety violations due to sensor noise, underscoring the importance of the ISS differentiator design. The proposed method, utilizing a robust differentiator, performs significantly better.

Limitations and Future Work

  • Model Simplification: The paper uses a simplified model for the mobile robot's dynamics (unicycle model with first-order actuator dynamics). While this is common for control design, extending the framework to more complex, high-fidelity robot models could be a future direction.
  • Assumptions on Parameter Dynamics: The method relies on the assumption that the time-varying parameters can be differentiated and that their derivatives are globally Lipschitz continuous. Relaxing these assumptions could broaden the applicability.
  • Experimental Scope: While the experiments demonstrate effectiveness on a specific tracked robot and slope, further validation on different robot platforms and more diverse challenging terrains would be beneficial.
  • Future Work: The authors suggest exploring integral CBFs for more complex control input-dependent safety conditions and investigating learning-based approaches for robust control barrier function synthesis.

Summary updated

Abstract

This paper develops rollover prevention guarantees for mobile robots using control barrier function (CBF) theory, and demonstrates the method experimentally. We consider a safety measure based on a zero moment point condition through the lens of CBFs. However, these conditions depend on time-varying and noisy parameters. To address this issue, we present a differentiator-based safety-critical controller that estimates these parameters and pairs Input-to-State Stable (ISS) differentiator dynamics with CBFs to achieve rigorous safety guarantees. Additionally, to ensure safety in the presence of disturbances, we utilize a time-varying extension of Projection-to-State Safety (PSSf). The effectiveness of the proposed method is demonstrated via experiments on a tracked robot with a rollover potential on steep slopes.
BibTeX
@article{das2024rollover,
  author = {Ersin Das and Aaron D. Ames and Joel W. Burdick},
  title = {{Rollover Prevention for Mobile Robots With Control Barrier Functions: Differentiator-Based Adaptation and Projection-to-State Safety}},
  journal = {IEEE Control. Syst. Lett},
  year = {2024},
  doi = {10.1109/lcsys.2024.3416239},
}
Figure