Safe Controller Synthesis With Tunable Input-to-State Safe Control Barrier Functions

Anil Alan, Andrew J. Taylor, Chaozhe R. He, Gábor Orosz, Aaron D. AmesIEEE Control. Syst. Lett 2022 · DOI

TL;DR

This paper introduces a new method called "Tunable Input-to-State Safe Control Barrier Functions" (TISSf-CBFs) to design controllers for complex systems that can guarantee safety even when facing uncertainties in their inputs. This approach offers more flexibility than previous methods, allowing for less conservative safety guarantees and improved system performance.

What Problem Does This Paper Address?

Ensuring the safety of complex systems, especially in real-world applications like autonomous vehicles, is paramount. These systems often operate under uncertain conditions, such as unpredictable environmental factors or internal system disturbances. Traditional control barrier functions (CBFs) provide a framework for guaranteeing safety by ensuring that the system's state remains within a predefined "safe set." However, existing methods for handling input disturbances, like Input-to-State Safety (ISSf) with CBFs, can be overly conservative. This means they might impose overly strict control constraints, leading to degraded performance. The core problem this paper addresses is how to design controllers that are provably safe under input disturbances while also being less conservative and more adaptable to varying levels of uncertainty.

Methods

The paper builds upon the concept of Control Barrier Functions (CBFs) and Input-to-State Safety (ISSf).

  • Standard CBFs: These functions define a "safe set" in the system's state space and provide conditions for controllers to keep the system within this set.
  • ISSf-CBFs: This extension handles bounded input disturbances by defining a larger, "input-to-state safe" set that remains invariant. However, this approach can lead to conservative controllers.
  • Tunable ISSf-CBFs (TISSf-CBFs): This is the paper's main contribution. TISSf-CBFs generalize ISSf-CBFs by introducing a tunable function, denoted by $\epsilon(h(x))$, into the safety condition. This tunability allows for a more flexible definition of the safe set under disturbance. The authors propose a new controller structure that leverages this tunable function. The core idea is to allow the safe set to expand in a controlled manner based on the uncertainty, rather than a fixed expansion as in traditional ISSf-CBFs. This flexibility aims to reduce conservatism without sacrificing safety.

Datasets

Not discussed in the paper. The paper uses simulated examples and a high-fidelity vehicle model for demonstration.

Key Insights and Ideas

  • Tunability is Key: The introduction of a tunable function $\epsilon(h(x))$ is the central innovation. This allows for a trade-off between the size of the safe set under disturbance and the conservatism of the controller.
  • Reduced Conservatism: By making the safe set's expansion tunable, the controller can be less restrictive, leading to better performance compared to fixed-expansion ISSf methods.
  • Flexibility in Design: The TISSf-CBF framework provides engineers with more knobs to turn when designing safety-critical controllers, allowing them to tailor safety guarantees to specific application needs and uncertainty levels.
  • Connection to Existing Methods: The paper shows how the standard ISSf-CBF approach is a special case of their TISSf-CBF framework, demonstrating its generality.

Key Results

The paper demonstrates the effectiveness of TISSf-CBFs through theoretical results and simulations:

  • Theoretical Guarantees: The authors provide theorems proving that controllers designed using TISSf-CBFs ensure forward invariance of a tunable safe set ($C_{\delta,T}$) under bounded input disturbances.
  • Example 1 (Simple System): A basic control system with input disturbance shows that TISSf-CBFs can maintain safety while allowing trajectories to approach the desired equilibrium more efficiently than ISSf-CBFs.
  • Example 2 (Disturbed System): This example highlights how ISSf-CBFs can lead to overly conservative behavior (larger distances maintained), while TISSf-CBFs can achieve safety with less performance degradation.
  • Example 3 (Connected Cruise Control): A high-fidelity simulation of a connected automated truck demonstrates that TISSf-CBFs can maintain a safe distance from a preceding vehicle during an emergency braking scenario, outperforming a standard CBF controller and showing better performance than ISSf-CBFs. The simulations show that the TISSf-CBF controller keeps the system within the safe set and avoids large headway distances at steady state.

Limitations and Future Work

  • Limitations:
    • The paper notes that the condition on the tunable function $\epsilon$ in Theorem 3, while easily verifiable, is stronger than strictly necessary.
  • The theoretical framework assumes bounded input disturbances.
  • Future Work:
    • Implementing the safety-critical control based on TISSf-CBFs on a real automated truck.
  • Ensuring safety experimentally.
  • Exploring the use of TISSf-CBFs with optimization-based controllers (e.g., quadratic programming) for more complex scenarios.

Summary updated

BibTeX
@article{alan2022safe,
  author = {Anil Alan and Andrew J. Taylor and Chaozhe R. He and Gábor Orosz and Aaron D. Ames},
  title = {{Safe Controller Synthesis With Tunable Input-to-State Safe Control Barrier Functions}},
  journal = {IEEE Control. Syst. Lett},
  year = {2022},
  doi = {10.1109/lcsys.2021.3087443},
}
Figure