TL;DR
This paper introduces a robust framework for ensuring safety in automated vehicles using Control Barrier Functions (CBFs) and Input-to-State Safety (ISSf). By filtering a nominal controller with CBFs, the system can maintain performance while guaranteeing safety, even in the presence of unmodeled disturbances, as demonstrated on a heavy-duty truck.
What Problem Does This Paper Address?
Balancing safety and performance is a critical challenge in modern control systems, especially for safety-critical applications like automated vehicles. Traditional control methods often struggle to provide rigorous safety guarantees without sacrificing performance, particularly when dealing with real-world complexities like unmodeled dynamics and external disturbances. This paper addresses the need for a control synthesis approach that can proactively ensure safety, intervening only when necessary, and can robustly handle uncertainties in the system model and external inputs. Specifically, it tackles the problem of ensuring safety for automated vehicles in scenarios where precise modeling is difficult, and performance must be maintained.
Methods
The core of the paper's methodology lies in the constructive synthesis of safety-critical controllers using Control Barrier Functions (CBFs). CBFs are mathematical tools that define a "safe set" in the system's state space and provide conditions for ensuring that the system's state remains within this set. The authors extend this by incorporating Input-to-State Safety (ISSf), which quantifies how the safe set might grow or shrink in the presence of disturbances.
The approach involves:
- CBF Formulation: Defining a safe set (e.g., maintaining a safe following distance) using a continuously differentiable function $h(x)$ and ensuring its 0-superlevel set is forward invariant.
- Nominal Controller Design: Designing a controller that prioritizes performance (e.g., a connected cruise controller for a truck) without explicit safety guarantees.
- Safety Filtering: Modifying the nominal controller using CBF conditions to create a safety-critical controller. This is achieved by solving an optimization problem that minimizes deviation from the nominal controller while satisfying the CBF constraints. The paper provides a closed-form solution for this optimization, resulting in a controller that only intervenes when safety is at risk.
- Robustness with ISSf: For systems with input disturbances, the authors leverage ISSf-CBFs. This allows for a formal guarantee that a slightly expanded safe set ($C_\delta$) remains forward invariant, scaling with the magnitude of the disturbance. A robust safety-critical controller is then synthesized using ISSf-CBF conditions.
- Design-Test-Redesign: The methodology emphasizes a practical design process where simulations and experiments are used to identify unmodeled disturbances, quantify their impact, and then use ISSf to robustify the controller.
Datasets
Not discussed in the paper. The paper focuses on modeling and simulation of a heavy-duty truck and an inverted pendulum, and experimental validation on a real truck.
Key Insights and Ideas
- CBFs as a Practical Safety Layer: CBFs offer an intuitive and constructive way to enforce safety, acting as a "safety filter" that minimally perturbs a performance-oriented controller.
- ISSf for Robustness: ISSf provides a crucial theoretical framework for understanding and quantifying how disturbances affect safety guarantees, allowing for the design of controllers that are robust to these uncertainties.
- Design-Test-Redesign Loop: The paper highlights the effectiveness of a practical design cycle where experimental observations of disturbances inform the robust controller design, bridging the gap between theory and real-world application.
- Automated Truck as a Challenging Testbed: Applying CBFs and ISSf to a heavy-duty automated truck demonstrates the scalability and practical relevance of the approach for complex, real-world systems.
- Trade-off Between Conservativeness and Performance: The choice of parameters in ISSf-CBFs allows for tuning the trade-off between the strength of the safety guarantee (size of $C_\delta$) and the controller's conservativeness, impacting performance metrics like tracking distance.
Key Results
- Inverted Pendulum Example: The paper successfully demonstrates that a nominal controller can lead to safety violations, while a CBF-filtered controller keeps the system within the safe set. The ISSf framework is also illustrated with a simulated disturbance.
- Automated Truck Simulation: Simulations show that the nominal controller fails to maintain safe following distances during a hard braking event, while the safety-critical controller (using CBFs) successfully avoids collisions.
- Automated Truck Experimentation:
- Experimental results confirm that the nominal controller and even the basic safety-critical controller can fail to meet safety requirements due to unmodeled input disturbances.
- The robust safety-critical controller (using ISSf-CBFs) successfully maintains safety (ensuring the CBF value remains positive) even in the presence of estimated input disturbances, both in simulation and experiments.
- The robust controller achieves safety without significantly compromising performance, as evidenced by the steady-state tracking distance.
Limitations and Future Work
- Disturbance Bound Estimation: The paper acknowledges that estimating the worst-case disturbance bound ($\delta$) can be challenging in practice, especially for complex systems like the truck's drive-by-wire system. Future work could explore more systematic methods for estimating these bounds.
- Controller Conservativeness: While ISSf-CBFs provide robustness, they can introduce conservativeness. The paper shows how parameter tuning can mitigate this, but further research could focus on optimizing this trade-off more effectively.
- Higher-Order Systems and Complex Dynamics: While the paper applies the framework to a simplified truck model, extending it to more complex vehicle dynamics or higher-order systems could be a future direction.
- Experimental Validation of ISSf: While the paper demonstrates the effectiveness of the robust controller experimentally, a more direct validation of the ISSf guarantees themselves (i.e., demonstrating the forward invariance of $C_\delta$) would be valuable.
Summary updated