Fast and Safe Path-Following Control using a State-Dependent Directional Metric

Zhichao Li, Omur Arslan, Nikolay Atanasov

Abstract

This paper considers the problem of fast and safe autonomous navigation in partially known environments. Our main contribution is a control policy design based on ellipsoidal trajectory bounds obtained from a quadratic state-dependent distance metric. The ellipsoidal bounds are used to embed directional preference in the control design, leading to system behavior that is adapted to the local environment geometry, carefully considering medial obstacles while paying less attention to lateral ones. We use a virtual reference governor system to adaptively follow a desired navigation path, slowing down when system safety may be violated and speeding up otherwise. The resulting controller is able to navigate complex environments faster than common Euclidean-norm and Lyapunov-function-based designs, while retaining stability and collision avoidance guarantees.

Introduction: Overcoming the "Corridor Effect"

A persistent challenge in autonomous robotics is the "corridor effect," where a robot slows significantly when navigating narrow passages like doorways or hallways. This behavior often stems from the use of non-directional distance metrics, such as the Euclidean norm, to guarantee safety. By treating all obstacles within a fixed radius as equally hazardous, standard controllers become overly cautious in environments where lateral obstacles (walls) are close but pose little risk to the robot's forward progress.

Methodology: State-Dependent Directional Metrics

To address this, we propose a two-layer control architecture centered on a State-Dependent Directional Metric (SDDM). Instead of a static safety circle, we employ a quadratic norm that generates an ellipsoidal safety bound. This ellipsoid is dynamically elongated in the direction of the robot's motion, allowing the system to prioritize "medial" obstacles directly in the path of travel while effectively ignoring "lateral" obstacles that the robot is moving parallel to.

Autonomous navigation in a complex environment using the proposed SDDM controller, showing the ellipsoidal trajectory bounds.

We utilize Semi-Definite Programming (SDP) to calculate tight ellipsoidal bounds on the robot’s future positions. These bounds represent the reachable set of the robot given its current dynamics. A virtual "Reference Governor" acts as a middleman, moving along the desired path and regulating the robot's speed. If the safety margin—the "leeway" between the ellipsoidal bound and an obstacle—shrinks, the governor slows down, ensuring the robot maintains stability and safety before proceeding.

Key Insight: Directionality as a Driver for Efficiency

The core advantage of the SDDM is its ability to decouple safety from proximity. In a narrow corridor, a Euclidean metric triggers a slowdown because the walls are physically close. In contrast, the SDDM stretches the metric along the velocity vector. As shown in the comparison below, the directional metric (Q-Dist) provides a larger safety margin for the same physical distance to a wall compared to the Euclidean metric (E-Dist).

Comparison of the proposed State-Dependent Directional Metric (Q-Dist) versus the standard Euclidean distance (E-Dist). The directional metric allows for higher speeds by reducing sensitivity to lateral obstacles.

Experimental Results

The proposed controller was evaluated in several simulated environments, demonstrating significant performance gains:

  • Significant Speedup: In a field of circular obstacles, the SDDM controller completed the path in 23.80 seconds, a 35% improvement over standard Lyapunov-based designs (36.45 seconds).
  • Robustness to Unknowns: The system successfully navigated complex spaces using real-time LIDAR data, proving that the SDDM can be integrated with local occupancy grids and dynamic re-planning (e.g., A*).
  • Motion Quality: The directional controller produced smoother trajectories, particularly during turn entries and exits, as it avoided the "panic" slowdowns typical of non-directional safety metrics.

Conclusion and Future Work

By embedding directional preference into the safety metric, we enable robots to navigate tight spaces with significantly higher efficiency without sacrificing safety guarantees. While current results rely on linearized dynamics and pre-defined paths, future work will focus on extending these proofs to highly non-linear systems and validating the approach on physical hardware to account for real-world sensor noise and actuator constraints.

Summary updated

Abstract

This paper considers the problem of fast and safe autonomous navigation in partially known environments. Our main contribution is a control policy design based on ellipsoidal trajectory bounds obtained from a quadratic state-dependent distance metric. The ellipsoidal bounds are used to embed directional preference in the control design, leading to system behavior that is adapted to the local environment geometry, carefully considering medial obstacles while paying less attention to lateral ones. We use a virtual reference governor system to adaptively follow a desired navigation path, slowing down when system safety may be violated and speeding up otherwise. The resulting controller is able to navigate complex environments faster than common Euclidean-norm and Lyapunov-function-based designs, while retaining stability and collision avoidance guarantees.
BibTeX
@inproceedings{li2020fast,
  author = {Zhichao Li and Omur Arslan and Nikolay Atanasov},
  title = {{Fast and Safe Path-Following Control using a State-Dependent Directional Metric}},
  booktitle = {IEEE International Conference on Robotics and Automation},
  year = {2020},
  url = {https://openreview.net/forum?id=PNzQ23eny8zA},
}
Figure