safe-cooperative-control-av-itsc2024

Safe Cooperative Control for AVs at Highway Ramps and Intersections

A leader-follower cooperative control framework for automated vehicles (AVs), combining second-order consensus control with a control-barrier- function (CBF) based safety filter to guarantee collision-free merging and intersection crossing.

Role & Attribution

Sidra Ghayour Bhatti — co-author, methodology and analysis. Lead author: Pei Yu Chang. Co-author: Nur Uddin Javed. Supervision: Qadeer Ahmed (PI, OSU Center for Automotive Research).

Abstract

This article focuses on improving the safety of real-world scenarios such as highway merging and intersections by using cooperative control strategies for automated vehicles (AVs). We utilize a cooperative control approach, integrating a safety filter to ensure that all vehicles can avoid conflicts in challenging environments. A leader-follower pattern is adopted in the framework, ensuring all AVs reach a consistent status with the leader reference AV. The safety filter utilizes control barrier functions (CBF) and a proposed safety set function to derive safe control inputs. To demonstrate the practicality of the safe cooperative framework, the simulation is implemented across three carefully chosen traffic scenarios: highway merging, opposite-direction merging, and a complex four-leg intersection, reflecting the diverse challenges encountered in real-world transportation systems. Implementing the safe cooperative controller achieves zero conflict rates in all three scenarios, demonstrating its effectiveness in preventing conflicts.

Method

Vehicles communicate over a leader-follower graph topology (encoded as a Laplacian matrix), and each AV is modeled as a double-integrator point mass. A second-order consensus law generates a nominal cooperative control input driving all followers to match the reference (leader) AV's state. A control barrier function (CBF)-based safety filter, built on a proposed safety-set function of relative position/velocity and braking deceleration, is layered on top: a quadratic program (QP) finds the control input closest to the nominal cooperative input while satisfying the CBF safety constraint, guaranteeing forward invariance of the safe set (i.e., provable collision avoidance) under the given dynamics.

Fig. 1: Framework of safe cooperative control — the vehicle model feeds AV state to a leader-follower cooperative controller and a safe-set function; a QP solver then finds the closest safe control input to the nominal cooperative command.

Results

Three traffic scenarios were simulated, each with and without the safety filter:

The safety filter reduces the conflict rate to zero in every scenario, including Scenario II where conflicts occurred at over 80% of waypoints without it. The trade-off: vehicles must deviate from their nominal paths to avoid collisions, which can increase control effort and travel time — noted as a direction for future work (e.g., adding a kinematic model and exploring alternative safety-set functions).

Fig. 4–5: Highway merging scenario without (left) and with (right) the safety filter. Without it, AV trajectories cross through several conflict points (×) near x = 0–10 m; with the safety filter applied, all AVs reach consensus with the reference AV while staying outside the conflict-prone region.

Published: P. Y. Chang, S. G. Bhatti, N. U. Javed, and Q. Ahmed, "Enhancing Safety at Highway Ramps and Intersections Using Safe Cooperative Controls in Automated Vehicles," 2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), Edmonton, Canada, pp. 902–907, 2024. DOI: 10.1109/ITSC58415.2024.10920020

Citation

@inproceedings{chang2024safecoop,
  author    = {Chang, Pei Yu and Bhatti, Sidra Ghayour and Javed, Nur Uddin and Ahmed, Qadeer},
  title     = {Enhancing Safety at Highway Ramps and Intersections Using Safe Cooperative Controls in Automated Vehicles},
  booktitle = {2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC)},
  pages     = {902--907},
  year      = {2024},
  doi       = {10.1109/ITSC58415.2024.10920020}
}