[Paper Review] Safety-Critical Lane-Change Control for CAV Platoons in Mixed Autonomy Traffic Using Control Barrier Functions

Fengqing Hu, Huan Yu|arXiv (Cornell University)|Feb 1, 2023

Traffic control and management4 citations

TL;DR

This paper proposes a safety-critical lane-change control framework for connected and autonomous vehicle (CAV) platoons in mixed autonomy traffic using Control Barrier Functions (CBFs) and Control Lyapunov Functions (CLFs). The method employs a two-layer controller—higher-level for lane-change decision and lower-level for safety-critical kinematic control—ensuring both longitudinal and lateral safety during maneuvers, with simulations showing collision avoidance in four safety-critical scenarios.

ABSTRACT

Platooning can serve as an effective management measure for connected and autonomous vehicles (CAVs) to ensure overall traffic efficiency. Current study focus on the longitudinal control of CAV platoons, however it still remains a challenging problem to stay safe under lane-change scenarios where both longitudinal and lateral control is required. In this paper, a safety-critical control method is proposed conduct lane-changing maneuvers for platooning CAVs using Control Barrier Functions (CBFs). The proposed method is composed of two layers: a higher-level controller for general lane change decision control and a lower-level controller for safe kinematics control. Different from traditional kinematics controllers, this lower-level controller conducts not only longitudinal safety-critical control but also critically ensures safety for lateral control during the platooning lane change. To effectively design this lower-level controller, an optimization problem is solved with constraints defined by both CBFs and Control Lyapunov Functions (CLFs). A traffic simulator is used to conduct numerical traffic simulations in four safety-critical scenarios and showed the effectiveness of the proposed controller.

Motivation & Objective

• Address the challenge of safe lane-changing in CAV platoons where both longitudinal and lateral control are required.
• Ensure safety in mixed autonomy traffic where human-driven vehicles (HDVs) may exhibit unpredictable or aggressive behaviors.
• Design a control framework that guarantees safety through barrier functions while maintaining platoon stability via Lyapunov functions.
• Enable cooperative platoon operations (split and join) to facilitate safe lane changes by adjusting inter-vehicle spacing.
• Demonstrate the effectiveness of the proposed controller in safety-critical scenarios through numerical simulation.

Proposed method

• Develop a two-layer control architecture: a higher-level controller for lane-change decision-making and a lower-level controller for safety-critical kinematic control.
• Formulate the lower-level controller as a quadratic program (QP) with constraints derived from both Control Barrier Functions (CBFs) and Control Lyapunov Functions (CLFs).
• Define CBF constraints to ensure safety by preventing collisions in both longitudinal and lateral directions during lane changes.
• Use CLF constraints to ensure asymptotic stability and desired convergence to the target state in the control trajectory.
• Integrate split and join maneuvers into the platoon operation to increase headway before lane changes and rejoin after completion.
• Solve the optimization problem online to generate real-time control inputs that satisfy safety and stability requirements.

Experimental results

Research questions

• RQ1 How can CAV platoons perform safe lane changes in mixed autonomy traffic where HDVs may act unpredictably?
• RQ2 What control framework ensures both longitudinal and lateral safety during platoon lane changes using CBFs and CLFs?
• RQ3 How does the proposed two-layer controller compare to a CLF-only controller in terms of safety and feasibility in critical scenarios?
• RQ4 Can cooperative platoon operations (split and join) improve the success rate and safety of lane changes in dense traffic?
• RQ5 What are the performance limitations of the controller under communication delays or infeasible QP conditions?

Key findings

• The CLF-CBF-QP controller successfully completed all four safety-critical lane-change scenarios (cut-in, FDEC, BACC, FFDEC) without collisions.
• In the BACC scenario, the CLF-CBF-QP controller avoided collision by switching back to the original lane when a rear vehicle accelerated, while the CLF-QP controller failed due to oscillation after collision.
• In the FFDEC scenario, the platoon-based CLF-CBF-QP controller succeeded in lane changing due to coordinated deceleration during split operations, whereas the single-vehicle CLF-CBF-QP controller failed due to QP infeasibility at t=2.2s.
• The failure in the single-vehicle controller was caused by conflicting control objectives: all CAVs accelerated to reach desired speed, increasing collision risk, while CAV1 decelerated to avoid FCV, and CAV3 did not adjust, reducing spacing.
• The proposed cooperative platoon operation (split and join) effectively increased headway and enabled safer lane changes by proactively managing inter-vehicle spacing.
• The integration of CBFs for lateral and longitudinal safety, combined with CLFs for stability, ensured both safety and convergence in all tested scenarios.