Safe and efficient lane-changing for autonomous vehicles: an improved double quintic polynomial approach with time-to-collision evaluation

Rui Bai, Rui Xu, Teng Rui, Jiale Liu, Hoi Leong Lee, Qi Wei Oung, Zhen Tian, Fujiang Yuan

TL;DR

Autonomous vehicles need to change lanes safely and comfortably in complex traffic. This paper introduces an improved double quintic polynomial trajectory planner that directly integrates a Time-to-Collision (TTC) safety evaluation into the planning process, ensuring proactive safety without sacrificing efficiency or comfort.

What Problem Does This Paper Address?

Autonomous vehicles (AVs) are rapidly advancing, but safe and comfortable interactions with human-driven vehicles (HDVs), especially during maneuvers like lane-changing, remain a significant challenge. Traditional trajectory planning methods often struggle with dynamic environments, either requiring frequent replanning or lacking explicit safety guarantees. This can lead to overly conservative or aggressive maneuvers, compromising safety, efficiency, and passenger comfort. The paper specifically addresses the need for a method that can generate safe, efficient, and comfortable lane-changing trajectories in mixed traffic, where AVs must coexist with unpredictable human drivers.

Methods

The core of the proposed method is an improved double quintic polynomial approach for trajectory generation. This builds upon existing double quintic polynomial methods by:

  • Integrating Time-to-Collision (TTC) directly into the optimization: Instead of a post-hoc safety check, a TTC-based penalty is embedded into the cost function of the trajectory planner. This means safety is a primary objective during trajectory generation, not an afterthought.
  • Constrained Optimization: The coefficients of the quintic polynomials are determined by minimizing a joint cost function (smoothness + TTC penalty) subject to boundary conditions, rather than a direct algebraic solution. This allows for explicit safety and feasibility constraints.
  • Differentiable Safety Constraint: A piecewise-quadratic penalty function is used for TTC violations, which is differentiable. This enables stable gradient-based optimization.
  • Double Quintic Polynomials: The lane-changing maneuver is decomposed into two quintic polynomial segments, allowing for more explicit control over different phases of the maneuver (initiation, transition, completion).

Datasets

The paper does not explicitly mention training datasets. For evaluation, they use:

  • Simulated Scenarios: Extensive simulations were conducted in MATLAB, covering various ramp variations, lane-changing timing variants, overtaking scenarios, and complex mixed-traffic environments.
  • NGSIM Dataset: For some simulations, real-world traffic trajectories from the NGSIM dataset were replayed to simulate more realistic interactions with surrounding vehicles, including human driving artifacts.

Key Insights and Ideas

  • Proactive Safety is Key: Embedding safety (TTC) directly into the trajectory generation process, rather than just validating it afterward, is crucial for robust and reliable autonomous driving. This shifts safety from a reactive measure to a proactive design principle.
  • Balancing Multiple Objectives: The proposed method effectively balances safety, comfort, and efficiency. It achieves this by jointly optimizing for smoothness (via jerk minimization) and collision avoidance (via TTC penalty), demonstrating that these objectives are not mutually exclusive.
  • Double Quintic for Maneuver Control: Decomposing the lane-change into two quintic segments provides finer control over the maneuver's dynamics, allowing for smoother and more adaptable trajectories.
  • TTC as a Soft Constraint: Using a differentiable TTC penalty rather than hard constraints allows the system to gracefully push trajectories away from unsafe regions without causing infeasibility, leading to more robust and less brittle behavior.

Key Results

  • Reduced Low-TTC Time Steps: The proposed method reduced low-TTC (<3s) time steps by approximately 35% compared to traditional methods.
  • Increased Minimum Obstacle Gap: The minimum obstacle gap was more than doubled, from 8.75m to 19.53m.
  • Shorter Maneuver Time: Total maneuver time was reduced from 13s to 8s while maintaining lower curvature (40-60%).
  • Improved Comfort: Lateral jerk was significantly reduced (peak jerk by 37.3%, RMS jerk by 38.6%), leading to smoother and more comfortable trajectories.
  • Generalization: The method demonstrated effective performance across various scenarios, including lane-changing, overtaking, and complex mixed-traffic interactions with diverse vehicle behaviors (braking, speed oscillations, real-world trajectories).
  • Numerical Stability: The optimization process converged reliably within 15-25 iterations, with monotonic cost reduction, indicating good numerical stability.

Limitations and Future Work

  • Simplified Vehicle Dynamics: The current model assumes constant longitudinal velocity and neglects lateral slip. Future work will incorporate more realistic vehicle dynamics, including steering constraints, maximum lateral/longitudinal accelerations, and tire-road interactions.
  • Simulation-Based Validation: While extensive simulations were performed, real-world testing is the next crucial step to validate the method's performance in actual driving conditions.
  • Uncertainty in Driving Environment: Future research will focus on addressing uncertainties in the driving environment, such as variability in human behavior, sensor noise, and environmental factors.
  • Parameter Tuning: While the paper provides guidance, adaptive parameter selection based on traffic density, vehicle capabilities, and mission requirements will be important for real-world deployment.

Summary updated

Abstract

Autonomous driving technology has made significant advancements in recent years, yet challenges remain in ensuring safe and comfortable interactions with human-driven vehicles (HDVs), particularly during lane-changing maneuvers. This paper proposes an improved double quintic polynomial approach for safe and efficient lane-changing in mixed traffic environments. The proposed method integrates a time-to-collision (TTC) based evaluation mechanism directly into the trajectory optimization process, ensuring that the ego vehicle proactively maintains a safe gap from surrounding HDVs throughout the maneuver. The framework comprises state estimation for both the autonomous vehicle (AV) and HDVs, trajectory generation using double quintic polynomials, real-time TTC computation, and adaptive trajectory evaluation. To the best of our knowledge, this is the first work to embed an analytic TTC penalty directly into the closed-form double-quintic polynomial solver, enabling real-time safety-aware trajectory generation without post-hoc validation. Extensive simulations conducted under diverse traffic scenarios demonstrate the safety, efficiency, and comfort of the proposed approach compared to conventional methods such as quintic polynomials, Bezier curves, and B-splines. The results highlight that the improved method not only avoids collisions but also ensures smooth transitions and adaptive decision-making in dynamic environments. This work bridges the gap between model-based and adaptive trajectory planning approaches, offering a stable solution for real-world autonomous driving applications.
BibTeX
@article{bai2026safe,
  author = {Rui Bai and Rui Xu and Teng Rui and Jiale Liu and Hoi Leong Lee and Qi Wei Oung and Zhen Tian and Fujiang Yuan},
  title = {{Safe and efficient lane-changing for autonomous vehicles: an improved double quintic polynomial approach with time-to-collision evaluation}},
  journal = {J. King Saud Univ. Comput. Inf. Sci.},
  year = {2026},
  doi = {10.1007/s44443-025-00300-2},
}
Figure