Efficient Speed Planning in the Path-Time Space for Urban Autonomous Driving

Duhautbout, Thibaud; Talj, Reine; Cherfaoui, Veronique; Aioun, Francois; Guillemard, Franck · 2022 · Crossref

DOI: 10.1109/itsc55140.2022.9921820

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of efficient speed planning for autonomous vehicles in dynamic urban environments, where constant adaptation to moving obstacles is required. While combined motion planning and path planning are well-studied, the authors note that dedicated speed planning algorithms often lack flexibility or computational efficiency. The study proposes a novel algorithmic method that computes safe and efficient speed profiles on predefined paths by utilizing the path-time space to represent interactions with other dynamic actors. This approach allows the vehicle to make behavioral decisions, such as passing before or after an obstacle, while respecting safety distances and physical constraints. The methodology relies on a path-time diagram where the horizontal axis represents the vehicle's position along a path and the vertical axis represents time. The authors define "collision zones" in this space based on motion predictions of other vehicles. These zones are discretized into a "collision matrix" using polygon intersection checks, which marks intervals of position and time that must be avoided. The speed planning algorithm first generates a comfortable trajectory and checks its validity against the collision matrix. If a collision is detected, the algorithm adjusts the trajectory using two primary strategies: "pass-before," which accelerates the vehicle to clear the obstacle earlier, and "pass-after," which decelerates to let the obstacle pass. These adjustments are computed using dichotomy searches to find minimal acceleration or braking values that satisfy safety constraints. The algorithm is designed to handle multiple obstacles by recursively applying these adjustments to segments of the path. Simulation results were conducted using MATLAB/Simulink and SCANeR Studio to evaluate the method in a four-branch crossing scenario. The proposed method was compared against a previous "stop-point" method. In scenarios involving single and multiple moving obstacles, the path-time method demonstrated superior performance by anticipating obstacles earlier and maintaining higher average speeds. Specifically, in a scenario with one obstacle, the proposed method achieved a 0.22-second faster maneuver with less braking and acceleration effort. In a complex scenario involving a vehicle and a pedestrian, the algorithm successfully navigated between the two obstacles by passing before the vehicle and slowing for the pedestrian, whereas the previous method failed to find a valid trajectory and stopped unnecessarily. Processing time analysis indicated that the algorithm operates efficiently, with most computations completed within milliseconds, making it suitable for high-frequency replanning. The significance of this work lies in providing a computationally efficient and flexible framework for speed planning that explicitly handles dynamic obstacles in urban settings. By leveraging the path-time space, the method enables autonomous vehicles to make nuanced driving decisions, such as overtaking or yielding, rather than simply stopping. This contributes to smoother, more natural driving behaviors and improved traffic flow in complex environments. The study demonstrates that separating speed planning from path planning and using spatiotemporal representations can yield better results than traditional optimization or graph-based approaches, particularly in terms of computational cost and behavioral flexibility.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• speed choice