On path planning methods for automotive collision avoidance

Madas, David; Nosratinia, Mohsen; Keshavarz, Mansour; Sundstrom, Peter; Philippsen, Rolland; Eidehall, Andreas; Dahlen, Karl-Magnus · 2013 · Crossref

DOI: 10.1109/ivs.2013.6629586

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 evaluates three distinct path planning methods for automotive collision avoidance systems, specifically targeting applications in Autonomous Emergency Braking (AEB) and Lane Keeping Aid (LKA). The research addresses the growing complexity of Advanced Driver Assistance Systems (ADAS), which require efficient online maneuver generation capable of handling diverse traffic scenarios and varying levels of maneuver severity. The study aims to provide qualitative and quantitative insights to guide the selection of appropriate planning algorithms, rather than declaring a single superior method. The problem is defined as finding a collision-free path while keeping the vehicle within road boundaries, assuming constant longitudinal velocity to isolate steering interventions. The authors compare three approaches: State Lattice planning, Predictive Constraint-Based Planning (using Model Predictive Control), and Spline-Based Search Tree methods. The State Lattice method discretizes the state space into a graph where nodes represent vehicle states and edges represent controllable trajectories, solved via graph search. The Predictive Constraint-Based method utilizes a "Safe Corridor" concept, where lateral position constraints derived from sensor data are used in an MPC formulation to optimize steering inputs while minimizing side slip angle and respecting dynamic limits. The Spline-Based method employs a search tree to generate smooth trajectories using cubic or quintic polynomials, designed to minimize jerk by ensuring continuous derivatives across path segments. The methods were evaluated in two specific scenarios: a highway situation involving a stationary obstacle at 80 km/h, and a city driving scenario involving a bicyclist and a stationary vehicle at 50 km/h. Results were analyzed based on lateral displacement, acceleration, and jerk. The State Lattice produced smooth paths but required careful tuning of spatial resolution to balance optimality and computational load. The MPC approach effectively managed constraints and provided smooth control inputs but required significant computational resources for real-time optimization. The Spline-Based methods, particularly the quintic splines, offered continuous jerk profiles, resulting in smoother paths compared to cubic splines, though they required solving for transition points based on object tangency. The study concludes that the choice of method depends on specific application requirements, including computational constraints, desired optimality criteria, and scenario complexity. State Lattices offer pre-computation benefits but struggle with varying road geometries. MPC provides robust constraint handling and dynamic fidelity but is computationally intensive. Spline-based methods offer high smoothness and low jerk, suitable for driver comfort, but may be less flexible in complex dynamic environments. The paper provides a comparative framework to assist developers in selecting the most appropriate algorithm for specific ADAS functions.

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.

• adas effectiveness