Evidential-Based Approach for Trajectory Planning With Tentacles, for Autonomous Vehicles

Mouhagir, Hafida; Talj, Reine; Cherfaoui, Veronique; Aioun, Francois; Guillemard, Franck · 2020 · Crossref

DOI: 10.1109/tits.2019.2930035

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 reactive local trajectory planning for autonomous vehicles operating in uncertain, dynamic environments. The authors argue that traditional probabilistic occupancy grids fail to distinguish between different types of uncertainty, such as ignorance (lack of data) and conflict (contradictory data). To resolve this, the study proposes an evidential-based approach using Dempster-Shafer theory to represent environmental uncertainty, combined with a clothoid tentacle method for trajectory generation and a Markov Decision Process (MDP) for decision-making. The methodology involves three main steps. First, the system generates a set of "clothoid tentacles"—smooth curves with linearly varying curvature—in the vehicle’s egocentric reference frame. These tentacles represent dynamically feasible trajectories that respect the vehicle’s kinematic limits and current state. Second, the environment is modeled using an evidential occupancy grid. Unlike probabilistic grids, this grid assigns belief masses to four states: Free, Occupied, Unknown (ignorance), and Conflict. The grid is constructed by fusing sensor data, road limit information, and longitudinal safety distance expansions for dynamic obstacles. Third, the best trajectory is selected using an MDP framework. Each tentacle is evaluated based on three criteria: occupancy (using the evidential grid to assess risk), distance from a global reference trajectory, and traffic rules (specifically overtaking safety). The occupancy reward utilizes a conjunctive rule to combine belief masses from grid cells intersecting the tentacle, allowing the system to differentiate between unknown space and conflicting evidence. The approach was validated using the SCANeR™Studio driving simulator in scenarios involving uncertain dynamic environments and multiple vehicles. The results demonstrate that the evidential grid provides significant advantages over probabilistic methods by explicitly handling ignorance and conflict. For instance, the system can distinguish between an unobserved cell (requiring exploration) and a cell with contradictory evidence (indicating potential danger), leading to more appropriate vehicle reactions. The clothoid tentacles ensure smooth, kinematically feasible paths, while the MDP-based selection ensures safety and compliance with traffic rules. The study concludes that integrating belief functions into trajectory planning allows autonomous vehicles to navigate safely and efficiently in partially observable environments, addressing the limitations of classical probability theory in representing complex uncertainty.

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.

• anticipation

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Theoretical Contribution: computational model