Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

C-TRAIL: A Commonsense World Framework for Trajectory Planning in Autonomous Driving

Zhihong Cui, Haoran Tang, Tianyi Li, Yushuai Li, Peiyuan Guan, Amir Taherkordi, Tor Skeie

Abstract

Trajectory planning for autonomous driving increasingly leverages large language models (LLMs) for commonsense reasoning, yet LLM outputs are inherently unreliable, posing risks in safety-critical applications. We propose C-TRAIL, a framework built on a Commonsense World that couples LLM-derived commonsense with a trust mechanism to guide trajectory planning. C-TRAIL operates through a closed-loop Recall, Plan, and Update cycle: the Recall module queries an LLM for semantic relations and quantifies their reliability via a dual-trust mechanism; the Plan module injects trust-weighted commonsense into Monte Carlo Tree Search (MCTS) through a Dirichlet trust policy; and the Update module adaptively refines trust scores and policy parameters from environmental feedback. Experiments on four simulated scenarios in Highway-env and two real-world levelXData datasets (highD, rounD) show that C-TRAIL consistently outperforms state-of-the-art baselines, reducing ADE by 40.2%, FDE by 51.7%, and improving SR by 16.9 percentage points on average. The source code is available at https://github.com/ZhihongCui/CTRAIL.

C-TRAIL: A Commonsense World Framework for Trajectory Planning in Autonomous Driving

Abstract

Trajectory planning for autonomous driving increasingly leverages large language models (LLMs) for commonsense reasoning, yet LLM outputs are inherently unreliable, posing risks in safety-critical applications. We propose C-TRAIL, a framework built on a Commonsense World that couples LLM-derived commonsense with a trust mechanism to guide trajectory planning. C-TRAIL operates through a closed-loop Recall, Plan, and Update cycle: the Recall module queries an LLM for semantic relations and quantifies their reliability via a dual-trust mechanism; the Plan module injects trust-weighted commonsense into Monte Carlo Tree Search (MCTS) through a Dirichlet trust policy; and the Update module adaptively refines trust scores and policy parameters from environmental feedback. Experiments on four simulated scenarios in Highway-env and two real-world levelXData datasets (highD, rounD) show that C-TRAIL consistently outperforms state-of-the-art baselines, reducing ADE by 40.2%, FDE by 51.7%, and improving SR by 16.9 percentage points on average. The source code is available at https://github.com/ZhihongCui/CTRAIL.

Paper Structure

This paper contains 23 sections, 15 equations, 7 figures, 10 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. Definition
  4. Problem Definition
  5. Model
  6. Trust-Aware Commonsense Recall
  7. Trust-Guided Planning
  8. Trust Calibration Update
  9. Joint Training
  10. Deployment Architecture
  11. Experiments
  12. Overall Settings
  13. Trajectory Planning Performance
  14. Ablation Study
  15. Safety Analysis
  16. ...and 8 more sections

Figures (7)

  • Figure 1: The commonsense world paradigm for trajectory planning in dynamic environments
  • Figure 2: The C-TRAIL, A Commonsense World Framework for Trajectory Planning in Autonomous Driving
  • Figure 3: Commonsense Graph Prompt.
  • Figure 4: Examples of correct and incorrect LLM outputs.
  • Figure 5: Trust score dynamics ($C_t$) under normal and error injection conditions on highway. Errors are injected at $t{=}4$.
  • ...and 2 more figures