HEAT: Heterogeneous End-to-End Autonomous Driving via Trajectory-Guided World Models

arXiv 2026 50.1 method, application

TLDR

A trajectory-guided world model enables a single end-to-end autonomous driving model to perform well across heterogeneous domains without retraining.

Reasoning

The paper addresses a practical multi-domain learning challenge in autonomous driving and proposes a novel trajectory-driven paradigm with a world model for feature consistency. Strengths include real-world benchmarks and clear problem motivation; weaknesses are limited methodological detail in the abstract and no explicit comparison to world simulator or RL approaches.

Read-first score

Read-first score 50.1, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 24.

Recency 6%

100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Citation impact 18%

77.5

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.77453184

Methodology quality 18%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=benchmark,dataset,result

Reproducibility 18%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=code,dataset

Topical relevance 29%

34.3

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Citation velocity 12%

Citation velocity estimates citations per publication-year to reduce old-paper bias. velocity=0.00

Field roles

FoundationFrontierBridgeMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 283.

Keyword Scores

world model

world dynamics prediction

generative world model

model-based reinforcement learning world model

world simulator

interactive world model

video world model

Deep Analysis

Innovations

Trajectory-driven learning paradigm that organizes training around planning trajectories to capture domain-invariant representations of driving intent.
Incorporation of a world model that predicts future latent features conditioned on ego actions to improve feature consistency and mitigate domain-induced biases.

Methodology

The paper proposes an end-to-end autonomous driving model that uses a trajectory-guided learning paradigm to organize training around planning trajectories, enabling capture of domain-invariant driving intent representations. It also incorporates a world model that predicts future latent features conditioned on ego actions to improve feature consistency and reduce domain biases. The model is trained jointly on multiple heterogeneous datasets (nuScenes, NAVSIM, Waymo) without domain-specific retraining.

Key Results

The approach shows substantial improvements over existing methods across all three benchmarks (nuScenes, NAVSIM, Waymo), demonstrating that a single unified model can be trained on heterogeneous datasets while maintaining strong performance within each domain.

Resources

Paper