PV-WM: A Heterogeneous Micro-Macro World Model for Articulated Pedestrian-Vehicle Co-Rollout

arXiv 2026 34.4 method, application

TLDR

PV-WM is a heterogeneous world model for co-rolling pedestrian articulation and vehicle states, reducing errors on Waymo with fewer parameters than modular specialists.

Reasoning

The paper introduces a novel recurrent world model that jointly predicts pedestrian articulation and vehicle states, with strong empirical gains over baselines on Waymo and clear efficiency benefits. However, the abstract lacks details on training procedures, baseline strengths, and potential failure modes.

Read-first score

Read-first score 34.4, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 38.

Recency 5%

100

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

Methodology quality 15%

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

Topical relevance 25%

54.3

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

Reproducibility 15%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=none

Graph prestige 15%

5.7

PageRank-style prestige over the local citation graph built from available OpenAlex references. local_pagerank=0.0016

Citation impact 15%

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. cited_by_count=0

Citation velocity 10%

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

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 278.

Keyword Scores

world model

world dynamics prediction

generative world model

world simulator

interactive world model

video world model

model-based reinforcement learning world model

Deep Analysis

Innovations

History-only heterogeneous micro-macro world model over structured post-perception tracks for articulated pedestrian-vehicle co-rollout
Synchronized heterogeneous state that recurrently advances pedestrian root motion, 15-joint articulation, and learned vehicle states
Self-feeding recurrent rollout where generated pedestrian and vehicle chunks supply the next recurrent boundary, with vehicle boxes reconstructed from predicted center and heading using observed extent and P-V geometry recomputed after every transition
Single-network formulation that unifies heterogeneous future state while preserving type-specific pedestrian and vehicle dynamics

Methodology

PV-WM is a recurrent world model over structured post-perception tracks that advances pedestrian root motion, 15-joint articulation, and learned vehicle states in a synchronized heterogeneous state. Generated pedestrian and vehicle chunks become the next recurrent boundary; vehicle boxes are reconstructed from predicted center and heading using observed extent, and pedestrian-vehicle geometry is recomputed after each transition. Evaluation uses aligned Waymo contexts and compares against a matched one-shot complete-state predictor and a validation-selected Modular Specialist using pedestrian, vehicle, and pedestrian-vehicle metrics plus compute efficiency measures.

Key Results

Recurrent execution reduces Root ADE by 12.7% and MPJPE by 14.8% relative to a matched one-shot complete-state predictor, and across 824 aligned Waymo contexts with 797 valid future vehicle support, PV-WM reduces Root ADE by 5.2%, MPJPE by 7.6%, P-V distance error by 11.9%, and oriented-box closest-approach error by 5.8% relative to a validation-selected Modular Specialist. The single-network model uses 57.1% fewer parameters, 96.5% lower average FLOPs per local scene, and 25.5% lower measured p95 latency; feedback interventions show later predictions depend on generated articulation content, temporal order, and pedestrian identity.

Resources

Paper