DeTrack: A Benchmark and Altitude-Aware Dual World Model for Drone-embodied Tracking

arXiv 2026 59.4 method, benchmark, application

TLDR

A benchmark and altitude-aware dual world model for drone-embodied tracking in interactive 3D environments.

Reasoning

The paper introduces a novel task and benchmark (DeTrack) with a large-scale dataset, and proposes AaDWorlds, a dual world model that addresses altitude-mediated trade-offs. Strengths include a well-defined problem and comprehensive evaluation; weaknesses include lack of real-world validation and potential over-reliance on simulated environments.

Read-first score

Read-first score 59.4, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 43.

Recency 6%

100

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

Citation impact 18%

91.1

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.91088706

Methodology quality 18%

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

Topical relevance 29%

61.4

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 18%

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

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: 394.

Keyword Scores

world model

interactive world model

generative world model

world dynamics prediction

model-based reinforcement learning world model

video world model

world simulator

Deep Analysis

Innovations

Definition of a new drone-embodied tracking task (DeTrack) requiring active perception and closed-loop flight control in interactive 3D environments
Construction of a large-scale benchmark with 11,368 target trajectories across diverse scenes, rendering conditions, semantic regions, and moving distractors, along with evaluation metrics for target visibility, tracking accuracy, and trajectory success
Proposal of AaDWorlds, an altitude-aware dual world model framework that uses pseudo altitude-aware observations and imagined future states to balance target visibility and flight safety

Methodology

The paper introduces the DeTrack benchmark and the AaDWorlds framework. AaDWorlds comprises an altitude-aware perception module and dual world models that generate imagined future states under both high- and low-altitude regimes. The framework combines these pseudo altitude-aware observations with imagined states to address the altitude-mediated trade-off between target visibility and flight safety. The benchmark includes 11,368 trajectories with diverse scenes, rendering conditions, semantic regions, and moving distractors, and uses metrics for target visibility, tracking accuracy, and trajectory success.

Key Results

Experiments on the DeTrack benchmark show that AaDWorlds improves closed-loop tracking performance across all evaluation metrics.

Resources

Paper