MAD: Mapping-Aware World Models for Agile Quadrotor Flight

arXiv 2026 60.8 method

TLDR

MAD introduces a geometry-aware world model for agile quadrotor flight, learning occupancy and visibility maps for collision avoidance, achieving real-world deployment.

Reasoning

The paper presents a novel mapping-aware world model that explicitly encodes geometry and visibility, leading to strong empirical results in simulation and real-world flight. However, the approach is tailored to depth-based occupancy maps and may not generalize to other sensing modalities or tasks beyond collision avoidance.

Read-first score

Read-first score 60.8, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 45.

Recency 6%

100

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

Citation impact 18%

96.3

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.9627694

Methodology quality 18%

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

Topical relevance 29%

64.3

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 18%

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

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: 459.

Keyword Scores

world model

world dynamics prediction

model-based reinforcement learning world model

interactive world model

generative world model

world simulator

video world model

Deep Analysis

Innovations

Mapping-aware world model (MAD) that reconstructs robocentric occupancy and visibility grid maps instead of raw images
Recurrent latent dynamics encoding local geometry, visibility history, and ego-motion
Training in DiffAero with a GPU-parallel map-construction module for high-throughput supervision
Three policy-learning modes: imagination-based MAD-Dreamer and feature-extractor variants based on PPO and SHAC
Demonstrated cross-task transfer between visual navigation and racing tasks

Methodology

MAD learns recurrent latent dynamics that reconstruct robocentric occupancy and visibility grid maps together with proprioceptive states. It is trained in DiffAero using a GPU-parallel map-construction module that provides supervision for occupancy and visibility. The learned representation is then used in three policy-learning modes: imagination-based MAD-Dreamer and feature-extractor variants based on PPO and SHAC.

Key Results

MAD-based agents achieve higher success rates, faster flight, and better cross-task transfer than corresponding vision-only baselines. The learned policy is deployed on a physical quadrotor, reaching 9.66 m/s in simulation and 5.05 m/s in real-world forest experiments.

Resources

Paper