VANDERER: Map-Free Exploration using Future-Aware and Visual-Curiosity-Guided Diffusion Policy

arXiv 2026 51.1 method, system

TLDR

VANDERER uses a visual curiosity module with a navigation world model to guide diffusion policies for map-free exploration using monocular images.

Reasoning

The paper presents a novel integration of curiosity-driven exploration with diffusion policies, showing strong simulated results. However, it lacks real-world validation and the world model is limited to navigation prediction, not a full generative or interactive simulator.

Read-first score

Read-first score 51.1, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 21.

Recency 6%

100

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

Citation impact 18%

96.5

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.96478361

Methodology quality 18%

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

Topical relevance 29%

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: 397.

Keyword Scores

world model

world dynamics prediction

model-based reinforcement learning world model

generative world model

interactive world model

world simulator

video world model

Deep Analysis

Innovations

Map-free exploration using only monocular image data, eliminating the need for occupancy maps
Visual Curiosity Module (VCM) that predicts action outcomes via a navigation world model and evaluates them through a curiosity cost
Future-aware and visual-curiosity-guided diffusion policy that uses the curiosity cost to generate exploration-maximizing actions
Demonstration of a direct correlation between visual and geometric curiosity in outdoor environments, enabling efficient exploration with sensor-constrained agents

Methodology

VANDERER employs a Visual Curiosity Module (VCM) that predicts the outcomes of proposed actions using a navigation world model and computes a curiosity cost. This cost guides a pre-trained diffusion policy to generate actions that maximize exploration. The framework is evaluated in diverse simulated environments against established baselines, including NoMaD, using only monocular camera input.

Key Results

VANDERER consistently outperforms baselines, exploring an average of 13.4% more area than NoMaD across diverse simulated environments. The results reveal a direct correlation between visual and geometric curiosity in outdoor environments.

Limitations

Only evaluated in simulated environments; real-world performance is not assessed
Relies solely on monocular camera input, limiting applicability to other sensor modalities
Requires a pre-trained diffusion policy, which may involve significant computational resources for training
The observed correlation between visual and geometric curiosity may be environment-specific and not generalize to indoor or other settings

Resources

Paper