DynoSLAM: Dynamic SLAM with Generative Graph Neural Networks for Real-World Social Navigation

arXiv 2026 49.9 method, system, application

TLDR

DynoSLAM integrates a stochastic GNN-based world model into SLAM for pedestrian motion forecasting, enabling robust navigation in dynamic environments.

Reasoning

The paper presents a novel approach combining graph neural networks with SLAM to handle dynamic pedestrians, addressing the static environment limitation. Its strength lies in the probabilistic formulation using Monte Carlo rollouts to capture uncertainty. However, the lack of real-world experiments limits validation of practical applicability.

Read-first score

Read-first score 49.9, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 36.

Recency 6%

100

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

Citation impact 18%

73.8

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.73795745

Methodology quality 18%

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

Topical relevance 29%

51.4

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

FrontierBridge

Rank sensitivity

Stability: volatile; rank range: 260.

Keyword Scores

world model

world dynamics prediction

interactive world model

generative world model

model-based reinforcement learning world model

world simulator

video world model

Deep Analysis

Innovations

Tightly-coupled Dynamic GraphSLAM architecture integrating socially-aware Graph Neural Networks (GNNs) directly into factor graph optimization
Stochastic World Model formulation for pedestrian motion forecasting using Monte Carlo rollouts from a trained GNN
Dynamic Mahalanobis distance factor that embeds multimodal epistemic uncertainty of human interactions into the SLAM graph
Probabilistic safety envelope derived from empirical mean and covariance of future pedestrian states for downstream local planners

Methodology

DynoSLAM replaces rigid constant-velocity heuristics or deterministic single-agent neural priors with a stochastic World Model. It uses Monte Carlo rollouts from a trained socially-aware GNN to capture multimodal epistemic uncertainty of human interactions, which is then embedded into the SLAM factor graph via a dynamic Mahalanobis distance factor. The framework is evaluated through extensive simulated experiments.

Key Results

The stochastic formulation maintains highly accurate retrospective tracking and prevents optimization failures caused by the deterministic 'argmax problem'. It enables anticipatory and collision-free robot navigation in densely crowded simulated environments.

Limitations

Only validated in simulated environments, not in real-world crowded spaces
Relies on accurate GNN predictions which may not generalize to unseen interaction patterns
Potential computational overhead from Monte Carlo rollouts may limit real-time applicability

Resources

Paper