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

arXiv cs.RO / 5/5/2026

📰 NewsIdeas & Deep AnalysisModels & Research

共有:

Key Points

DynoSLAM addresses a core limitation of traditional SLAM by removing the “static environment” assumption, targeting scenes where pedestrians and other moving entities affect localization and mapping.
The method uses a tightly coupled Dynamic GraphSLAM design that integrates socially aware GNNs directly into factor-graph optimization rather than relying on fixed heuristics or single-agent neural priors.
It treats pedestrian motion forecasting as a stochastic world model, using Monte Carlo rollouts from the trained GNN to capture multimodal uncertainty and represent it in the SLAM graph via a dynamic Mahalanobis-distance factor.
The authors report simulated experiments showing strong retrospective tracking accuracy and improved robustness, including avoiding optimization failures associated with deterministic argmax-style approaches.
By producing an empirical mean and covariance for future pedestrian states, DynoSLAM yields a probabilistic “safety envelope” that can improve downstream local planners for anticipatory, collision-free navigation in dense crowds.

Abstract

Traditional Simultaneous Localization and Mapping (SLAM) algorithms rely heavily on the static environment assumption, which severely limits their applicability in real-world spaces populated by moving entities, such as pedestrians. In this work, we propose DynoSLAM, a tightly-coupled Dynamic GraphSLAM architecture that integrates socially-aware Graph Neural Networks (GNNs) directly into the factor graph optimization. Unlike conventional approaches that use rigid constant-velocity heuristics or deterministic single-agent neural priors, our framework formulates pedestrian motion forecasting as a stochastic World Model. By utilizing Monte Carlo rollouts from a trained GNN, we capture the multimodal epistemic uncertainty of human interactions and embed it into the SLAM graph via a dynamic Mahalanobis distance factor. We demonstrate through extensive simulated experiments that this stochastic formulation not only maintains highly accurate retrospective tracking but also prevents the optimization failures caused by the deterministic "argmax problem". Ultimately, extracting the empirical mean and covariance matrices of future pedestrian states provides a mathematically rigorous, probabilistic safety envelope for downstream local planners, enabling anticipatory and collision-free robot navigation in densely crowded environments.