Pondering the Way: Spatial-perceiving World Action Model for Embodied Navigation
For embodied navigation, SWAM addresses inefficiencies and inconsistencies in existing world model-based planners by enabling goal-consistent trajectory generation without candidate sampling.
SWAM proposes a single-pass framework that jointly generates RGB-D sequences and action trajectories from start and goal images, outperforming two-stage planners in success rate, trajectory accuracy, and inference efficiency with zero-shot generalization.
Existing world model-based planners for visual navigation typically follow a verification-centric paradigm, decoupling goal intent from trajectory synthesis. This approach suffers from candidate dependence, heavy computational overhead, and inconsistencies between sampled actions and predicted visuals. To address these issues, we propose SWAM (Spatial-perceiving World Action Model), a task-centric joint observation-action generation framework. Given start and goal RGB observations, SWAM performs single-pass inference to simultaneously generate intermediate RGB-D sequences and corresponding action trajectories, promoting goal-consistent trajectory generation and improved spatial feasibility. While SWAM leverages depth pseudo-labels during training to internalize spatial priors, it requires only monocular RGB input at inference time. We further introduce a visual-guided action refinement module and a trajectory-scale regularization loss to enforce fine-grained alignment between motion and visual cues while stabilizing predictions across varying distances. Extensive experiments show that SWAM significantly outperforms state-of-the-art two-stage planners in success rate, trajectory accuracy, and inference efficiency, while demonstrating robust zero-shot generalization to unseen environments.