Embodied World Models Emerge from Navigational Task in Open-Ended Environments
This addresses the challenge of developing interpretable and transferable navigation policies for AI agents in partially observable environments, though it is incremental in building on embodied cognition theories.
The study tackled the problem of whether an agent trained with sparse rewards in procedurally generated mazes can internalize spatial concepts like direction and distance, and found that it produced near-optimal paths in unseen mazes, with evidence from dynamical and causal analyses showing emergent embodied world models.
Spatial reasoning in partially observable environments has often been approached through passive predictive models, yet theories of embodied cognition suggest that genuinely useful representations arise only when perception is tightly coupled to action. Here we ask whether a recurrent agent, trained solely by sparse rewards to solve procedurally generated planar mazes, can autonomously internalize metric concepts such as direction, distance and obstacle layout. After training, the agent consistently produces near-optimal paths in unseen mazes, behavior that hints at an underlying spatial model. To probe this possibility, we cast the closed agent-environment loop as a hybrid dynamical system, identify stable limit cycles in its state space, and characterize behavior with a Ridge Representation that embeds whole trajectories into a common metric space. Canonical correlation analysis exposes a robust linear alignment between neural and behavioral manifolds, while targeted perturbations of the most informative neural dimensions sharply degrade navigation performance. Taken together, these dynamical, representational, and causal signatures show that sustained sensorimotor interaction is sufficient for the spontaneous emergence of compact, embodied world models, providing a principled path toward interpretable and transferable navigation policies.