ResearchResearch paperRobotics & Embodied AI1 source · Oct 13, 2026

RingNav: Task-Shaped Circular Memory for Visual Heading

We present RingNav, a $10.6$k--parameter visual heading estimator that constrains its memory to the circular manifold $S^1$.

Tiny robots have limited computation and memory, yet visual navigation systems often maintain more state than a task requires.
We ask whether persistent memory can instead be shaped by the task.
RingNav learns only the turn between consecutive images, while heading is composed directly on the circle.
On the TartanAir Visual SLAM Challenge, RingNav reduces mean absolute turn error by $27%$ and accumulated heading error over $4$ sec by $15%$ relative to a similarly sized learned-memory model.

[1]RingNav: Task-Shaped Circular Memory for Visual Heading
arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 13, 06:33 PM
We present RingNav, a $10.6$k--parameter visual heading estimator that constrains its memory to the circular manifold $S^1$.
Tiny robots have limited computation and memory, yet visual navigation systems often maintain more state than a task requires.

Extractive summary: sentences quoted from the sources.

Before this