## Problem 317 — rank:116

**Title:** How animals construct navigation maps

**Added binding assumptions (LaTeX):**

```latex
The displayed Bayesian filtering recursion is an established benchmark under the specified hidden-Markov observation assumptions. Approximating that posterior does not by itself predict the animal's decisions, so add an independent movement/heading loss for the inferred decision law. Define movement states consistently across geographic charts and use the declared geodesic metric. Learning or cue-history dependence requires augmentation of the Markov state rather than future-information leakage.
```

**Direct edits:** 0. **Fresh external checks:** None; supplied sources retained as provenance.