Selective memory: Recalling relevant experience for long‐term visual localization
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Abstract Visual navigation is a key enabling technology for autonomous mobile vehicles. The ability to provide large‐scale, long‐term navigation using low‐cost, low‐power vision sensors is appealing for industrial applications. A crucial requirement for long‐term navigation systems is the ability to localize in environments whose appearance is constantly changing over time—due to lighting, weather, seasons, and physical changes. This paper presents a multiexperience localization (MEL) system that uses a powerful map representation—storing every visual experience in layers—that does not make assumptions about underlying appearance modalities and generators. Our localization system provides real‐time performance by selecting online, a subset of experiences against which to localize. We achieve this task through a novel experience‐triage algorithm based on collaborative filtering, which selects experiences relevant to the live view , outperforming competing techniques. Based on classical memory‐based recommender systems, this technique also enables landmark‐level recommendations, is entirely online, and requires no training data. We demonstrate the capabilities of the MEL system in the context of long‐term autonomous path following in unstructured outdoor environments with a challenging 100‐day field experiment through day, night, snow, spring, and summer. We furthermore provide offline analysis comparing our system to several state‐of‐the‐art alternatives. We show that the combination of the novel methods presented in this paper enable full use of incredibly rich multiexperience maps, opening the door to robust long‐term visual localization.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it