Property Optimized GNN: Improving Data Association Performance Using Cost Function Optimization for Sensor Fusion In High Density Environments
Bibliographic record
Abstract
To reliably track objects in safety critical applications such as autonomous vehicles, the chosen data association algorithm must be capable of handling scenarios where objects are in close proximity to each other and/or frequently cross paths. This paper introduces Property Optimized GNN (POG), a novel generalized property-based data association strategy to increase accuracy of track-level sensor fusion in high density scenarios. POG improves upon the Global Nearest Neighbor (GNN) strategy by accessing track properties from the previous time-step. The current implementation accomplishes this by checking if the ID of the incoming sensor track matches that of the previous confirmed track, rather than associating tracks strictly based on relative distance. To compare performance, the POG algorithm was implemented with both Euclidean and Mahalanobis distance, and compared to a GNN using the same distance equations. These algorithms are evaluated on labelled data collected from a 2023 Cadillac LYRIQ’s stock sensor suite, on a drive cycle containing both city and highway sections. The performance of each method is compared using sGOSPA metrics, accounting for location accuracy, track creation accuracy, and number of ID swaps. The results indicate that POG with Mahalanobis distance outperforms standard Euclidean and Mahalanobis association strategies on nearly all components of sGOSPA.
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.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.005 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 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 itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".