arxiv: 2602.21954 · v2 · submitted 2026-02-25 · ⚛️ physics.soc-ph · cs.RO

Recognition: 1 theorem link

· Lean Theorem

The Swarm Intelligence Freeway-Urban Trajectories (SWIFTraj) Dataset -- Part II: A Graph-Based Approach for Trajectory Connection

Xinkai Ji , Pan Liu , Ying Yang , Yu Han

Authors on Pith no claims yet

Pith reviewed 2026-05-15 19:25 UTC · model grok-4.3

classification ⚛️ physics.soc-ph cs.RO

keywords UAV swarmtrajectory connectiongraph-based approachtime alignmentvehicle matchingHungarian algorithmtraffic trajectories

0 comments

The pith

A graph-based approach connects vehicle trajectories across UAV swarm videos by aligning times and matching vehicles.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces a graph-based method to link vehicle paths from multiple UAV videos into long continuous trajectories covering both freeways and urban roads. By modeling UAV positions as an undirected graph, it automatically determines time offsets between videos through cost minimization on trajectory matches. The Hungarian algorithm then pairs trajectories of the same vehicle observed in different views. Sympathetic readers would care because such continuous paths over several kilometers enable better analysis of traffic patterns across integrated networks, which fragmented short clips cannot provide.

Core claim

By representing the UAV swarm layout as an undirected graph, optimal time offsets are found by minimizing trajectory matching costs across videos, after which the Hungarian algorithm constructs a vehicle matching table to associate trajectories of identical vehicles, yielding alignment errors within three frames and F1-scores near 0.99 on real data.

What carries the argument

Undirected graph of UAV layouts with cost-minimization time alignment and Hungarian algorithm for vehicle matching.

If this is right

Long-distance trajectories exceeding 4.5 km can be generated from swarm UAV footage.
Integrated freeway and urban road networks can be monitored with continuous vehicle paths.
Time synchronization between videos achieves accuracy within 0.1 seconds.
Vehicle re-identification across views reaches F1-score of 0.99.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

This technique could apply to fusing data from other distributed camera systems in traffic or surveillance.
Scaling the swarm size might allow even longer trajectories or coverage of larger areas without additional infrastructure.
The cost-based alignment may offer robustness in scenarios with partial overlaps or varying UAV densities.

Load-bearing premise

The approach relies on accurate vehicle detection and tracking being already solved within each separate UAV video, plus enough overlap between views for reliable cost matching.

What would settle it

Collecting ground-truth time offsets and vehicle identities from real UAV swarm footage and finding that the computed alignments deviate by more than three frames or that matching F1-scores fall below 0.95 would disprove the reported performance.

Figures

Figures reproduced from arXiv: 2602.21954 by Pan Liu, Xinkai Ji, Ying Yang, Yu Han.

**Figure 1.** Figure 1: Example of video graph. (a) (Left) The flight plan of pNEUMA dataset (the image is from Barmpounakis and Geroliminis (2020)). [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗

**Figure 2.** Figure 2: The GCVT framework. The subsets of trajectory data from video [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

**Figure 3.** Figure 3: The spatial transformation of images. The four orange points are the selected feature points. [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

**Figure 4.** Figure 4: Study area and the video graph. (Up) Satellite map of the study area and the recording area of each UAV. (Down) Video graph for the [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗

**Figure 5.** Figure 5: Variation of the mean trajectory matching cost defined in Eq. (1) for di [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗

**Figure 6.** Figure 6: The video stitching after time alignment. The yellow dashed lines represent the seams where two videos are stitched together [PITH_FULL_IMAGE:figures/full_fig_p013_6.png] view at source ↗

**Figure 7.** Figure 7: Time alignment results in Exp2. 5.3. Evaluation of vehicle matching Vehicle matching is a critical component of the trajectory connection framework. Its accuracy depends not only on the matching algorithm itself, but also on the preceding spatial transformation and time alignment stages, which ensure spatial and temporal consistency across videos. Therefore, the evaluation results reflect the overall effec… view at source ↗

**Figure 8.** Figure 8: The original trajectory and connected trajectory. [PITH_FULL_IMAGE:figures/full_fig_p016_8.png] view at source ↗

**Figure 9.** Figure 9: Space–time diagram of vehicle trajectories after connection, covering videos [PITH_FULL_IMAGE:figures/full_fig_p016_9.png] view at source ↗

**Figure 10.** Figure 10: Spatial trajectories illustrating connections across videos [PITH_FULL_IMAGE:figures/full_fig_p017_10.png] view at source ↗

**Figure 11.** Figure 11: Vehicle matching results in Exp2: (a) F1-score under di [PITH_FULL_IMAGE:figures/full_fig_p017_11.png] view at source ↗

read the original abstract

In Part I of this companion paper series, we introduced SWIFTraj, a new open-source vehicle trajectory dataset collected using a unmanned aerial vehicle (UAV) swarm. The dataset has two distinctive features. First, by connecting trajectories across consecutive UAV videos, it provides long-distance continuous trajectories, with the longest exceeding 4.5 km. Second, it covers an integrated traffic network consisting of both freeways and their connected urban roads. Obtaining such long-distance continuous trajectories from a UAV swarm is challenging, due to the need for accurate time alignment across multiple videos and the irregular spatial distribution of UAVs. To address these challenges, this paper proposes a novel graph-based approach for connecting vehicle trajectories captured by a UAV swarm. An undirected graph is constructed to represent flexible UAV layouts, and an automatic time alignment method based on trajectory matching cost minimization is developed to estimate optimal time offsets across videos. To associate trajectories of the same vehicle observed in different videos, a vehicle matching table is established using the Hungarian algorithm. The proposed approach is evaluated using both simulated and real-world data. Results from real-world experiments show that the time alignment error is within three video frames, corresponding to approximately 0.1 s, and that the vehicle matching achieves an F1-score of about 0.99. These results demonstrate the effectiveness of the proposed method in addressing key challenges in UAV-based trajectory connection and highlight its potential for large-scale vehicle trajectory collection.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper gives a workable graph-based method to stitch UAV-swarm trajectories across freeway-urban roads, with reported real-world alignment under 0.1 s and 0.99 F1 matching.

read the letter

This paper describes a graph-based pipeline to stitch together vehicle trajectories from a swarm of UAVs covering both freeways and urban roads. The core idea is to model the UAV layout as an undirected graph, align the videos by minimizing a trajectory-matching cost, and then match vehicles across views with the Hungarian algorithm. What it does well is deliver concrete numbers from real-world experiments: time alignment error within three frames (about 0.1 seconds) and vehicle matching F1-score near 0.99. Those results come from actual data rather than just simulations, which is a step up for this kind of work. The approach handles irregular UAV positions better than fixed setups, and the companion dataset aims for long continuous tracks over 4.5 km. The main gaps are the lack of detail on per-video detection and tracking accuracy, plus the exact form of the cost function used for alignment. The paper assumes reliable single-video tracking and enough overlap between views, which are reasonable but not tested in edge cases here. If those hold, the method looks effective; otherwise the numbers might not generalize. This is aimed at researchers building large-scale trajectory datasets from aerial video or working on multi-drone surveillance. A reader interested in practical multi-view fusion for traffic data would get value from the reported performance. I would recommend sending it for peer review. The empirical claims are specific enough to warrant a closer look, even if the write-up needs more on the implementation steps.

Referee Report

0 major / 2 minor

Summary. The paper proposes a graph-based approach for connecting vehicle trajectories across UAV swarm videos in the SWIFTraj dataset. It constructs an undirected graph to represent flexible UAV layouts, develops an automatic time alignment method based on trajectory matching cost minimization to estimate optimal time offsets, and applies the Hungarian algorithm to build a vehicle matching table for associating trajectories of the same vehicle across videos. Evaluation on simulated and real-world data reports time alignment errors within three video frames (~0.1 s) and vehicle matching F1-scores of ~0.99.

Significance. If the empirical results hold, the work is significant for enabling long continuous trajectories (exceeding 4.5 km) across integrated freeway-urban networks using UAV swarms, a challenging task due to irregular UAV distributions and multi-video alignment needs. The approach leverages standard algorithms with direct experimental validation on real data, supporting its utility for large-scale traffic trajectory collection and analysis.

minor comments (2)

[Method description] The exact formulation of the trajectory matching cost function (used for time alignment) is referenced but not fully specified with equations or components, which would aid reproducibility even if the overall performance metrics are reported.
[Evaluation] The paper could add a brief discussion of detection/tracking accuracy within individual videos and any observed failure modes, as these are prerequisites for the connection pipeline.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive review and the recommendation to accept the manuscript. We appreciate the recognition of the significance of the graph-based trajectory connection method for enabling long continuous trajectories across integrated freeway-urban networks using UAV swarms.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper proposes a graph-based trajectory connection method using an undirected graph for UAV layouts, cost-minimization for time alignment, and the Hungarian algorithm for vehicle matching. These steps apply standard algorithms to the dataset from the companion Part I paper. Central performance claims (time alignment error within three frames and F1-score ~0.99) are obtained via direct empirical measurement on simulated and real-world data rather than any fitted parameter being reused as a prediction. No equations reduce to self-definition, no uniqueness theorems are imported from the authors' prior work, and the single reference to Part I is not load-bearing for the method's validity. The derivation chain is self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The method rests on standard graph theory and the Hungarian algorithm; no new free parameters, invented physical entities, or ad-hoc axioms are introduced in the abstract description.

axioms (2)

domain assumption Undirected graph representation suffices to capture flexible UAV spatial layouts for trajectory matching
Invoked when constructing the graph to model possible video overlaps.
standard math Hungarian algorithm yields the optimal assignment for vehicle trajectory matching
Standard use of the bipartite matching algorithm.

pith-pipeline@v0.9.0 · 5576 in / 1307 out tokens · 42951 ms · 2026-05-15T19:25:34.042212+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

An undirected graph is constructed to represent flexible UAV layouts, and an automatic time alignment method based on trajectory matching cost minimization is developed... Hungarian algorithm

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

16 extracted references · 16 canonical work pages

[1]

Communica- tions in Transportation Research 4, 100133

Vehicle trajectory dataset from drone videos including off-ramp and congested traffic – analysis of data quality, traffic flow, and accident risk. Communica- tions in Transportation Research 4, 100133. URL:https://www.sciencedirect.com/science/article/ pii/S2772424724000167, doi:https://doi.org/10.1016/j.commtr.2024.100133. Bock, J., Krajewski, R., Moers,...

work page doi:10.1016/j.commtr.2024.100133 2024
[2]

The ind dataset: A drone dataset of naturalistic road user trajectories at german intersections, in: 2020 IEEE Intelligent Vehicles Symposium (IV), IEEE. pp. 1929–1934. Chaudhari, A.A., Treiber, M., Okhrin, O.,

work page 2020
[3]

Chen, D., Ahn, S., Laval, J., Zheng, Z.,

doi:10.1038/s41597-025-05472-0. Chen, D., Ahn, S., Laval, J., Zheng, Z.,

work page doi:10.1038/s41597-025-05472-0
[4]

arXiv preprint arXiv:2503.12562

History-aware transformation of reid features for multiple object tracking. arXiv preprint arXiv:2503.12562 . Gloudemans, D., Wang, Y ., Ji, J., Zachar, G., Barbour, W., Hall, E., Cebelak, M., Smith, L., Work, D.B.,

work page arXiv
[5]

IEEE Transactions on Intelligent Transportation Systems 26, 59–76

Openvter: An open vehicle trajectory extraction framework based on rotated bounding boxes. IEEE Transactions on Intelligent Transportation Systems 26, 59–76. doi:10. 1109/TITS.2024.3481256. Kim, Z., Cao, M.,

work page arXiv 2024
[6]

The highd dataset: A drone dataset of naturalistic vehicle trajectories on german highways for validation of highly automated driving systems, in: 2018 21st International Conference on Intelligent Transportation Systems (ITSC), IEEE. pp. 2118–2125. Krajewski, R., Moers, T., Bock, J., Vater, L., Eckstein, L.,

work page 2018
[7]

The round dataset: A drone dataset of road user trajectories at roundabouts in germany, in: 2020 IEEE 23rd International Conference on Intelligent Transportation Systems (ITSC), IEEE. pp. 1–6. Kuhn, H.W.,

work page 2020
[8]

Communications in Transportation Research 5, 100200

Enhanced trajectory reconstruction from sparse and noisy gps data: A progres- sive chunked transformer approach. Communications in Transportation Research 5, 100200. URL:https:// www.sciencedirect.com/science/article/pii/S277242472500040X, doi:https://doi.org/10.1016/ j.commtr.2025.100200. Liu, Z., He, J., Zhang, C., Yan, X., Wang, C., Qiao, B.,

work page arXiv 2025
[9]

Transportation Research Part E: Logistics and Transportation Review 192, 103799

Physics-informed neural network for cross-dynamics vehicle trajectory stitching. Transportation Research Part E: Logistics and Transportation Review 192, 103799. doi:https://doi.org/10. 1016/j.tre.2024.103799. Ma, W., Zhong, H., Wang, L., Jiang, L., Abdel-Aty, M.,

work page arXiv 2024
[10]

The exid dataset: A real-world trajectory dataset of highly interactive highway scenarios in germany, in: 2022 IEEE Intelligent Vehicles Symposium (IV), IEEE. pp. 958–964. Rajput, S., Venkateshappa, S., Kanagaraj, V ., Asaithambi, G., Treiber, M.,

work page 2022
[11]

Transportation Research Part C: Emerging Technologies 182, 105431

Spt: Obtaining long trajectory data of disordered traffic using a swarm of unmanned aerial vehicles. Transportation Research Part C: Emerging Technologies 182, 105431. doi:10.1016/j.trc.2025.105431. Raju, N., Arkatkar, S., Easa, S., Joshi, G.,

work page doi:10.1016/j.trc.2025.105431 2025
[12]

Transportation Research Part C: Emerging Technologies 160, 104520

Automatic vehi- cle trajectory data reconstruction at scale. Transportation Research Part C: Emerging Technologies 160, 104520. doi:https://doi.org/10.1016/j.trc.2024.104520. Wojke, N., Bewley, A., Paulus, D.,

work page doi:10.1016/j.trc.2024.104520 2024
[13]

3645–3649

Simple online and realtime tracking with a deep association metric, in: 2017 IEEE International Conference on Image Processing (ICIP), pp. 3645–3649. doi:10.1109/ICIP.2017.8296962. Xu, Y ., Shao, W., Li, J., Yang, K., Wang, W., Huang, H., Lv, C., Wang, H.,

work page doi:10.1109/icip.2017.8296962 2017
[14]

Sind: A drone dataset at signalized intersection in china, in: 2022 IEEE 25th International Conference on Intelligent Transportation Systems (ITSC), IEEE. pp. 2471–2478. Xu, Y ., Yu, G., Wu, X., Wang, Y ., Ma, Y .,

work page 2022
[15]

arXiv preprint arXiv:1910.03088

Interaction dataset: An international, adversarial and cooperative motion dataset in interactive driving scenarios with semantic maps. arXiv preprint arXiv:1910.03088 . Zhang, Y ., Wang, C., Yu, R., Wang, L., Quan, W., Gao, Y ., Li, P.,

work page arXiv 1910
[16]

Transportation Research Record 2678, 606–621

Citysim: A drone-based vehicle trajectory dataset for safety-oriented research and digital twins. Transportation Research Record 2678, 606–621. doi:10.1177/03611981231185768. 21

work page doi:10.1177/03611981231185768