arxiv: 2604.22973 · v1 · submitted 2026-04-24 · 💻 cs.RO

Collaborative Trajectory Prediction via Late Fusion

Nadya Abdel Madjid , Murad Mebrahtu , Zakhar Yagudin , Bilal Hassan , Naoufel Werghi , Jorge Dias , Dzmitry Tsetserukou , Majid Khonji This is my paper

Pith reviewed 2026-05-08 11:13 UTC · model grok-4.3

classification 💻 cs.RO

keywords collaborative trajectory predictionlate fusionvehicle-to-vehicletrajectory forecastingautonomous drivingmiss ratetrajectory success rate

0 comments p. Extension

The pith

Late fusion of independent vehicle forecasts reduces prediction errors with lower communication costs.

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

The paper introduces a late-fusion framework that lets each vehicle run its own trajectory predictor and then combine the final forecasts rather than merging raw sensor data or feature maps early on. This design treats collaborating vehicles as independent and asynchronous agents, so only compact prediction outputs need to be shared. A sympathetic reader would care because the approach sidesteps the high bandwidth and synchronization demands that have limited earlier collaborative methods. The authors show that the resulting forecasts achieve lower miss rates and higher trajectory success rates (TSR_0.5) on both simulated and real-world datasets.

Core claim

By fusing trajectory predictions after each vehicle has independently forecasted, the method compensates for occlusions and perception errors in a model-agnostic way, leading to lower miss rates and higher TSR_0.5 on multiple datasets, including gains of 1.69% and 1.22% on the real-world V2V4Real dataset for the two vehicles.

What carries the argument

Late fusion framework that combines shared final trajectory forecasts from independent asynchronous agents.

If this is right

Communication overhead drops because only compact final predictions are exchanged instead of high-dimensional feature maps.
Any existing single-vehicle predictor can be used without retraining or architectural changes.
Performance gains appear consistently across OPV2V, V2V4Real, and DeepAccident datasets.
The method tolerates asynchronous operation and does not require perfect synchronization between vehicles.

Where Pith is reading between the lines

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

Output-level fusion can compensate for individual occlusions even when one agent has a severely limited view.
The same late-fusion idea could be tested in multi-agent settings beyond vehicles, such as robot teams with partial observability.
Hybrid systems that combine selective early fusion with this late fusion might achieve still lower error at modest extra bandwidth cost.

Load-bearing premise

That the independent predictions from each vehicle are accurate enough for their fusion to meaningfully improve results without adding errors from bad forecasts.

What would settle it

A controlled test where one vehicle is given deliberately degraded input so its individual prediction is poor, then checking whether the fused output still beats the better of the two single-vehicle predictions.

Figures

Figures reproduced from arXiv: 2604.22973 by Bilal Hassan, Dzmitry Tsetserukou, Jorge Dias, Majid Khonji, Murad Mebrahtu, Nadya Abdel Madjid, Naoufel Werghi, Zakhar Yagudin.

**Figure 1.** Figure 1: Occlusion scenarios, where collaboration between vehicles is useful. view at source ↗

**Figure 2.** Figure 2: System architecture of the proposed collaborative trajectory prediction framework. (a) Broadcast–listen communication setup and packet structure used view at source ↗

**Figure 4.** Figure 4: Per-broadcast message sizes for Vehicles 1 and 2 on V2V4Real (val view at source ↗

**Figure 5.** Figure 5: Per-broadcast message sizes for all seven vehicles on OPV2V (vali view at source ↗

**Figure 6.** Figure 6: Per-broadcast message sizes for all seven vehicles on OPV2V (test view at source ↗

**Figure 7.** Figure 7: Per-broadcast message sizes for four vehicles on DeepAccident (valid view at source ↗

**Figure 8.** Figure 8: Per-broadcast communication delay as a function of packet size. view at source ↗

**Figure 9.** Figure 9: Distribution of communication delay across three broadcast size cat view at source ↗

read the original abstract

Predicting future trajectories of surrounding traffic agents is critical for safe autonomous navigation and collision avoidance. Despite all advances in the trajectory forecasting realm, the prediction models remains vulnerable to uncertainty caused by occlusions, limited sensing range, and perception errors. Collaborative vehicle-to-vehicle (V2V) approaches help reduce this uncertainty by sharing complementary information. Existing collaborative trajectory prediction methods typically fuse feature maps at the perception stage to construct a holistic scene view. Further this holistic representation is decoded into the future trajectories. Such design incurs substantial communication overhead due to the exchange of high-dimensional feature representations and often assumes idealized bandwidth and synchronization, limiting practical deployment. We address these limitations by shifting collaboration from perception to the prediction module and introducing a late-fusion framework for shared forecasts. The framework is model-agnostic and treats collaborating vehicles as independent asynchronous agents. We evaluate the approach on the OPV2V, V2V4Real, and DeepAccident datasets, comparing individual and collaborative forecasting. Across all datasets, late fusion consistently reduces miss rate and improves trajectory success rate ($\mathrm{TSR}_{0.5}$), defined as the fraction of ground-truth agents with final displacement error below 0.5 m. On the real-world V2V4Real dataset, collaborative prediction improves the success rate by $1.69\%$ and $1.22\%$ for both intelligent vehicles, respectively, compared with individual forecasting.

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

Late fusion shifts collaboration to the prediction stage to cut comms costs, but the reported gains stay small and the handling of async agent matching stays underspecified.

read the letter

The paper moves collaborative trajectory prediction from early feature fusion at perception to late fusion of independent forecasts. This is the main practical move: vehicles share only their own predicted trajectories instead of high-dimensional maps, which should lower bandwidth needs and relax synchronization assumptions. They frame the setup as model-agnostic and explicitly asynchronous, then test on OPV2V, V2V4Real, and DeepAccident. On the real-world V2V4Real set the success rate rises by roughly 1.7 % and 1.2 % for the two vehicles relative to solo forecasting, with similar patterns on the other sets for miss rate and TSR_0.5. That is the concrete evidence they provide. The shift itself is a clean architectural choice that prior V2V work has not emphasized in the same way. The multi-dataset evaluation, including real data, gives the claim some grounding beyond simulation. The modest size of the lift is reported plainly rather than overstated. The soft spots sit in the details that are not visible from the abstract. Late fusion still requires some form of cross-vehicle association and time alignment for the predictions to be combined usefully; if that step leans on shared perception outputs or position heuristics, the claimed overhead reduction shrinks. The abstract gives no error bars, no ablation on the fusion rule itself, and no breakdown of how asynchrony is resolved in practice. Those gaps make it hard to judge whether the 1-2 % gains are robust or mainly an artifact of favorable matching. This is the kind of paper that matters to groups building deployable V2V prediction stacks who already have individual forecasters and want to add collaboration without blowing up the link budget. A reader who cares about communication-aware autonomy will find the framing useful even if the numbers stay incremental. It is coherent on its own terms and engages the right prior literature on perception-stage fusion, so it clears the bar for serious refereeing. I would send it out, but with the expectation that reviewers will press on the association mechanism and whether the effect sizes justify the added coordination layer.

Referee Report

3 major / 2 minor

Summary. The paper introduces a late-fusion framework for collaborative trajectory prediction that fuses independent forecasts from asynchronous V2V agents at the prediction stage rather than fusing perception features. It is presented as model-agnostic and is evaluated on the OPV2V, V2V4Real, and DeepAccident datasets, where it reports consistent reductions in miss rate and gains in TSR_0.5 (fraction of agents with final displacement error below 0.5 m). On the real-world V2V4Real dataset the method improves success rate by 1.69% and 1.22% for the two intelligent vehicles relative to individual forecasting.

Significance. If the gains are reproducible, the late-fusion design could reduce communication bandwidth relative to early-fusion baselines while remaining compatible with existing single-vehicle predictors. The use of public datasets and direct comparison to individual forecasting are positive aspects. However, the modest size of the reported improvements and the absence of statistical validation limit the immediate practical significance.

major comments (3)

[Abstract] Abstract: the central claim of consistent improvement is supported only by point estimates (e.g., +1.69% and +1.22% TSR_0.5 on V2V4Real) with no error bars, ablation studies, or statistical significance tests. Without these, it is impossible to determine whether the observed gains exceed run-to-run variability or dataset-specific noise.
[Late-fusion framework] Late-fusion framework description: the manuscript states that vehicles are treated as 'independent asynchronous agents' yet provides no explicit procedure for solving the agent-association and temporal-alignment problem before fusion. Because the claimed benefit is compensation for occlusions and perception errors via fusion of independent predictions, the lack of detail on correspondence is load-bearing for the central claim.
[Experiments] Evaluation on V2V4Real: the reported success-rate gains are presented without any analysis of how association errors or communication latency in the real-world asynchronous setting affect the fusion step. This directly bears on the weakest assumption that independent predictions are sufficiently accurate for late fusion to compensate without additional mechanisms.

minor comments (2)

[Abstract] The threshold of 0.5 m used to define TSR_0.5 should be justified by reference to prior trajectory-prediction literature or safety requirements.
[Experiments] Implementation details (network architectures, training hyperparameters, exact fusion operator) are not provided, hindering reproducibility.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments and the recommendation for major revision. We address each major comment below, indicating the revisions we will incorporate to strengthen the manuscript.

read point-by-point responses

Referee: [Abstract] Abstract: the central claim of consistent improvement is supported only by point estimates (e.g., +1.69% and +1.22% TSR_0.5 on V2V4Real) with no error bars, ablation studies, or statistical significance tests. Without these, it is impossible to determine whether the observed gains exceed run-to-run variability or dataset-specific noise.

Authors: We agree that the abstract presents point estimates without error bars or statistical tests. The full manuscript reports consistent gains across OPV2V, V2V4Real, and DeepAccident with direct comparisons to individual forecasting and includes ablations on fusion components in the experiments section. To address the concern, we will revise the abstract and results to include error bars from multiple runs, ablation expansions, and statistical significance tests such as paired t-tests on the TSR_0.5 improvements. revision: yes
Referee: [Late-fusion framework] Late-fusion framework description: the manuscript states that vehicles are treated as 'independent asynchronous agents' yet provides no explicit procedure for solving the agent-association and temporal-alignment problem before fusion. Because the claimed benefit is compensation for occlusions and perception errors via fusion of independent predictions, the lack of detail on correspondence is load-bearing for the central claim.

Authors: The late-fusion approach assumes standard V2V communication provides agent IDs and timestamps, enabling association via shared identifiers. Temporal alignment uses linear interpolation of predictions to a common horizon, consistent with the asynchronous setting in the evaluated datasets. We will add an explicit subsection in the revised Section 3 detailing this association and alignment procedure, including pseudocode, to make the implementation clear. revision: yes
Referee: [Experiments] Evaluation on V2V4Real: the reported success-rate gains are presented without any analysis of how association errors or communication latency in the real-world asynchronous setting affect the fusion step. This directly bears on the weakest assumption that independent predictions are sufficiently accurate for late fusion to compensate without additional mechanisms.

Authors: Our V2V4Real experiments used the dataset's ground-truth associations to isolate the fusion benefit from individual predictions. We acknowledge the value of robustness analysis for real-world deployment. In the revision, we will add a new experiment subsection simulating association errors (e.g., random mismatches at 5-15% rates) and communication latency (0-500ms) on the V2V4Real data to quantify their impact on TSR_0.5 gains. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical gains measured directly against baselines on public datasets

full rationale

The paper introduces a late-fusion framework for collaborative trajectory prediction, describes it as model-agnostic and treating vehicles as independent asynchronous agents, then reports measured improvements in miss rate and TSR_0.5 on OPV2V, V2V4Real, and DeepAccident by direct comparison to individual forecasting. No equations, derivations, or first-principles claims appear that reduce the reported success-rate gains to fitted parameters, self-definitions, or self-citation chains. The evaluation uses public datasets and standard metrics without renaming known results or smuggling ansatzes via prior self-work. The modest real-world deltas are presented as observed outcomes, not forced by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

No free parameters, axioms, or invented entities are described in the abstract; the contribution is an empirical framework relying on standard trajectory prediction models and public datasets.

pith-pipeline@v0.9.0 · 5582 in / 1087 out tokens · 22994 ms · 2026-05-08T11:13:46.660902+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

43 extracted references · 19 canonical work pages

[1]

H. Zhao, J. Gao, T. Lan, C. Sun, B. Sapp, B. Varadarajan, Y . Shen, Y . Chai, C. Schmid, Tnt: Target-driven trajectory prediction, in: Proceedings of the Conference on Robot Learning, CoRL, 2020, pp. 1–12

2020
[2]

D. Li, Q. Zhang, Z. Xia, Y . Zheng, K. Zhang, M. Yi, W. Jin, D. Zhao, Planning-inspired hierarchical trajec- tory prediction via lateral-longitudinal decomposition for autonomous driving, IEEE Transactions on Intelligent Vehicles 9 (2024) 692–702. doi:10.1109/TIV.2023. 3307116

work page doi:10.1109/tiv.2023 2024
[3]

S. Moon, H. Woo, H. Park, H. Jung, R. Mahjourian, H.-g. Chi, H. Lim, S. Kim, J. Kim, Visiontrap: Vision- augmented trajectory prediction guided by textual descriptions, in: Computer Vision – ECCV 2024: 18th European Conference, Milan, Italy, September 29–Oc- tober 4, 2024, Proceedings, Part VI, Springer-Verlag, Berlin, Heidelberg, 2024, p. 361–379. URL:ht...

work page doi:10.1007/978-3-031-72658-3_21 2024
[4]

T.-H. Wang, S. Manivasagam, M. Liang, B. Yang, W. Zeng, R. Urtasun, V2vnet: Vehicle-to-vehicle com- munication for joint perception and prediction, in: A. Vedaldi, H. Bischof, T. Brox, J.-M. Frahm (Eds.), Computer Vision – ECCV 2020, Springer International Publishing, Cham, 2020, pp. 605–621

2020
[5]

T. Wang, S. Kim, J. Wenxuan, E. Xie, C. Ge, J. Chen, Z. Li, P. Luo, Deepaccident: a motion and accident pre- diction benchmark for v2x autonomous driving, in: Pro- ceedings of the Thirty-Eighth AAAI Conference on Ar- tificial Intelligence and Thirty-Sixth Conference on In- novative Applications of Artificial Intelligence and Four- teenth Symposium on Educ...

work page doi:10.1609/aaai.v38i6 2025
[6]

R. Xu, Z. Tu, H. Xiang, W. Shao, B. Zhou, J. Ma, Cobevt: Cooperative bird’s eye view semantic segmentation with sparse transformers, in: Conference on Robot Learning (CoRL), 2022

2022
[7]

H. Ruan, H. Yu, W. Yang, S. Fan, Y . Tang, Z. Nie, Learn- ing cooperative trajectory representations for motion fore- casting, 2023.arXiv:2311.00371

work page arXiv 2023
[8]

Z. Wu, Y . Wang, H. Ma, Z. Li, H. Qiu, J. Li, Cmp: Co- operative motion prediction with multi-agent communica- tion, 2024.arXiv:2403.17916

work page arXiv 2024
[9]

Z. Lei, Z. Ni, R. Han, S. Tang, D. Wang, C. Feng, S. Chen, Y . Wang, Robust collaborative perception without external localization and clock devices, 2024. URL:https://arxiv.org/abs/2405.02965. arXiv:2405.02965

work page arXiv 2024
[10]

Y . Li, S. Ren, P. Wu, S. Chen, C. Feng, W. Zhang, Learn- ing distilled collaboration graph for multi-agent percep- tion, in: Proceedings of the 35th International Confer- ence on Neural Information Processing Systems, NIPS ’21, Curran Associates Inc., Red Hook, NY , USA, 2024

2024
[11]

Y . Xia, Q. Yuan, G. Luo, X. Fu, Y . Li, X. Zhu, T. Luo, S. Chen, J. Li, One is plenty: A polymorphic fea- ture interpreter for immutable heterogeneous collabora- tive perception, 2025. URL:https://arxiv.org/ abs/2411.16799.arXiv:2411.16799. 16

work page arXiv 2025
[12]

Y . Xu, L. Li, J. Wang, B. Yang, Z. Wu, X. Chen, J. Wang, Codyntrust: Robust asynchronous collab- orative perception via dynamic feature trust modu- lus, 2025. URL:https://arxiv.org/abs/2502. 08169.arXiv:2502.08169

work page arXiv 2025
[13]

Z. Lei, S. Ren, Y . Hu, W. Zhang, S. Chen, Latency- aware collaborative perception, in: S. Avidan, G. Bros- tow, M. Cissé, G. M. Farinella, T. Hassner (Eds.), Com- puter Vision – ECCV 2022, Springer Nature Switzerland, Cham, 2022, pp. 316–332

2022
[14]

H. Yu, Y . Tang, E. Xie, J. Mao, P. Luo, Z. Nie, Flow-based feature fusion for vehicle-infrastructure cooperative 3d object detection, in: Thirty-seventh Conference on Neu- ral Information Processing Systems, 2023. URL:https: //openreview.net/forum?id=gsglrhvQxX

2023
[15]

Y .-C. Liu, J. Tian, C.-Y . Ma, N. Glaser, C.-W. Kuo, Z. Kira, Who2com: Collaborative perception via learn- able handshake communication, in: 2020 IEEE Interna- tional Conference on Robotics and Automation (ICRA), 2020, pp. 6876–6883. doi:10.1109/ICRA40945. 2020.9197364

work page doi:10.1109/icra40945 2020
[16]

T. Wang, G. Chen, K. Chen, Z. Liu, B. Zhang, A. Knoll, C. Jiang, Umc: A unified bandwidth-efficient and multi-resolution based collaborative perception frame- work, in: 2023 IEEE/CVF International Conference on Computer Vision (ICCV), IEEE Computer Soci- ety, Los Alamitos, CA, USA, 2023, pp. 8153–8162. URL:https://doi.ieeecomputersociety. org/10.1109/ICCV...

work page doi:10.1109/iccv51070.2023.00752 2023
[17]

K. Yang, D. Yang, J. Zhang, H. Wang, P. Sun, L. Song, What2comm: Towards communication-efficient collabo- rative perception via feature decoupling, in: Proceedings of the 31st ACM International Conference on Multimedia, Association for Computing Machinery, New York, NY , USA, 2023, p. 7686–7695

2023
[18]

D. Yang, K. Yang, Y . Wang, J. Liu, Z. Xu, R. Yin, P. Zhai, L. Zhang, How2comm: Communication- efficient and collaboration-pragmatic multi-agent percep- tion, in: Thirty-seventh Conference on Neural In- formation Processing Systems, 2023. URL:https:// openreview.net/forum?id=Dbaxm9ujq6

2023
[19]

Y . Hu, S. Fang, Z. Lei, Y . Zhong, S. Chen, Where2comm: communication-efficient collaborative perception via spa- tial confidence maps, in: Proceedings of the 36th Interna- tional Conference on Neural Information Processing Sys- tems, NIPS ’22, Curran Associates Inc., Red Hook, NY , USA, 2024

2024
[20]

Z. Chen, Y . Shi, J. Jia, Transiff: An instance-level fea- ture fusion framework for vehicle-infrastructure coopera- tive 3d detection with transformers, in: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2023, pp. 18205–18214

2023
[21]

R. Xu, H. Xiang, Z. Tu, X. Xia, M.-H. Yang, J. Ma, V2x-vit: Vehicle-to-everything cooperative perception withÂ vision transformer, in: S. Avidan, G. Bros- tow, M. Cissé, G. M. Farinella, T. Hassner (Eds.), Com- puter Vision – ECCV 2022, Springer Nature Switzerland, Cham, 2022, pp. 107–124

2022
[22]

J. Xu, Y . Zhang, Z. Cai, D. Huang, Cosdh: Communication-efficient collaborative perception via supply-demand awareness and intermediate-late hybridization, 2025. URL:https://arxiv.org/ abs/2503.03430.arXiv:2503.03430

work page arXiv 2025
[23]

X. Gao, R. Xu, J. Li, Z. Wang, Z. Fan, Z. Tu, STAMP: Scalable task- and model-agnostic collaborative percep- tion, in: The Thirteenth International Conference on Learning Representations, 2025. URL:https:// openreview.net/forum?id=8NdNniulYE

2025
[24]

Glaser, Y .-C

N. Glaser, Y .-C. Liu, J. Tian, Z. Kira, Overcoming ob- structions via bandwidth-limited multi-agent spatial hand- shaking, in: 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2021, pp. 2406–

2021
[25]

doi:10.1109/IROS51168.2021.9636761

work page doi:10.1109/iros51168.2021.9636761 2021
[26]

S. Hong, Y . Liu, Z. Li, S. Li, Y . He, Multi-agent collabo- rative perception via motion-aware robust communication network, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024, pp. 15301–15310

2024
[27]

Y . Tao, S. Hu, Z. Fang, Y . Fang, Direct-cp: Di- rected collaborative perception for connected and autonomous vehicles via proactive attention, 2025. URL:https://arxiv.org/abs/2409.08840. arXiv:2409.08840

work page arXiv 2025
[28]

R. Chen, Y . Mu, R. Xu, W. Shao, C. Jiang, H. Xu, Y . Qiao, Z. Li, P. Luo, CO3: Cooperative unsupervised 3d rep- resentation learning for autonomous driving, in: The Eleventh International Conference on Learning Represen- tations, 2023. URL:https://openreview.net/ forum?id=QUaDoIdgo0

2023
[29]

S. Wei, Y . Wei, Y . Hu, Y . Lu, Y . Zhong, S. Chen, Y . Zhang, Asynchrony-robust collaborative perception via bird’s eye view flow, in: Thirty-seventh Conference on Neural Information Processing Systems, 2023. URL:https: //openreview.net/forum?id=UHIDdtxmVS

2023
[30]

D. Qiao, F. Zulkernine, Adaptive feature fusion for co- operative perception using lidar point clouds, in: 2023 IEEE/CVF Winter Conference on Applications of Com- puter Vision (W ACV), 2023, pp. 1186–1195. doi:10. 1109/WACV56688.2023.00124. 17

work page arXiv 2023
[31]

Y . Zhou, J. Xiao, Y . Zhou, G. Loianno, Multi-robot col- laborative perception with graph neural networks, IEEE Robotics and Automation Letters 7 (2022) 2289–2296. doi:10.1109/LRA.2022.3141661

work page doi:10.1109/lra.2022.3141661 2022
[32]

K. Yang, D. Yang, J. Zhang, M. Li, Y . Liu, J. Liu, H. Wang, P. Sun, L. Song, Spatio-Temporal Domain Awareness for Multi-Agent Collaborative Perception , in: 2023 IEEE/CVF International Conference on Computer Vision (ICCV), IEEE Computer Society, Los Alamitos, CA, USA, 2023, pp. 23326–23335. URL:https://doi.ieeecomputersociety. org/10.1109/ICCV51070.2023....

work page doi:10.1109/iccv51070.2023.02137 2023
[33]

J. Cui, H. Qiu, D. Chen, P. Stone, Y . Zhu, Cooper- naut: End-to-end driving with cooperative perception for networked vehicles, 2022 IEEE/CVF Confer- ence on Computer Vision and Pattern Recogni- tion (CVPR) (2022) 17231–17241. URL:https: //api.semanticscholar.org/CorpusID: 248512704

2022
[34]

N. M. Glaser, Z. Kira, Communication-critical plan- ning via multi-agent trajectory exchange, in: 2023 IEEE International Conference on Robotics and Au- tomation (ICRA), 2023, pp. 3468–3475. doi:10.1109/ ICRA48891.2023.10160880

work page arXiv 2023
[35]

Y . Li, J. Zhang, D. Ma, Y . Wang, C. Feng, Multi- robot scene completion: Towards task-agnostic collabo- rative perception, in: 6th Annual Conference on Robot Learning, 2022. URL:https://openreview.net/ forum?id=hW0tcXOJas2

2022
[36]

Zhang, Z

X. Zhang, Z. Zhou, Z. Wang, Y . Ji, Y . Huang, H. Chen, Co-mtp: A cooperative trajectory prediction frame- work with multi-temporal fusion for autonomous driv- ing, 2025. URL:https://arxiv.org/abs/2502. 16589.arXiv:2502.16589

work page arXiv 2025
[37]

Y . Hu, J. Peng, S. Liu, J. Ge, S. Liu, S. Chen, Communication-efficient collaborative perception via information filling with codebook, 2024. URL: https://arxiv.org/abs/2405.04966. arXiv:2405.04966

work page arXiv 2024
[38]

Z. Ding, J. Fu, S. Liu, H. Li, S. Chen, H. Li, S. Zhang, X. Zhou, Point cluster: A compact message unit for communication-efficient collaborative perception, in: The Thirteenth International Conference on Learning Rep- resentations, 2025. URL:https://openreview. net/forum?id=54XlM8Clkg

2025
[39]

C. E. Rasmussen, C. K. I. Williams, Gaussian Processes for Machine Learning, Adaptive Computation and Ma- chine Learning, MIT Press, Cambridge, MA, 2006

2006
[40]

Lakshminarayanan, A

B. Lakshminarayanan, A. Pritzel, C. Blundell, Simple and scalable predictive uncertainty estimation using deep ensembles, in: Proceedings of the 31st International Conference on Neural Information Processing Systems, NIPS’17, Curran Associates Inc., Red Hook, NY , USA, 2017, p. 6405–6416

2017
[41]

R. Xu, H. Xiang, X. Xia, X. Han, J. Li, J. Ma, Opv2v: An open benchmark dataset and fusion pipeline for per- ception with vehicle-to-vehicle communication, in: 2022 International Conference on Robotics and Automation (ICRA), IEEE, 2022, pp. 2583–2589

2022
[42]

R. Xu, X. Xia, J. Li, H. Li, S. Zhang, Z. Tu, Z. Meng, H. Xiang, X. Dong, R. Song, et al., V2v4real: A real- world large-scale dataset for vehicle-to-vehicle coopera- tive perception, in: Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition, 2023, pp. 13712–13722

2023
[43]

de Ponte Müller, I

F. de Ponte Müller, I. Rashdan, M. Schmidhammer, S. Sand, Its-g5 and c-v2x link level performance measurements, in: Proceedings of the 27th ITS World Congress, Hamburg, Germany, 2021. URL: https://elib.dlr.de/142341/1/ITS-G5_ and_C-V2X_Link_Level_Performance_ Measurements_dePonteMueller_v4.0.pdf, german Aerospace Center (DLR). Field-test report/paper. 18

2021