Tsukada Lab. Computer Network Group

Implementation and Evaluation of Late Stage Cooperative Multi-Object Tracking in Urban Scenarios

Mikel García, Ehsan Javanmardi, Manabu Tsukada, Marcos Nieto, Naiara Aginako, Oihana Otaegui

IEEE Transactions on Vehicular Technology, Jul 27, 2026

Summary in three lines

  • A cooperative multi-object tracker integrates local detections with positions received through standard C-ITS messages.
  • The implementation evaluates four association strategies while accounting for communication and processing delays.
  • Urban evaluations report more accurate tracking than local-only perception in the reported scenarios.

Abstract

Future mobility based on Cooperative Intelligent Transport Systems (C-ITS) will allow Vehicle-to-Everything (V2X) communications to exchange information with their surrounding infrastructure and traffic participants. Cooperative Perception (CP) extends the sensing capabilities of each Intelligent Transport System Station (ITS-S) by exchanging standardized data such as Collective Perception Messages (CPM). An example is Cooperative Perception Services (CPS), which extend the sensing capabilities of C-ITS by exchanging standardised perception data. The challenge is to efficiently and robustly fuse received data with own data, accounting for the communication and processing latencies. In this work, we implement a Cooperative Multi Object Tracker (C-MOT) for CPS that fuses its local perception information with the information received from multiple surrounding ITS-S via V2X. Specifically, our method leverages object positions from standardized ETSI messages, including Cooperative Awareness Message (CAM) and cooperative perception data from CPMs. We compare four different association strategies and evaluate the proposed C-MOT in simulations across multiple scenario configurations. Comparing the results obtained by using only local perception data against the proposed C-MOT shows improvements of up to 60% in accuracy while also obtaining more precise positions of the tracks in certain scenarios.

How to cite

M. García, E. Javanmardi, M. Tsukada, M. Nieto, N. Aginako, and O. Otaegui, “Implementation and Evaluation of Late Stage Cooperative Multi-Object Tracking in Urban Scenarios,” IEEE Transactions on Vehicular Technology, 2026, doi: 10.1109/TVT.2026.3718064.

BibTeX

@article{García2026,
  title         = {Implementation and Evaluation of Late Stage Cooperative Multi-Object Tracking in Urban Scenarios},
  author        = {Mikel García and Ehsan Javanmardi and Manabu Tsukada and Marcos Nieto and Naiara Aginako and Oihana Otaegui},
  doi           = {10.1109/TVT.2026.3718064},
  year          = {2026},
  date          = {2026-07-27},
  urldate       = {2026-07-27},
  journal       = {IEEE Transactions on Vehicular Technology},
  abstract      = {Future mobility based on Cooperative Intelligent Transport Systems (C-ITS) will allow Vehicle-to-Everything (V2X) communications to exchange information with their surrounding infrastructure and traffic participants. Cooperative Perception (CP) extends the sensing capabilities of each Intelligent Transport System Station (ITS-S) by exchanging standardized data such as Collective Perception Messages (CPM). An example is Cooperative Perception Services (CPS), which extend the sensing capabilities of C-ITS by exchanging standardised perception data. The challenge is to efficiently and robustly fuse received data with own data, accounting for the communication and processing latencies. In this work, we implement a Cooperative Multi Object Tracker (C-MOT) for CPS that fuses its local perception information with the information received from multiple surrounding ITS-S via V2X. Specifically, our method leverages object positions from standardized ETSI messages, including Cooperative Awareness Message (CAM) and cooperative perception data from CPMs. We compare four different association strategies and evaluate the proposed C-MOT in simulations across multiple scenario configurations. Comparing the results obtained by using only local perception data against the proposed C-MOT shows improvements of up to 60% in accuracy while also obtaining more precise positions of the tracks in certain scenarios.},
  keywords      = {ITS},
  pubstate      = {published},
  tppubtype     = {article}
}