Cooperative perception among autonomous agents overcomes the limitations of single-agent sensing, but bandwidth constraints in vehicle-to-everything (V2X) networks require efficient communication policies. Existing approaches rely on reactive mechanisms, such as confidence maps, learned gating, or sparse masks, to decide what to transmit, without reasoning about why a message benefits the receiver. We introduce Reason-to-Transmit (R2T), a framework that equips each agent with a lightweight transformer-based module that reasons over local scene context, estimated neighbor information gaps, and bandwidth budget to make per-region transmission decisions. Trained end-to-end with a bandwidth-aware objective, R2T is evaluated against nine baselines in a multi-agent bird's-eye-view perception environment. Any communication improves performance by about 58% AP over no communication. At low bandwidth, all selective methods perform similarly, but R2T shows clear gains under high occlusion, where information asymmetry is greatest, approaching oracle performance. All methods degrade gracefully under packet drops up to 50%, showing robustness to communication failures. These results indicate that while fusion design dominates performance, deliberative communication provides additional gains in challenging scenarios. R2T introduces a reasoning-based approach to communication, enabling more efficient and context-aware information sharing in cooperative perception.
Figures & tables
Figure 1: Overview of the Reason-to-Transmit (R2T) framework. Each agent encodes local observations into a BEV feature map, generates spatial region candidates, and applies a reasoning-based transmit policy that jointly considers local features, neighbor state estimates, and the bandwidth budget. Selected regions are transmitted to neighbors and fused via gated cross-attention before final detection.
Figure 2: Communication decision pipeline. The reasoning module attends over region tokens augmented with agent context, neighbor state estimates, and a budget token, producing per-region transmit probabilities under the bandwidth constraint.
Method
10% (2.4 KB)
50% (12 KB)
100% (24 KB)
No Comm.
0.1452 ± 0.0534
Random
0.2306 ± 0.0195
0.2297 ± 0.0188
0.2304 ± 0.0198
Confidence
0.2274 ± 0.0200
0.2294 ± 0.0189
0.2304 ± 0.0198
Uncertainty
0.2272 ± 0.0195
0.2291 ± 0.0189
0.2304 ± 0.0198
Where2Comm
0.2278 ± 0.0193
0.2299 ± 0.0202
0.2304 ± 0.0198
IC3Net Gate
0.2316 ± 0.0209
0.2296 ± 0.0194
0.2304 ± 0.0198
Table 1: Detection performance (AP) at various bandwidth budgets. Mean ± std over 5 seeds. Communication yields ∼58% improvement over isolation. All selective policies are competitive; bold indicates best per column.
Figure 3: AP vs. bandwidth budget. All selective methods dramatically outperform No Communication. Error bands show ± 1 std over 5 seeds. The secondary axis shows bandwidth in KB.
Method
Low
Medium
High
No Comm.
0.188 ± 0.003
0.181 ± 0.002
0.159 ± 0.003
Confidence
0.236 ± 0.005
0.223 ± 0.006
0.202 ± 0.006
Where2Comm
0.235 ± 0.006
0.226 ± 0.005
0.203 ± 0.005
IC3Net
0.234 ± 0.006
0.227 ± 0.004
0.202 ± 0.007
R2T
0.232 ± 0.006
0.225 ± 0.006
0.205 ± 0.006
Always
0.236 ± 0.005
0.223 ± 0.006
0.203 ± 0.004
Table 2: AP by occlusion level at 50% bandwidth. Mean ± std over 3 seeds. R2T leads under high occlusion, closest to oracle.
Drop Rate
R2T
Where2Comm
IC3Net
Conf.
0%
0.225 ± 0.006
0.226 ± 0.005
0.227 ± 0.004
0.223 ± 0.006
10%
0.225 ± 0.005
0.226 ± 0.005
0.226 ± 0.004
0.224 ± 0.006
20%
0.224 ± 0.005
0.227 ± 0.007
0.227 ± 0.006
0.224 ± 0.006
30%
0.224 ± 0.004
0.227 ± 0.004
0.227 ± 0.005
0.225 ± 0.005
50%
0.223 ± 0.006
0.225 ± 0.004
0.227 ± 0.005
0.223 ± 0.005
Table 3: AP under varying packet drop rates at 50% bandwidth. Mean ± std over 3 seeds. R2T degrades gracefully with increasing packet loss.
Figure 4: Qualitative visualization of communication decisions. R2T selectively transmits regions containing occluded objects that the receiver cannot observe, while suppressing regions with redundant information.
Communication-efficient collaborative perception studies how networked agents share useful sensor information with minimal bandwidth while improving their joint view of the scene.
Collaborative perception improves autonomous perception by sharing intermediate Bird's-Eye-View (BEV) features across connected agents, but dense feature exchange is difficult to deploy under strict Vehicle-to-Everything (V2X) bandwidth limits. Existing efficient methods typically either compress the full feature map uniformly, spending bits on low-value background, or sparsify communication, risking the loss of useful context. We propose a coverage-refinement design for byte-constrained cooperative perception: each agent transmits a highly compressed coarse layer over the full BEV map and allocates the remaining budget to selected high-resolution patches. A Task-Aware Benefit Selector ranks cells by estimated downstream utility, enabling deterministic budgeted refinement and zero-retraining adaptation to changing bandwidth. The receiver reconstructs a dense BEV tensor compatible with standard fusion modules. Experiments on DAIR-V2X and OPV2V show strong accuracy-payload trade-offs at kilobyte-scale budgets. On DAIR-V2X, our method reaches 0.60 [email protected] at only 1.87 KB per non-ego agent, compared with 0.52 at 4.61 KB for uniform SimVQ compression. Controlled diagnostics further show that the gain arises from coverage-refinement allocation rather than quantization alone. Code will be published.
Melih Yazgan, Timon Müller, J. Marius Zöllner
FZI Research Center for Information Technology · Karlsruhe Institute of Technology
Collaborative-perception enables multi-robot systems to enhance situational awareness by sharing perceptual information. Existing collaborative-perception systems face an inherent trade-off between communication bandwidth requirements and perception accuracy, where methods that exchange more information achieve better perception results at the cost of increased communication overhead. However, real-world communication networks impose bandwidth constraints that require minimizing communication overhead without sacrificing perception performance. To address this challenge, we propose HydraCollab, an adaptive collaborative-perception framework that (i) selectively transmits the most informative sensor features and (ii) dynamically employs collaboration strategies (intermediate or late) based on spatial confidence maps. Extensive evaluations on the V2X-R, V2X-Radar and UAV3D-mini datasets demonstrate that HydraCollab achieves the best overall trade-off between accuracy and communication cost among existing collaborative-perception methods. Relative to SOTA Where2comm, HydraCollab uses only 41% of the bandwidth on V2X-R and 26% on V2X-Radar while improving performance by 0.78% and 0.75% respectively. Our code and models are available at https://github.com/AICPS/HydraCollab.
Luke Chen, Cheng-Ju Wu, David R. Martin +3
Department of Electrical Engineering and Computer Science, University of California, Irvine, USA.
Vehicle-to-Everything (V2X) cooperative perception improves 3-D detection by sharing intermediate features, but dense remote features may repeat context that the ego agent can infer locally. Most communication-efficient designs optimize masks or codes empirically, leaving a more basic question open: which remote evidence is indispensable given the receiver's own observation? We introduce a closure-fidelity perspective on ego conditioned remote perception. Under a finite deductive abstraction and explicit conditions, its rate--distortion function decomposes over an irredundant core, and the exact zero-distortion rate becomes PAH(πA). This analysis suggests a concrete design principle: transmit compact evidence and recover derivable context with bounded receiver-side inference. Guided by this principle, SemRD-V2X is an operational neural proxy that combines exact-budget BEV support selection, pointwise channel compression, and masked shared-weight reconstruction before standard fusion. Experiments on simulated V2XSet and real-world DAIR-V2X validate the resulting design. In a controlled five-run V2XSet comparison against a locally reproduced V2X-ViT-v1 baseline on one Tesla V100, SemRD-V2X reduces the analytical feature payload by 26.6× while improving [email protected]/[email protected] by 4.13/8.57 points, with 3.81% additional mean compute latency. These results position closure fidelity as both an analytical lens and an actionable design principle for communication-efficient cooperative perception.