EcoCert-V2X: Risk-Aware Adaptive Cooperation for Communication-Efficient V2X Cooperative Perception

Cooperative perception can improve the situational awareness of connected and autonomous vehicles by exchanging complementary sensing information among nearby agents. However, systematic multi-vehicle cooperation also increases vehicle-to-everything (V2X) communication demand, processing overhead, and operational resource use. This study introduces EcoCert-V2X, a risk-aware adaptive cooperation framework that selects the level of cooperative feature acquisition according to action-admissibility scores derived from the pre-communication operating context. Rather than transmitting all available peer information systematically, EcoCert-V2X selects among a frozen set of cooperation actions while accounting for perception quality, communication volume, latency, graphics processing unit (GPU) action energy, modeled communication energy, and operational carbon. The framework is evaluated using a leakage-controlled development, calibration, and locked confirmatory protocol on the V2V4Real benchmark, which contains real-world vehicle-to-vehicle light detection and ranging (LiDAR) observations. Across 1846 confirmatory samples from eight scenarios and 30 paired operational seeds, EcoCert-V2X achieved a scenario-balanced task risk of 0.6513 compared with 0.6433 for full cooperation. The resulting risk difference of +0.0080 satisfied the prespecified non-inferiority criterion, with a one-sided 95% upper bound of +0.0188 below the margin of 0.03. Communication volume decreased by 14.94%, while operational carbon decreased by 4.07%. The observed GPU action energy reduction remained statistically inconclusive, and the absolute risk target of 0.5730 was not reached. Under six frozen packet erasure and signal-to-noise ratio (SNR) perturbation conditions, aggregate non-inferiority was preserved, although localized scenario–seed sensitivity remained. These results show that risk-aware adaptive V2X cooperation can reduce communication demand while preserving relative cooperative perception performance in connected-vehicle environments.