Mitigating Packet Collisions by Predicting Collective Perception Message Transmissions in Cellular V2X Environment

Collective perception is known to effectively boost awareness in vehicle-to-everything (V2X) environment. However, the Intelligent Transportation System (ITS) frequency band is narrow, and channel congestion may occur due to frequent exchange of Collective Perception Messages (CPM). For this reason, European Telecommunications Standards Institute (ETSI) is standardizing a facilities-layer congestion control for CPM traffic, called the CPM generation rule. Unfortunately, it changes the periodic CPM traffic into an aperiodic one, causing disruption for the Sensing-based Semi-Persistent Scheduling (SPS) algorithm used in New Radio (NR) Sidelink Mode 2 that reserves periodic transmit resources. Therefore, to solve this problem, this paper proposes a deep learning-based scheme that predicts when the sensed vehicles will qualify to be included in the CPM. Based on the prediction, the proposed scheme reserves transmit resource where there will be vehicles to report about. It shows that the number of wasted resources and packet collisions can be greatly reduced compared to using standard SPS. Consequently, the packet reception ratio (PRR) is increased and the target PRR for safety applications can be achieved at much longer distances.