Performance Evaluation of Object Detection Considering C-V2X Communication Errors

Abstract

To cope with the issue of high computational loads in vehicles equipped with multiple sensors for safe autonomous driving, vehicular edge computing (VEC) has been proposed. When applying VEC, it is important to analyze the recognition performance considering communication errors because data quality can be degraded due to communication errors, which can result in decreased recognition or processing performance. Previous studies have mainly focused on aspects such as delay and cost. However, there is a need for analysis specifically addressing recognition performance in light of communication errors. This paper suggests the optimal sensor resolution to enhance object detection performance based on the signal-to-noise ratio (SNR). We analyze the performance of each resolution according to SNR. Based on these results, we anticipate that selecting the optimal resolution for object detection based on channel conditions during offloading can further improve driving safety. © 2023 IEEE.