IoV-Assisted Semantic Communication for Occlusion-Aware Edge Autonomous Driving

Engineering notes

Key topics: autonomous driving system, autonomous driving, bev, perception, prediction, control. See the paper for implementation details and experimental results.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为端到端自动驾驶、BEV感知、3D目标检测、轨迹预测、路径规划、LiDAR感知等高价值论文补充中文说明。

Original abstract

In edge-enabled autonomous driving systems, onboard sensors alone are insufficient to handle occluded objects and unpredictable traffic participants, fundamentally limiting control reliability in complex urban environments. In this paper, we propose an extension of the task-oriented co-design framework for communication, computing, and control by integrating roadside units (RSUs) equipped with cameras via Internet of Vehicles (IoV) infrastructure as a complementary perception source. Building upon the Delay-Aware Trajectoryguided Control Prediction (DTCP) with Joint Source-Channel Coding (JSCC), we introduce an RSU-side occlusion-aware semantic perception module based on BEVFormer. Unlike methods that mainly rely on explicit coordinate alignment between the RSU and the vehicle, the proposed RSU module uses the requesting vehicle state m as a conditioning signal to predict semantic information that is relevant to regions likely to be occluded from the ego vehicle’s viewpoint. To reduce the computational overhead of BEV feature generation, we apply vehicleconditioned dynamic token sparsification based on DynamicViT directly to the BEVFormer perception pipeline. Furthermore, we introduce an occlusion-aware preservation loss that penalizes the sparsification module when tokens containing occlusion-critical objects are removed. The RSU transmits the estimated semantic representation $\hat{s}^{r}$ to the edge agent, which is incorporated into DTCP as complementary information with minimal architectural modification, while explicitly accounting for the RSU processing delay in the overall system delay budget.

Links and sources

• Official / arXiv page

Comments