Accuracy-Aware Cooperative Sensing and Computing for Connected Autonomous Vehicles

Xuehan Ye,Kaige Qu,Weihua Zhuang,Xuemin Shen

2024-03-25

Abstract:To maintain high perception performance among connected and autonomous vehicles (CAVs), in this paper, we propose an accuracy-aware and resource-efficient raw-level cooperative sensing and computing scheme among CAVs and road-side infrastructure. The scheme enables fined-grained partial raw sensing data selection, transmission, fusion, and processing in per-object granularity, by exploiting the parallelism among object classification subtasks associated with each object. A supervised learning model is trained to capture the relationship between the object classification accuracy and the data quality of selected object sensing data, facilitating accuracy-aware sensing data selection. We formulate an optimization problem for joint sensing data selection, subtask placement and resource allocation among multiple object classification subtasks, to minimize the total resource cost while satisfying the delay and accuracy requirements. A genetic algorithm based iterative solution is proposed for the optimization problem. Simulation results demonstrate the accuracy awareness and resource efficiency achieved by the proposed cooperative sensing and computing scheme, in comparison with benchmark solutions.

Networking and Internet Architecture,Signal Processing

What problem does this paper attempt to address?

The focus of this paper is on achieving high-precision perception performance in autonomous vehicles (CAVs). Existing onboard sensors such as cameras and LiDAR may be limited by line-of-sight obstructions, leading to incomplete or inaccurate perception of the surrounding environment. To address this issue, the paper proposes a raw-level collaborative perception and computation scheme based on accuracy-awareness and resource efficiency.

This scheme allows for fine-grained selection, transmission, fusion, and processing of raw perception data at each object level, leveraging parallelism between object-related subtasks. A supervised learning model is used to capture the relationship between object classification accuracy and the quality of selected object perception data, enabling accuracy-aware data selection. The paper also formulates an optimization problem aiming to minimize total resource cost while meeting latency and accuracy requirements. To solve this problem, an iteration-based solution using a genetic algorithm is proposed.

The research demonstrates that the proposed collaborative perception and computation scheme is both accuracy-aware and resource-efficient compared to baseline approaches. Furthermore, the paper explores optimizations for task allocation and computational resource allocation, as well as distributed computing between edge servers or vehicles to reduce latency and enhance overall resource efficiency.

In conclusion, this paper attempts to address the issue of how to efficiently and accurately acquire and process environment perception data in autonomous vehicles through collaborative perception and computation techniques, overcoming the limitations of individual onboard sensors and improving autonomous driving perception performance.