Density-Adaptive RSU Sleeping and Opportunistic Offloading for Energy Efficiency Optimization in Connected Autonomous Vehicles
Year: 2025
Type: preprint
Abstract: This study addresses energy consumption in connected autonomous vehicle (CAV) networks by exploiting underutilized computing resources in charging CAVs and optimizing roadside unit (RSU) operations. Prior works typically address RSU sleep scheduling and dynamic voltage and frequency scaling (DVFS) for CAVs as separate problems, overlooking the critical interdependencies among RSU operations, on-road traffic, and the computing capacity of charging CAVs. Consequently, these approaches result in significant energy inefficiencies under realistic traffic conditions. To address this challenge, we first formulate a minimization problem that captures the interdependencies between RSU activation decisions, coupled with continuous task partitioning and DVFS for both on-road and charging CAVs while ensuring latency and computation constraints for perception tasks. Then, we propose a traffic density-adaptive framework that jointly integrates RSU sleep scheduling, DVFS, and load partitioning tailored for perception tasks. By leveraging the Benders decomposition (BD) method, the resulting mixed-integer nonlinear programming problem is efficiently decomposed into a master problem (that handles binary sleep decisions) and a convex subproblem (that manages continuous load partitioning and DVFS), then solved iteratively until convergence. Simulation results under realistic traffic profiles and heterogeneous resource scenarios reveal that the proposed BD method achieves optimal energy savings, reducing the normalized per-device energy consumption by up to 16.2% relative to the standalone baseline and by 12% relative to the always-on RSU configurations while outperforming sequential and heuristic approaches.
Authors: Minh Thao Chan, Tan Chen, Zhaojun Nan, Sheng Zhou, Zhisheng Niu
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W4412703003