Effects of V2V Adoption Rates on DDPG-Based Cooperative Adaptive Cruise Control for Electric Vehicles

Cooperative Adaptive Cruise Control (CACC) has the potential to improve vehicular safety and energy efficiency through the use of Vehicle-to-Vehicle (V2V) communication. Although prior studies typically assume either complete absence or full deployment of V2V technology, this study addresses the challenges of its incremental integration. Six reinforcement learning agents based on Deep Deterministic Policy Gradient (DDPG) are trained under varying V 2 V adoption levels, ranging from 0 to 100 %, and evaluated in different adoption scenarios using the FTP-75 and WLTP drive cycles. Simulation results showed that each reinforcement learning agent achieved its highest energy efficiency when operating in environments that matched its training conditions. The agent trained under full V2V adoption achieved the best overall performance in the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$100 \% ~\mathrm{V} 2 ~\mathrm{V}$</tex> scenario, but exhibited the poorest performance when evaluated under the 0 % adoption scenario. Agents trained at moderate adoption rates (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$40-60 \%$</tex>) showed the most consistent performance across all testing scenarios, suggesting greater adaptability to mixed real-world conditions.