Wang Yukang

Wuhan Business University

Abstract：

In modern intelligent vehicles, the increasing complexity of driving environments and control requirements has exposed the limitations of traditional rule-based automotive control systems. Although the Controller Area Network (CAN) bus provides a reliable and efficient communication mechanism for distributed electronic control units (ECUs), its integration with conventional control strategies lacks adaptability to nonlinear and uncertain driving conditions. To address this issue, this paper proposes an automotive drive control system that integrates artificial intelligence algorithms within a CAN bus-based distributed architecture. A neural network-based control model is designed to approximate nonlinear vehicle dynamics, while a reinforcement learning framework is introduced to enable adaptive optimization under varying operational conditions. The communication mechanism is optimized to support real-time AI-driven decision-making within the constraints of CAN bandwidth. A hardware-in-the-loop (HIL) simulation platform is constructed to validate the proposed system. Experimental results demonstrate that the proposed method significantly improves response speed, tracking accuracy, and energy efficiency compared with traditional PID-based control approaches, while maintaining stable CAN communication performance. The proposed design provides an effective solution for intelligent vehicle control systems and demonstrates the feasibility of integrating AI into resource-constrained in-vehicle networks.

Key Words：

CAN bus; intelligent control; reinforcement learning; automotive electronics; neural network