汽车工程 ›› 2025, Vol. 47 ›› Issue (9): 1655-1664.doi: 10.19562/j.chinasae.qcgc.2025.09.002

• • 上一篇    

自动驾驶车辆模型预测路径跟踪控制的FPGA硬件加速实现

李文昌,赵治国(),梁凯冲,赵坤,于勤   

  1. 同济大学汽车学院,上海 201804
  • 收稿日期:2025-03-14 修回日期:2025-04-17 出版日期:2025-09-25 发布日期:2025-09-19
  • 通讯作者: 赵治国 E-mail:zhiguozhao@tongji.edu.cn
  • 基金资助:
    国家自然科学基金(52172390)

FPGA Hardware-Accelerated Implementation of Model Predictive Path Tracking Control for Autonomous Vehicles

Wenchang Li,Zhiguo Zhao(),Kaichong Liang,Kun Zhao,Qin Yu   

  1. College of Automotive Studies,Tongji University,Shanghai 201804
  • Received:2025-03-14 Revised:2025-04-17 Online:2025-09-25 Published:2025-09-19
  • Contact: Zhiguo Zhao E-mail:zhiguozhao@tongji.edu.cn

摘要:

针对模型预测控制(model predictive control,MPC)在线求解复杂度高、难以在既有自动驾驶车载控制器上实时应用的问题,本文提出一种基于现场可编程门阵列(field-programmable gate array,FPGA)的MPC路径跟踪硬件加速实现方法。首先,设计了自动驾驶车辆MPC路径跟踪控制器;其次,为便于算法的简化求解,将MPC问题转化为约束二次规划优化问题,并引入了Hildreth方法进行求解;之后,为提升控制算法的实时性及硬件部署效率,基于Xilinx System Generator工具,提出了MPC路径跟踪算法的FPGA便捷部署方案;最后,在不同工况下,开展了MATLAB/Simulink-CarSim联合仿真及硬件在环测试。结果表明,该方法可实现自动驾驶车辆精准跟踪期望路径,且FPGA平均计算时间低于0.1 ms,验证了其有效性和实时性。

关键词: 自动驾驶车辆, 模型预测控制, FPGA硬件实现, 路径跟踪

Abstract:

For the high complexity of online solving in model predictive control (MPC) and its challenges in real-time implementation on existing autonomous vehicle onboard controllers, an FPGA hardware-accelerated implementation of MPC-based path tracking method is proposed in this paper. Firstly, an MPC-based path tracking controller for autonomous vehicles is designed. Then, to simplify the solution process, the MPC problem is transformed into a constrained quadratic programming problem, and the Hildreth method is introduced for solving it. Furthermore, to improve the real-time performance and deployment efficiency of the control algorithm, a convenient FPGA implementation scheme for the MPC path tracking algorithm is developed based on the Xilinx System Generator tool. Finally, MATLAB/Simulink-CarSim co-simulation and hardware-in-the-loop (HIL) tests are conducted under different conditions. The results show that the proposed method enables autonomous vehicles to accurately track the desired path, with an average FPGA computation time of less than 0.1 ms, validating its effectiveness and real-time performance.

Key words: autonomous vehicles, model predictive control, FPGA implementation, path tracking