基于改进连续型Hopfield神经网络的CAN总线负载率优化  

CAN Bus Load Rate Optimization Based on Improved Continuous Hopfield Neural Network

作　　者：何智成[1] 杜磊浩 周恩临 覃高峰 黄晋 He Zhicheng;Du Leihao;Zhou Enlin;Qin Gaofeng;Huang Jin(Hunan University,State Key Laboratory of Advanced Design and Manufacturing for Vehicle Body,Changsha 410082;SAIC GM Wuling Automobile Co.,Ltd.,Liuzhou 545007;School of Vehicle and Mobility,Tsinghua University,Beijing 100084)

机构地区：[1]湖南大学,汽车车身先进设计制造国家重点实验室,长沙410082 [2]上汽通用五菱汽车股份有限公司,柳州545007 [3]清华大学车辆与运载学院,北京100084

出　　处：《汽车工程》2023年第12期2338-2347,共10页Automotive Engineering

基　　金：国家自然科学基金联合基金(U20A20285);湖南省杰出青年基金(2021JJ10016)资助。

摘　　要：CAN总线负载率对于车载总线的安全性与时延性有至关重要的作用。鉴于传统的连续型Hopfield神经网络(CHNN)在解决此类问题时存在惩罚参数鲁棒性差和所得解易陷入局部最优的缺陷,本文借助模拟退火算法中的蒙特卡洛思想,提出应用于CAN总线负载率优化问题的改进连续性Hopfield神经网络算法(SA-CHNN)。选取微型电动车中99条通信信号作为实验数据进行测试,结果表明,SA-CHNN算法成功解决传统CHNN算法求解CAN总线负载率优化问题的不足,具有明显的优越性。最后,基于搭建的Simulink-Speedgoat CAN总线实验平台对SACHNN算法得到的最优信号分配的报文进行实时负载率仿真验证,结果表明SA-CHNN算法的准确性。The Busload rate of CAN bus is of vital importance to the security and delay bounds of Vehicle-Bus.Whereas,the traditional continuous Hopfield neural network(Continuous Hopfield Neural Network,CHNN)has the defects of poor penalty parameter robustness and the resulting solution that easy to be trapped in local opti-mality when solving such problems,based on the Metropolis thought in the simulated annealing algorithm,an im-proved continuous Hopfield neural network algorithm(SA-CHNN)applied to the CAN bus load rate optimization problem is proposed in this paper.Micro-electric vehicle’s ninety-nine communication signals are selected and test-ed as experimental data.The results show that the SA-CHNN algorithm successfully solves the problems of the tradi-tional CHNN algorithm in solving the CAN bus load rate optimization,which has obvious advantages.Finally,based on the Simulink-Speedgoat CAN bus experimental platform,the real-time load rate simulation with the opti-mal signal distribution messages is conducted and the result reveals the accuracy of the SA-CHNN algorithm.

关 键 词：CAN总线 负载率 神经网络 模拟退火法 METROPOLIS准则 

分 类 号：U463.6[机械工程—车辆工程]