基于AUTOSAR的汽车控制器软件优化部署研究  

Optimized Deployment Development of Automotive Controller Software Based on AUTOSAR

作　　者：邹渊[1] 马文斌 张旭东[1] 翟建阳 张兆龙 ZOU Yuan;MA Wenbin;ZHANG Xudong;ZHAI Jianyang;ZHANG Zhaolong(School of Mechanical Engineering,Beijing Institute of Technology,Beijing 100081,China;Beijing Electric Vehicle Co.,Ltd.,Beijing 100176,China)

机构地区：[1]北京理工大学机械与车辆学院,北京100081 [2]北京新能源汽车股份有限公司,北京100176

出　　处：《北京理工大学学报》2024年第11期1192-1198,共7页Transactions of Beijing Institute of Technology

基　　金：国家重点研发计划(2021YFB2500900)资助。

摘　　要：针对基于AUTOSAR的汽车控制器软件开发过程中SW-C到ECU、Runnable到OsTask以及OsTask到多核ECU中Core的软件优化部署问题,面向工程应用需求,建立了基于AUTOSAR的汽车控制器软件拓扑和优化部署模型,提出了一种基于D2RL和PER改进的SAC深度强化学习求解框架.仿真实验显示所提方法相比于常用启发式算法在ECU核心负载均衡、OsTask栈空间利用率以及ECU之间和Core之间通信带宽利用率等具有优越性和稳定性.In order to deal with the optimal deployment of software from SW-C(SoftWare-Component)to ECU(Electric Control Unit),from Runnable to OsTask(Operation System Task)and from OsTask to Core in multi-core ECU in the software development process for AUTOSAR-based automotive controller,AUTOSAR-based software topology and optimal deployment model of automotive controller were constructed for practical engin-eering application requirements.Firstly,an improved SAC(Soft Actor-Critic)deep reinforcement learning solv-er framework was proposed based on D2RL(deep dense architecture in reinforcement learning)and PER(priorit-ized experience replay).And then some simulation experiments were carried out to demonstrate the proposed method.Results show the superior performance and stability of the new method,compared with commonly used heuristic algorithms in terms of ECU core load balancing,OsTask stack space utilization,as well as the utiliza-tion of communication bandwidth between ECUs and among cores.

关 键 词：汽车开放系统架构 软件优化部署 深度强化学习 改进SAC 

分 类 号：U462.1[机械工程—车辆工程]