基于UML活动图的无人车变道可靠性预测  

Reliability Prediction of Unmanned Vehicle Lane Changing Based on UML Activity Diagram

作　　者：吴兆贤 吴培培 WU Zhao-xian;WU Pei-pei(Information Academy,Zhejiang Sci-Tech University,Hangzhou 310018,China)

机构地区：[1]浙江理工大学信息学院,浙江杭州310018

出　　处：《软件导刊》2022年第1期101-107,共7页Software Guide

基　　金：国家自然科学基金项目(61210004,61170015)。

摘　　要：车辆变道是很频繁的驾驶行为,对无人车的变道行为进行可靠性预测尤为必要。为更好地实现可靠性预测,使用UML活动图对基于长短期记忆网络(LSTM)的无人车变道行为进行建模。考虑到LSTM神经网络的鲁棒性,基于6000组实验数据对LSTM模块的错误渗透率加以计算,从而度量模块间的故障传播概率。将UML模型转换为离散时间马尔可夫链(DTMC)模型,通过PRISM工具得到可靠性预测值。实验结果显示,LSTM模块的错误渗透率为0.3025。通过基于UML模型映射和基于构件转移图所得的DTMC模型进行预测时,可靠性值分别为76.47%和90.19%。结果表明,对基于LSTM的无人车变道行为进行可靠性计算时,LSTM模块的错误渗透率不可忽视。通过映射所得的DTMC模型对模块的刻画更为细致,更适用于无人车变道可靠性预测。Vehicle lane changing is a very frequent driving behavior,so it is necessary to predict the reliability of unmanned vehicle lane changing behavior.In order to better predict the reliability,UML activity diagram is used to model the lane changing behavior of unmanned vehicle based on long term memory network(LSTM).Considering the robustness of LSTM neural network,6000 sets of ex⁃perimental data are used to calculate the error permeability of LSTM modules,so as to measure the fault propagation probability be⁃tween modules.The UML model is transformed into a discrete-time Markov chain(DTMC)model,and the reliability prediction value is obtained by prism.The experimental results show that the error permeability of LSTM module is 0.3025.When the DTMC model based on UML model mapping and component transfer diagram is used for prediction,the reliability values are 76.47%and 90.19%re⁃spectively.The results show that the error permeability of LSTM module can not be ignored in the reliability calculation of lane chang⁃ing behavior of unmanned vehicle based on LSTM.The DTMC model obtained by mapping is more detailed to describe the modules,which is more suitable for the reliability prediction of unmanned vehicle lane changing.

关 键 词：无人车变道 可靠性 UML 错误渗透率 离散时间马尔可夫链 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]