The Improvement of SOME/IP Service Discovery via Association Rule Mining
The Improvement of SOME/IP Service Discovery via Association Rule Mining
复制标题
通过关联规则挖掘改进SOME/IP服务发现
DOI：
--
复制
发表时间：
2022
期刊：
影响因子：
--
通讯作者：
Berkay Saydam
中科院分区：
文献类型：
--
作者：
Berkay Saydam
IP上可扩展的面向服务的中间件是一种中间件解决方案，它实现了电子控制单元之间的面向服务的体系结构。这种体系结构降低了数据路径的复杂性，提供了所需的灵活性，并轻松地向车辆系统添加了新的软件功能。服务发现是自动检测网络上的服务和设备的技术。请求数据的节点成为客户端，提供数据的节点成为服务器。在服务发现期间，服务可能会错过客户端的订阅尝试，因为它们的节点在网络中的客户端节点之后运行。当没有组织协调服务和客户端时，这种情况会增加系统启动时间。关联规则挖掘是一种机器学习方法，用于发现数据集之间的关联和关系。通过这种方法可以确定服务和客户端之间的关系。可以根据频繁使用的服务和客户订单来提高启动时间。通过调整，降低了订阅尝试的错失率。这项研究的目的是减少网络启动时间，而服务发现过程会增加网络启动时间。该网络具有四个服务和八个客户端，是作为四个电子控制单元的模拟创建的。这项研究的结果是，即使只有八个客户端，网络启动时间也几乎缩短了三倍。研究中使用的算法具有可接受的计算复杂度，这是嵌入式系统具有较低运行时间开销的一个重要参数。
Scalable service-Oriented MiddlewarE over IP is a middleware solution that enables service-oriented architecture between the Electronic Control Units. This architecture provides reduced complexity of the data path, the required flexibility and adding new software functions to the vehicle system easily. Service discovery is the technology to automatically detect services and devices on a network. The node which requests for data becomes the client and the node which provides the data becomes the server. Services can miss subscription attempts by a client during service discovery since their nodes are running after client node in a network. This situation increases system startup time when there is no organization to align services and clients. Association Rule Mining is a machine learning method to find associations and relationships among sets of data. The relationships between services and clients can be determined via this method. The startup time can be improved according to frequently used services and clients order. The rate of missing subscription attempts is reduced with alignment. The purpose of this study is reducing network startup time, which is increased by the service discovery process. The network, which has four services and eight clients, is created as a simulation of four Electronic Control Units. As a result of this study, the network startup time is improved almost three times even for only eight clients. The algorithm used in the study has an acceptable computational complexity which is an important parameter for embedded systems to have low runtime overhead as a novelty.