嵌入式系统研究毕业论文文献翻译.doc

嵌入式系统研究1前言 智能软件Agent是能够为用户执行特定的任务、具有一定程度的智能、能够自主的执行部分任务并以一种合适的方式和环境相互作用的软件程序。Agent 有自主性、响应性、学习能力和社会性等特性。这使得它适合在高度动态的环境下做出及时的响应。嵌入式技术和新一代移动通讯网络的发展使得嵌入式智能设备大量的涌现。这些设备的大都具有嵌入式操作系统的支持, 并运行着越来越丰富的应用程序。如何对这些应用程序进行测试，就成为一个需要研究的课题。本文将Agent技术引入嵌入式智能设备的测试中，使用目标设备Agent, 测试控制Agent, 网络环境Agent分别模拟和处理测试设备，测试工程师和测试环境的复杂性，利用Agent自身具有的特点，提出了一种有效的自动化测试的方法。2相关的研究及本文的思路Agent 所具有的自主性、响应性、学习能力和社会性等特性，使得它很适合处理复杂测试系统中的问题。将Agent技术应用于测试领域已经有一些相关的研究，下面是具体的介绍。Jeongeun Choi 和 Byoungju Choi 1提出了一个基于Agent技术的测试工具，它通过使用用户接口Agent来处理和测试者的交互，使用测试用例选择Agent来进行测试用例的选择，并使用回归测试Agent进行回归测试，从而很好的实现了自动化的软件测试。但是他们提出的测试用例选择技术仅能在大量已有的测试用例中选择最佳的用例，不能减少编写测试用例本身的复杂性。另外一些研究将Agent 技术应用到了某一领域的测试中。Yu Qi、 David Hung 和 Eric Wong 3 提出了一个基于Agent 技术的Web 应用程序测试方法。他们使用Agent技术降低了网络应用测试的复杂性。他们的方法不仅仅适用于Web应用程序的测试，也适合于嵌入式智能设备的测试。在实际测试过程中，测试人员很大的一项任务仍然是编写各种测试用例，要真正的提高测试效率，就要提高测试脚本的通用性，减少测试脚本的变化2 4。常见的嵌入式智能设备测试工具(比如TestQuest)使用图像比对来判断目标设备的状态, 这种方法虽然实现了非侵入性的测试,但是存在两个问题,图片的抓取和传送消耗了大量测试资源,不同手机的用户界面风格变化很大,这使得测试脚本在用来进行新设备的测试时需要对脚本进行维护.然而,实际的设备中,当重要的事件发生时,系统会产生敏感事件来激励相关模块进行处理.如果能够在测试过程中捕获这些敏感事件,就能更加高效的进行测试. 并且,只要操作系统相同,敏感事件的捕获方式也是一样,测试用例不会因为界面的变化而变化。本文用目标设备Agent抽象和捕获测试过程中的敏感事件，并将敏感事件发送给测试控制Agent。而测试控制Agent收到敏感事件后，根据自己的知识，采取相应的动作的机制（如异常处理，重新调度测试等）加以处理。这正好可以作为Agent的推理规则，用Agent的智能性来屏蔽测试过程的复杂性。此外，考虑到目标设备处于复杂的网络环境中，我们利用网络环境Agent来控制目标设备所处的网络信号，从而实现对设备所处网络环境的控制。3  基于Agent的测试系统3.1 总体结构测试系统的最终目标是实现一个移动数字终端的自动化测试平台，能够对移动数字终端上的系统软件和应用层软件建立模块化、可重用的测试脚本库，提高各种黑盒测试的效率，自动化测试结果的生成，并实现测试结果的回放。如图1所示：测试平台分为四层，包括用户接口层，测试控制层，通讯层和设备Agent层。 其中，过程监控模块，测试环境控制模块和设备Agent层采用Agent的思想进行设计，以更好的支持嵌入式设备的测试。图1 测试系统的整体架构3.2 系统基本执行流程测试过程是对真实用户使用手机时“输入反馈”模型的一个模拟。系统的基本的使用用例如下，首先用户通过脚本编辑器或者虚拟手机创建测试脚本并进行基本的配置，然后过程监控模块调度脚本开始执行，当脚本解释器解释执行脚本语句时，脚本解释器通常使用通讯模块向手机发送模拟按键，然脚本解释器进入等待状态，目标设备Agent从通讯模块得到按键，执行按键动作，截取屏幕并返回给测试控制Agent，或者如果手机有敏感世事件的发生（如来电，短信，异常等）也会反馈给测试控制Agent，测试控制Agent接到数据，再次激励脚本运行，并同时将测试结果输入存入测试资源库。脚本继续运行下面的语句，直到运行结束。 3.3 目标设备Agent目标设备Agent驻留在被测试的智能设备中，实现对目标设备的建模。本系统中它接收从PC机中接收到的控制命令，然后在智能设备中进行相应的操作，包括模拟键盘事件，截取屏幕并返回给PC，以及根据知识库中的配置的测试目标，通过推理，有所选择的将必需的状态信息主动通知测试宿主机。3.3.1设备Agent 依赖的决策规则系统事件报告决策规则（SERD-RULE）这个决策规则决定设备Agent在获取到状态信息时，是否将状态信息反馈给上层系统。异常处理规则（EM-RULE）这个决策规则决定了当Agent发现测试中出现异常，要采取的处理方式。3.3.2 Agent 属性目标设备Agent具有下面的属性： 自主性目标设备Agent使得测试用户无需关注设备的细节，它是设备功能和状态的一个很好的抽象，通过目标代理Agent，上层测试框架只需要执行相应的虚拟操作，则目标设备Agent会自动将这些虚拟操作映射为对手机的实际操作，并且手机的响应和其它状态也能够自动的反馈给客户，这使得上层测试和底层的设备具体操作无关。 智能性当目标设备Agent获取到一个变化的被测系统状态或信息时，它会根据推理规则，仅仅把和本次目标相关的信息发送给测试控制Agent。目标设备Agent还能够根据异常处理规则，合理的处理测试中出现的异常。 3.4  测试控制 Agent测试控制Agent 实现了对测试过程的建模。它主要实现三个目标： 在测试过程中，对测试人员感兴趣的事件进行捕获，当这些敏感事件发生后，系统立刻执行预定义的动作，并为测试提供准确和详尽的日志。 通过和脚本解释器交互，实现异步的事件通知和交互测试用例的执行。 处理测试过程中的异常，使得测试过程可以自我修复，进而多个脚本能够按照预先的调度正常运行.3.4.1测试控制Agent的决策规则 事件条件行动规则（ECARULE）这是测试控制Agent所依据的核心规则，它定义了某个测试所关心的事件，并根据比对条件，采取相应的动作。 异常处理规则（EM-RULE）测试控制Agent也遵守异常处理规则，它能够检测测试过程中发生的异常事件，并执行相应的解决方案，使得测试能够顺利的进行。3.4.2测试控制Agent 的属性自主性自主性使得测试人员无需过分关注测试的细节，测试人员仅仅需要选择测试目标，测试控制Agent就能根据知识库中的信息，自动生成ECA规则表，并在脚本运行时，按照ECA规则，捕捉敏感的事件并执行相应的操作。智能性测试控制Agent能够根据用户选择的测试目标，自动生成ECA规则表，并根据ECA规则，进行推理并采取相应的动作。测试控制Agent还能根据异常处理规则来处理测试过程中发身的异常，保障测试过程的自动，平稳进行。  3.5 网络环境Agent网络环境Agent 实现了对网络环境的模拟和控制。为了测试手机的射频性能，通常使用基站模拟系统来提供设备所处的无线网络环境，并且利用编程接口，通过软件对设备的环境进行控制，使得复杂网络环境下对目标设备的测试成为可能。4 系统评估为了评价本系统的有效性，我们设计了一个比较实验。试验在信息产业部泰尔实验室真实的测试项目中进行。我们将测试工程师分为两组，第一组使用本文的系统MobileTest进行测试，第二组使用业界著名的TestQuest Pro 进行测试。测试的依据是中国通讯行业标准的数字移动台测试部分。测试的内容是根据这两个系统各自已经建立好的回归测试用例对新的智能手机进行回归测试，从而比较这两个测试工具的测试效率和脚本的可维护性。 实验度量了两个工具在测试不同种类的测试用例时，自动化执行的用例所占的百分比。从表中可见，虽然TestQuest 在功能测试上有更高的自动化率，在压力测试，多状态测试，多任务测试，临界测试和总计中，MobileTest有更好的测试覆盖率。这是因为在后四种测试中，测试用例需要不断追踪系统的状态，使用图片比对机制的TestQuest 很难完成这些工作，而MobileTest 却能使用Agent返回的敏感事件来进行处理。此外，MobileTest 完成测试任务的时间时15天，比TestQuest 的测试效率要高。5总结本文提出了一个基于Agent技术的嵌入式智能设备的测试方法，它用Agent技术对嵌入式智能设备测试过程中的被测试设备，测试者和网络环境进行模拟和控制，很好的屏蔽了嵌入式智能设备的复杂性，提高了测试的效率和脚本的可重用性。将来的研究工作中，我们会进一步拓展整个测试系统，使之能够支持整个测试的生命周期。本文作者的创新点：本文将软件Agent技术引入到嵌入式智能设备的自动化测试中，利用Agent的特性，很好的屏蔽了嵌入式智能设备测试中，测试者，待测设备，网络环境的复杂性，提高了自动化测试的效率。 来源于： Journal of Electronic Science and Technology附：英文原文Embedded system research1 Introduction Intelligent Software Agent is the ability to perform specific tasks for the user, with a certain degree of intelligence, able to perform some tasks and autonomy in a proper manner and environment interactions software programs. Agent has autonomy, responsiveness, learning and social and other characteristics. This makes it suitable for highly dynamic environment to make a timely response. Embedded technology and the development of new generation mobile communication networks makes the emergence of a large number of embedded intelligent devices. Most of these devices with embedded operating system support, and run an increasingly rich applications. How to test for these applications to be a need to study. This article Intelligent Agent technology into embedded devices test, using the target device Agent, test control Agent, Network Agent and treatment were simulated test equipment, test engineers and test complexity of the environment, the use of Agent has its own characteristics, proposed An effective automated testing. 2 Related research and ideas of this article Agent possess autonomy, responsiveness, learning and social and other features, makes it very suitable for handling complex problems in the test system. Agent technology in the test area will already have some relevant research, the following is a specific description. Jeongeun Choi and Byoungju Choi 1 proposed a testing tool based on Agent technology, which through the use of Agent to handle user interface and test those interactions, the use of Agent to carry out the test case test case selection choices, and use the Agent for regression testing regression testing, so a good automated software testing. But they can only be made in a large number of test selection techniques select test cases has been the best use cases, write test cases can not reduce its complexity. Other studies will Agent technology applied to a field test. Yu Qi, David Hung and Eric Wong 3 proposed a Web-based Agent technology application testing methods. Agent technology to reduce their use of the network application testing complexity. Their method not only for Web application testing, but also for the testing of embedded intelligent devices. In the actual testing process, testers great task is still the preparation of various test cases, to really improve test efficiency, we must improve the general test scripts, reduce test script changes 2 4. Common test tools for embedded smart devices (such as TestQuest) using image comparison to determine the status of the target device, although this method to achieve a non-invasive test, but there are two problems, the image capture and transmission consumes a lot of test resources, the different style of phone's user interface has changed dramatically, which makes the test script used for testing new equipment required for maintenance of the script. However, the actual device, when an important event occurs, the system will generate sensitive events to encourage the relevant module for processing. If during the test capture these sensitive matters, will be able to test more efficient. And, as long as the operating system the same sensitive way to capture the same event, the interface test cases will not change. In this paper, the abstract and the Agent target device during the test capture of sensitive events and control of sensitive events are sent to test Agent. Agent receives the test control sensitive incident, according to their knowledge, take appropriate action mechanisms (such as exception handling, re-scheduling the test, etc.) to be addressed. This is precisely the inference rules as Agent, with the intelligence of Agent to shield the complexity of the testing process. In addition, given the target device is a complex network environment, we use the Internet Agent to control the target device environment in which the network signal, enabling the network environment in which the control device. 3 Agent-Based Test System 3.1 The overall structure Test system is to achieve the ultimate goal of a mobile digital terminal's automated testing platforms to the mobile digital terminal on the system software and application layer software build modular, reusable test script library, to improve the efficiency of various black box testing, automation the generation of test results and test results to achieve playback. Figure 1: Test platform divided into four layers, including user interface layer, the test control layer, communication layer and the equipment Agent layer. Among them, the process monitoring module, test equipment, environmental control module and the use of Agent Agent layer design ideas to better support embedded devices testing.  Figure 1 The overall architecture test system3.2 The basic system implementation process 待添加的隐藏文字内容3Testing process is a real cell phone users "input - feedback" model of a simulation. The system's basic use with the example, the first user through the Script Editor or the virtual phone to create test scripts and make basic configuration, then the process monitoring module dispatch script started executing, when the script interpreter interprets the script statement, the script interpreter is commonly used communication module to transmit analog phone keypad, then the script interpreter to enter a wait state, the target device by Agent button from the communication module, the implementation of key actions, the interception of the screen and return to the Test Control Agent, or if the phone has a sensitive world events (such as call , text messages, exceptions, etc.) will be fed back to the test control Agent, Agent receiving the test control data, run the script again to inspire, and while the test results entered into the repository. The script continues to run the following statement, until the end of operation. 3.3 Target Device Agent Agent resides on the target device being tested intelligence equipment, to achieve the target device modeling. This system it receives from the PC, the received control command, and then in the corresponding smart device operations, including simulated keyboard events, the interception of the screen and return to the PC, and configured according to the knowledge base test objectives, by reasoning , have a choice to take the initiative to notify the state information necessary to test the host. 3.3.1 Device dependent decision rules Agent Event reporting system decision rules (SERD-RULE) the decision rule to determine the device status Agent in obtaining information, whether the state information back to the upper system. Exception handling rules (EM-RULE) The decision rules that determine when the Agent abnormal test, to take the treatment. 3.3.2 Agent Properties Agent target device has the following attributes: autonomy Agent makes the test target device users do not concern the details of the device, which is the device function and status of a good abstraction, through the target agent Agent, the upper test framework need only perform the appropriate virtual operation, the target device, these virtual Agent automatically operation mapped to the actual operation of the mobile phone and mobile phone and other states can respond automatically to the customer feedback, which makes the top and bottom of the device test has nothing to do specific operations. intelligence Agent when the target device under test to obtain a change of system state or the information, it will under the rules of inference, and the second goal just to send information related to the test control Agent. Target Device Agent also exception handling according to the rules appear reasonable exception handling test. 3.4 Test Control Agent Test Control Agent to achieve the modeling of the testing process. It is mainly to achieve three objectives: in the testing process, the test to capture the events of interest, when these sensitive events, the system pre-defined implementation of the action immediately, and to provide accurate and detailed test log. and script interpreter through interactive, asynchronous event notification and interactive execution of test cases. exception handling testing process, making the testing process can be self-healing, and then several scripts in advance of the schedule in accordance with normal operation. 3.4.1 Test Control Agent of the decision rules event - condition - action rules (ECA-RULE) Control Agent This is a test based on the core rules, which defines a test case of interest, and according to comparison conditions, take appropriate action. exception handling rules (EM-RULE) Test Control Agent also comply with exception handling rules, which occurred during the test can detect abnormal events and to implement appropriate solutions to enable testing to proceed smoothly. 3.4.2 Test Control Agent's Properties autonomy Autonomy to make the tester without too much attention to the details of the test, testers just need to select the test objectives, test control Agent information can be based on the knowledge base, automatically generate ECA rules table, and run the script, according to ECA rules to capture sensitive event and perform the appropriate action. intelligence Test Control Agent to the test objectives based on user selection, automatically generates ECA rules table, and according to ECA rules, reasoning and take appropriate action. Test Control Agent can exception handling rules based on the testing process to deal with abnormal body fat to protect the testing process automatically and smoothly. 3.5 Agent Network Network Agent implements the network environment simulation and control. In order to test the phone's RF performance, usually the base station simulator to provide a device in which the wireless network environment, and use programming interface, the software environment to control the device, making the complex network environment, the target device under test as possible. 4 System Evaluation In order to evaluate the effectiveness of the system, we designed a comparative experiment. Tell testing laboratory in the Ministry of Information Industry of the test project for real. We will test engineers divided into two groups, the first group to use this system MobileTest test, the second group uses the industry's leading TestQuest Pro test. Test is based on the number of China's telecommunications industry standard part of mobile station testing. Test content is based on both systems return to their already established on the new test cases for regression testing of smart phones to compare these two test tools, test efficiency and maintainability of the script. Table 1 regression test results:  Two experimental tools to measure different types of test cases in the test, the automated i