乘着创新的浪潮: 协同运输机器人编队的基于模型的开发实施方案(Riding the wave of innovation: model-based development and implementation scheme of cooperative transportation robot formation)

作为由德国联邦教育和研究部(BMBF)资助的研究项目,协同嵌入式系统(CrESt)专注于复杂的嵌入式系统的开发,这些系统须在不同背景下与不同成员有效合作以完成给定任务。该项目的目标是使用基于模型的系统和系统上下文描述,为动态和可扩展的应用定义方法和架构。协同嵌入式系统(CES)和适应性系统架构将在不久的将来对技术发展产生重大影响。通过这些系统及其架构,目前的工厂流程可以重新规划并从高度自动化中获益。它们还可以帮助工厂以更灵活的方式对不断变化的生产条件做出反应。例如,学习型机器可以接管以前由人类完成的重复性任务,以目标导向的方式优化这些流程。此外,协同嵌入式系统可以通过控制中央订单管理系统,自主协调和组织运输订单。以基于模型的系统设计为中心,以安全协作为重点,此系统可以运行分析,记录系统将会需要的所有功能。这种技术可应用于自主机器人、学习控制系统和适应性工厂等。

基于模型的协同嵌入式系统开发

Systems Automatio公司和MES模赛思是22个来自工业和科学背景的国际合作者的其中两个,两个公司在联合研究项目中合作探寻新技术和新方法,共同开发CES以及协同系统组(CSG),如自主机器人编队 。协同系统重点关注系统中的有效合作,避免相互竞争。CrESt项目分为六个”工程挑战 “和六个跨学科的主题。前三个工程挑战涉及的首要问题是如何最好地设计灵活、动态和具有适应性的系统架构 。在动态架构项目方面,InSystems Automation和MES模赛思的主要任务是研究各种利用即插即用等机制将机器人整合到已有队伍中的可能性,而无需暂停整个生产过程。此外,InSystems的目标是改进他们的proAnt运输机器人,这些机器人自2012年以来投入至生产中。总的来说,项目重点在于技术的改进和创造和评估新方法来实现其整体概念。更确切地说,关键议题包括分散的车队管理、各个系统之间的通信和环境数据分析。

由于所涉及的情景高度复杂,导致了许多困难,开发自适应系统需要有良好基础的方法。机器人队伍必须对制造执行系统的要求还有其成员的数量和性质的动态变化作出反应,以确保CSG的整体功能和效率得到保障。基于模型的自动化系统开发过程的一致应用提供了各种有益的特性。最重要的是,CSG可执行模型形式的规范和协同AGV控制器(cac)允许机器人编队成员及其协同自适应行为的完全虚拟呈现(数字孪生)。这种虚拟呈现为有效开发、维护和扩展实际系统及其硬件、软件和机械组件提供了良好的基础。为充分挖掘其潜力,基于模型的方法首先依赖于模型和测试平台在不同开发阶段的可重用性,包括功能、系统和系统组件的开发。其次,基于模型的开发过程建立在一个完全集成的工具链上,相关的开发活动高度自动化,包括需求管理、建模和仿真,以及集成的质量保证任务,特别是基于模型的静态分析和基于需求的测试工具。

运输机器人编队的自适应系统架构

在自适应嵌入式系统的设计和维护中,对系统环境的考量几乎 是所有建模和分析方法的一个关键方面。在协同运输机器人的用例中,工作分配程序是协同的最关键部分。因此,为了能够在工作任务出现时确定负责执行的机器人,需要确定和应用一个或一组策略。换句话说,CSG必须能够完成工厂要求的所有工作任务,同时使机器人的行为适应环境的变化。在这种情况下,项目面临如下挑战: CSG要如何分配工作给机器人,使它们既能实现局部目标,也能达成全局目标。前一目标专门针对单个机器人,如最小的电池充电状态,后一目标则是由制造执行系统给出的合适的生产策略所决定。协同运输机器人的生产策略决定了CSG的共同目标,该目标由制造执行系统传达给机器人编队。

在与InSystems的合作中,MES模赛思的软件工程师研究并评估了由制造执行系统动态传达的四个具体全局目标,即经济性(使所有CAC的总行驶距离最小化)、稳健性(使每个机器人的作业队列长度尽可能小)、性能(使每个时间单位的作业执行数量最大化)和维护性(分配任务,使所有机器人的行驶距离相似),并使用投标参数向量对这些全局目标进行编码。需要注意的是,因为在设计之初没有给出预设信息,这些目标可能随系统运行而发生变化。动态变化的目标必须作为具体的CSG策略来实现,这些策略将由机器人编队自主解决。其他全局性的CSG目标,包括对工作分配的时间限制以及局部目标,也需要被考虑在内。环境中的任何动态变化都可能触发一个重新配置阶段,比如当新的工作任务被制造执行系统广泛采用,当机器人加入或离开车队,或者当检测到一个新的障碍物,在这些情况下,CSG均需要重新配置,以便符合新的CSG目标。例如,根据给定的策略,如果机器人在没有完成其任务列表的情况下离开车队,那么每个 CAC 处理的作业队列都必须进行调整甚至重新分配。为了完成像这样的任务,需要重新配置单元。

在Simulink中建立运输机器人编队模型

InSystems和MES模赛思共同开发了一个MATLAB/Simulink模型,该模型是一个由协同运输机器人组成的自适应机队,如InSystems公司开发的proANT机器人。该模型旨在捕捉所需的自适应系统行为,并更有效地处理上述CSG目标和挑战。独立于领域的半形式化语言Simulink适合描述机器人机队及其成员的CSG/CAC行为,以及包括制造执行系统在内的环境。此外,Simulink模型能够与典型的机器人中间件或通信框架(如机器人操作系统(ROS))连接。

Simulink提供了一个平台来设计、模拟和验证各种抽象层次的(动态)系统的行为,包括功能/规格、系统和软件模型。该软件在工业界广泛使用,因为它为动态系统提供了一个特定域无关的建模工具。典型的应用包括信号处理、控制工程问题和系统工程。特别是,Simulink提供了几种仿真模式,从准连续到离散或基于事件的执行和采样率,各种求解器选项,允许在精度、内存消耗和执行时间之间进行定制的权衡,它支持常见的数据类型概念,包括浮点、定点和枚举类型。Simulink的内置库和各种插件,包括有限状态机、特定领域模型集成和FMI支持,非常适合作为快速创建、仿真和测试CSG原型的工具软件。此外,它在不同开发阶段有很大潜力可以重复使用,非常有利于开发。

在基于模型的开发过程中,为了确保功能、可维护性以及整体有效的工作流程,遵循建模规范也至关重要。例如,系统分解模型必须只包含子系统块和信号路由元素。特别是,在分解模型中不进行数值计算。这保证了CSG的需求可以完全映射到每个机器人的CAC组件上的需求 。此外,分解模型必须指定复杂度受限的组件,对于这些组件需采取合适的措施。建模规范进一步解决了许多方面的问题,包括安全主题、变体控制和强数据类型。在这个项目过程中,静态分析工具(如MES Model Examiner)负责自动检查和纠正合规性。

从形式化的需求到使用MES Test Manager/MARS进行自动评估

为了开发一种实用的方法来实现、操作和验证CSG的自适应行为,MES模赛思将重点放在了基于Simulink的CSG原型上。前文所提到的系统要求必须使用测试驱动的方法进行验证,并在运行时进行监控指示可能的系统故障,以便采取适当的对策。策略更改会导致协议更改,这需要在系统规范中得到适当处理。最重要的是,在制造执行系统动态变化的策略影响下,预期的适应性系统响应必须充分体现在CSG需求中。一致的系统规范必须是明确的、统一的和易懂的,因此形式化需求尤为重要。

与实践中广泛使用的基于自然语言的方法相比,形式化的需求格式产生了对CSG需求的明确表示。此外,形式化的需求格式,如Test Manager可评估需求语法(MARS),可以与基于模型的方法完全整合,即基于状态或事件的触发器和所需的信号响应可以通过引用模型实体(如信号规格或设计参数)来完全定义。与适当的测试用例的有效定义相结合,自适应CSG行为的虚拟验证可以在自动测试执行和评估的基础上实现自动化。特别是,MARS弥合了基于形式化语言的需求和不同背景的系统工程师使用的可以轻松制定和处理的自然语言需求 之间的差距。作为模型测试工具MES Test Manager的一部分,MARS进一步为基于模型的开发中测试自动化的各个方面提供了基础,包括自动生成评估、测试用例和测试过程监控。

总体而言,上述CSG的开发和维护可以基于完全虚拟的对象,该对象以交互的Simulink模型的形式表现运输机器人编队的行为。因此,质量保证方法,例如基于需求的测试方法,可以基于机器人编队的完全虚拟原型,使用MES Test Manager等工具,使质量保证任务高度自动化。这种”前置”方法有助于在开发过程早期发现设计和实施错误。此外,它有助于快速测试和验证新机器人类型的集成或新协同协议的实施。

关于MES模赛思: 软件质量尽在控制之中

模赛思软件技术有限公司(Model Engineering Solutions),简称MES)是一家来自德国柏林的高科技软件公司,专为软件项目的质量保障提供解决方案。

MES为客户基于模型的软件开发提供技术支持,使其符合IEC 61508、ISO 26262和ASPICE等行业标准。MES模赛思成立于2006年,总部位于德国柏林。Hartmut Pohlheim博士作为基于模型的开发领域最著名的专家之一,自2008年起任公司常务董事。MES的主要客户包括整车厂如戴姆勒、大众、丰田和吉利等以及博世、西门子和三星等行业供应商。在汽车行业中,除少数几家公司外,全球数十家顶尖制造商及供应商均在他们的开发环境中使用MES的解决方案。为支持其全球客户,MES已在美国和中国建立了子公司,并与全球分销商网络紧密合作。

MES的产品包括4种质量工具软件:MXAM、MES Test Manager、MoRe和MQC,它们共同构成了一个工具链,全面保障基于模型的软件开发过程中所有阶段的质量。通过MES Jenkins Plugin,该工具链也可以在持续集成环境中使用。工具链主要应用平台为MATLAB/Simulink。除了MES质量工具外,MES测试中心和MES学院的专家们还为全球客户提供关于质量保证和开发流程优化的定制咨询服务及培训课程。

2Bierbaum
Unternehmensgruppe
proANT AGVs车队,运输敞口桶

图2:Bierbaum
Unternehmensgruppe的proANT AGVs车队,运输敞口桶

在自适应嵌入式系统的设计和维护中,对系统环境的考量几乎
是所有建模和分析方法的一个关键方面。在协同运输机器人的用例中,工作分配程序是协同的最关键部分。因此,为了能够在工作任务出现时确定负责执行的机器人,需要确定和应用一个或一组策略。换句话说,CSG必须能够完成工厂要求的所有工作任务,同时使机器人的行为适应环境的变化。在这种情况下,项目面临如下挑战:
CSG要如何分配工作给机器人,使它们既能实现局部目标,也能达成全局目标。前一目标专门针对单个机器人,如最小的电池充电状态,后一目标则是由制造执行系统给出的合适的生产策略所决定。协同运输机器人的生产策略决定了CSG的共同目标,该目标由制造执行系统传达给机器人编队。

在与InSystems的合作中,MES模赛思的软件工程师研究并评估了由制造执行系统动态传达的四个具体全局目标,即经济性(使所有CAC的总行驶距离最小化)、稳健性(使每个机器人的作业队列长度尽可能小)、性能(使每个时间单位的作业执行数量最大化)和维护性(分配任务,使所有机器人的行驶距离相似),并使用投标参数向量对这些全局目标进行编码。需要注意的是,因为在设计之初没有给出预设信息,这些目标可能随系统运行而发生变化。动态变化的目标必须作为具体的CSG策略来实现,这些策略将由机器人编队自主解决。其他全局性的CSG目标,包括对工作分配的时间限制以及局部目标,也需要被考虑在内。环境中的任何动态变化都可能触发一个重新配置阶段,比如当新的工作任务被制造执行系统广泛采用,当机器人加入或离开车队,或者当检测到一个新的障碍物,在这些情况下,CSG均需要重新配置,以便符合新的CSG目标。例如,根据给定的策略,如果机器人在没有完成其任务列表的情况下离开车队,那么每个 CAC 处理的作业队列都必须进行调整甚至重新分配。为了完成像这样的任务,需要重新配置单元。

————————

As a research project funded by the German Federal Ministry of education and Research (BMBF), collaborative embedded system (CREST) focuses on the development of complex embedded systems. These systems must effectively cooperate with different members in different contexts to complete a given task. The goal of the project is to define methods and architectures for dynamic and scalable applications using model-based system and system context description. Collaborative embedded system (CES) and adaptive system architecture will have a significant impact on the development of technology in the near future. Through these systems and their architectures, current plant processes can be re planned and benefit from a high degree of automation. They can also help factories respond to changing production conditions in a more flexible way. For example, learning machines can take over repetitive tasks previously completed by humans and optimize these processes in a goal-oriented manner. In addition, the collaborative embedded system can independently coordinate and organize transportation orders by controlling the central order management system. Centered on model-based system design and focusing on security cooperation, this system can run analysis and record all functions that the system will need. This technology can be applied to autonomous robots, learning control systems and adaptive factories.

Model based collaborative embedded system development

Systems automation and MES are two of 22 international collaborators from industrial and scientific backgrounds. The two companies cooperate to explore new technologies and methods in joint research projects, and jointly develop ces and collaborative system group (CSG), such as autonomous robot formation. Collaborative system focuses on effective cooperation in the system to avoid mutual competition. The crisis project is divided into six “engineering challenges” and six interdisciplinary themes. The first three engineering challenges involve how to best design a flexible, dynamic and adaptive system architecture. In terms of dynamic architecture projects, the main task of insystems automation and MES is to study the possibility of integrating robots into existing teams by using plug and play mechanisms without suspending the whole production process. In addition, insystems aims to improve their proant transport robots, which have been put into production since 2012. In general, the project focuses on the improvement of technology and the creation and evaluation of new methods to realize its overall concept. Rather, key issues include decentralized fleet management, communication between systems, and environmental data analysis.

Because the scenarios involved are highly complex, which leads to many difficulties, the development of adaptive system needs a good basic method. The robot team must respond to the requirements of the manufacturing execution system and the dynamic changes in the number and nature of its members, so as to ensure the overall function and efficiency of CSG. The consistent application of model-based automation system development process provides a variety of useful features. Most importantly, the specification in the form of CSG executable model and the cooperative AGV controller (CAC) allow the complete virtual presentation (Digital twinning) of robot formation members and their cooperative adaptive behavior. This virtual presentation provides a good foundation for the effective development, maintenance and expansion of the actual system and its hardware, software and mechanical components. In order to fully tap its potential, the model-based method first depends on the reusability of the model and test platform in different development stages, including the development of functions, systems and system components. Secondly, the model-based development process is based on a fully integrated tool chain, and the relevant development activities are highly automated, including requirements management, modeling and simulation, as well as integrated quality assurance tasks, especially model-based static analysis and requirements based testing tools.

Adaptive system architecture of transportation robot formation

In the design and maintenance of adaptive embedded systems, the consideration of system environment is almost a key aspect of all modeling and analysis methods. In the use case of cooperative transportation robot, work assignment program is the most key part of cooperation. Therefore, in order to determine the robot responsible for execution when a work task occurs, one or a set of strategies need to be determined and applied. In other words, CSG must be able to complete all work tasks required by the factory and adapt the behavior of the robot to the changes of the environment. In this case, the project faces the following challenges: how CSG assigns work to robots so that they can achieve both local and global goals. The former goal is specific to a single robot, such as the minimum battery state of charge, while the latter goal is determined by the appropriate production strategy given by the manufacturing execution system. The production strategy of cooperative transportation robot determines the common goal of CSG, which is communicated to robot formation by manufacturing execution system.

In cooperation with insystems, MES software engineers studied and evaluated four specific global objectives dynamically conveyed by the manufacturing execution system, namely, economy (minimizing the total travel distance of all CACS), robustness (minimizing the job queue length of each robot as much as possible), performance (maximizing the number of job executions per time unit) and maintainability (assign tasks so that all robots travel similar distances) , and use the bidding parameter vector to encode these global objectives. It should be noted that because there is no preset information at the beginning of the design, these objectives may change with the operation of the system. The dynamically changing objectives must be realized as specific CSG strategies, which will be solved by the robot formation independently. Other global CSG objectives include work assignment The time limit and local objectives also need to be taken into account. Any dynamic change in the environment may trigger a reconfiguration stage. For example, when a new work task is widely adopted by the manufacturing execution system, when the robot joins or leaves the fleet, or when a new obstacle is detected, the CSG needs to be reconfigured to meet the new CSG objectives For example, according to a given strategy, if the robot leaves the fleet without completing its task list, the job queue processed by each CAC must be adjusted or even reassigned. In order to complete tasks like this, the unit needs to be reconfigured.

The formation model of transportation robot is established in Simulink

Insystems and MES have jointly developed a Matlab / Simulink model, which is an adaptive fleet composed of cooperative transportation robots, such as proant robot developed by insystems. The model aims to capture the required adaptive system behavior and deal with the above CSG objectives and challenges more effectively. The domain independent semi formal language Simulink is suitable for describing the CSG / CAC behavior of robot fleet and its members, as well as the environment including manufacturing execution system. In addition, the Simulink model can be connected with typical robot middleware or communication framework, such as robot operating system (ROS).

Simulink provides a platform to design, simulate and verify the behavior of various abstract levels of (dynamic) systems, including function / specification, system and software model. The software is widely used in industry because it provides a domain independent modeling tool for dynamic systems. Typical applications include signal processing, control engineering problems and system engineering. In particular, Simulink provides several simulation modes, from quasi continuous to discrete or event based execution and sampling rate, and various solver options, allowing customized trade-offs between precision, memory consumption and execution time. It supports common data type concepts, including floating-point, fixed-point and enumeration types. Simulink’s built-in libraries and various plug-ins, including finite state machines, domain specific model integration and FMI support, are very suitable as tool software for rapid creation, simulation and testing of CSG prototypes. In addition, it has great potential in different development stages and can be reused, which is very conducive to development.

In the process of model-based development, in order to ensure function, maintainability and overall effective workflow, it is also very important to follow the modeling specification. For example, the system decomposition model must contain only subsystem blocks and signal routing elements. In particular, numerical calculations are not performed in the decomposition model. This ensures that the requirements of CSG can be fully mapped to the requirements of CAC components of each robot. In addition, the decomposition model must specify the components with limited complexity, and appropriate measures shall be taken for these components. The modeling specification further solves many problems, including security topics, variation control and strong data types. During this project, static analysis tools (such as MES model examiner) are responsible for automatically checking and correcting compliance.

从形式化的需求到使用MES Test Manager/MARS进行自动评估

In order to develop a practical method to realize, operate and verify the adaptive behavior of CSG, MES Marcus focuses on the CSG prototype based on Simulink. The system requirements mentioned above must be verified by test driven method, and monitored during operation to indicate possible system faults, so as to take appropriate countermeasures. Policy changes will lead to protocol changes, which need to be properly handled in the system specification. Most importantly, under the influence of the dynamic change strategy of manufacturing execution system, the expected adaptive system response must be fully reflected in CSG requirements. Consistent system specifications must be clear, unified and easy to understand, so formal requirements are particularly important.

Compared with the natural language based method widely used in practice, the formal requirement format produces a clear representation of CSG requirements. In addition, formal requirements formats, such as test manager evaluable requirements syntax (MARS), can be fully integrated with model-based methods, that is, state or event based triggers and required signal responses can be fully defined by referencing model entities (such as signal specifications or design parameters). Combined with the effective definition of appropriate test cases, the virtual verification of adaptive CSG behavior can be automated on the basis of automatic test execution and evaluation. In particular, Mars bridges the gap between formal language based requirements and natural language requirements that can be easily formulated and processed by system engineers from different backgrounds. As a part of the model testing tool MES test manager, Mars further provides the basis for all aspects of test automation in model-based development, including automatic generation and evaluation, test cases and test process monitoring.

In general, the development and maintenance of the above CSG can be based on a completely virtual object, which represents the behavior of transportation robot formation in the form of interactive Simulink model. Therefore, quality assurance methods, such as demand-based testing methods, can be based on the full virtual prototype of robot formation and use tools such as MES test manager to highly automate quality assurance tasks. This “front-end” approach helps to identify design and implementation errors early in the development process. In addition, it helps to quickly test and verify the integration of new robot types or the implementation of new collaborative protocols.

About MES: software quality is under control

Model engineering solutions (MES) is a high-tech software company from Berlin, Germany, which provides solutions for the quality assurance of software projects.

MES provides technical support for customer model-based software development to make it comply with industry standards such as IEC 61508, ISO 26262 and aspire. MES was founded in 2006 and headquartered in Berlin, Germany. As one of the most famous experts in the field of model-based development, Dr. Hartmut Pohlheim has been the managing director of the company since 2008. The main customers of MES include vehicle manufacturers such as Daimler, Volkswagen, Toyota and Geely, as well as industry suppliers such as Bosch, Siemens and Samsung. In the automotive industry, except for a few companies, dozens of top manufacturers and suppliers around the world use MES solutions in their development environment. To support its global customers, MES has established subsidiaries in the United States and China and works closely with the global distributor network.

MES products include four quality tool software: mxam, MES test manager, moreand MQC, which together form a tool chain to comprehensively ensure the quality of all stages in the process of model-based software development. Through MES Jenkins plugin, the tool chain can also be used in the continuous integration environment. The main application platform of tool chain is Matlab / Simulink. In addition to MES quality tools, experts from MES test center and MES college also provide customized consulting services and training courses on quality assurance and development process optimization for customers all over the world.

2Bierbaum
Unternehmensgruppe
proANT AGVs车队,运输敞口桶

图2:Bierbaum
Unternehmensgruppe的proANT AGVs车队,运输敞口桶

In the design and maintenance of adaptive embedded system, the consideration of system environment is almost
Is a key aspect of all modeling and analysis methods. In the use case of cooperative transportation robot, work assignment program is the most key part of cooperation. Therefore, in order to determine the robot responsible for execution when a work task occurs, one or a set of strategies need to be determined and applied. In other words, CSG must be able to complete all work tasks required by the factory and adapt the behavior of the robot to the changes of the environment. In this case, the project faces the following challenges:
How should CSG assign work to robots so that they can achieve both local and global goals. The former goal is specific to a single robot, such as the minimum battery state of charge, while the latter goal is determined by the appropriate production strategy given by the manufacturing execution system. The production strategy of cooperative transportation robot determines the common goal of CSG, which is communicated to robot formation by manufacturing execution system.

In cooperation with insystems, MES software engineers studied and evaluated four specific global objectives dynamically conveyed by the manufacturing execution system, namely, economy (minimizing the total travel distance of all CACS), robustness (minimizing the job queue length of each robot as much as possible), performance (maximizing the number of job executions per time unit) and maintainability (assign tasks so that all robots travel similar distances) , and use the bidding parameter vector to encode these global objectives. It should be noted that because there is no preset information at the beginning of the design, these objectives may change with the operation of the system. The dynamically changing objectives must be realized as specific CSG strategies, which will be solved by the robot formation independently. Other global CSG objectives include work assignment The time limit and local objectives also need to be taken into account. Any dynamic change in the environment may trigger a reconfiguration stage. For example, when a new work task is widely adopted by the manufacturing execution system, when the robot joins or leaves the fleet, or when a new obstacle is detected, the CSG needs to be reconfigured to meet the new CSG objectives For example, according to a given strategy, if the robot leaves the fleet without completing its task list, the job queue processed by each CAC must be adjusted or even reassigned. In order to complete tasks like this, the unit needs to be reconfigured.