The use of software-driven product development is on the rise in the manufacturing industry. Within that software-driven development, a great deal of attention is being turned to Model-Based Development (MBD) as a methodology for transforming design, development, and testing processes. Model-Based Development is being fused with simulation technologies and is gradually being used throughout the manufacturing industry. With the rise of next-generation mobility, such as autonomous vehicles, the automobile industry is entering a period of great transformation and is making extensive use of Model-Based Development. Toshiba Digital Solutions has made it possible to further increase the efficiency of overall product development by using Model-Based Development for individual components, connecting different models and development tools through cyberspace. This enables collaborative simulation by a large number of systems. In this issue, we will introduce VenetDCP, a Distributed Co-simulation Platform for building integrated simulation environments.

In recent years, product development in the manufacturing industry has been shifting from traditional hardware-driven development to software-driven development. Well-known examples include the next-generation mobility efforts of the automobile industry, such as the development of electric vehicles (EVs), software-defined vehicles (SDVs), and autonomous vehicles. Behind this lie the demands being placed on the manufacturing industry as a whole to tackle various social issues, from micro- to macro-sale issues. These include improving the safety of industrial products, digital transformation (DX), promoting social and economic transformations such as the Fourth Industrial Revolution, reducing CO2 emissions, and addressing related energy problems.

There is an ongoing, rapid shift toward using software in next-generation mobility, and the architecture of electronic control units (ECUs) is changing. As the number of ECUs used in automobiles rises, software has come to account for an even larger percentage of total automobile development costs. In 2020, the number of lines of source code per automobile was 200 million. By 2025, that number is forecast to rise to 600 million. This is an extremely large amount of code, even compared to software in other fields.*

* Reference: Ministry of Economy, Trade and Industry (https://www.meti.go.jp/shingikai/economy/daiyoji_sangyo_skill/pdf/005_04_00.pdf (PDF)(3.99MB))

Major changes must also be made to the processes used in next-generation automotive development, which involves this massive scale of software development. Traditionally, automotive development has been led by the manufacturers of engines, transmissions, and other hardware. ECUs, sensors, and the like have been positioned as accessories, and software development has been based on physical equipment and performed using manpower. Back when vehicle control was simple, this was enough. However, automotive development has changed due to advances in engine control, automatic brakes, Advanced Driver Assistance Systems (ADAS), connected functions, and the like. Now, hardware must be designed with the assumption that it will be controlled by complex software. Modern automotive development must be performed with an eye towards entire automotive systems. Using conventional methods of software development and then rolling back from the testing and evaluation stages to the design and specification stages has a significant impact on vehicle costs and development schedules.

A great deal of attention is being turned to the use of simulation technologies and Model-Based Development (MBD) to deal with these issues. Model-Based Development itself is widely used throughout the industrial world, primarily for software. It is transforming design, development, and testing processing through its use of using object-oriented approaches and Unified Modeling Language (UML) models.

In Model-Based Development, formulas and the like are used to define models of phenomena and behavior. Development is performed by using these models in simulations. The feasibility of the specifications of these models is thoroughly evaluated through multiple simulations, and software (source code) is automatically generated based on these well-vetted models. Conventional development methodologies encompass various stages, including capability planning, system design, component prototyping, and software development. Physical prototypes are then assembled and used to perform verification. Unlike conventional development, which emphasizes downstream processes in which evaluation is performed, Model-Based Development emphasizes up-stream processes, carrying out verification in advance, during the design phase. This makes it possible to shorten development times and curb development costs (Fig. 1).

That said, simply introducing Model-Based Development is not enough to resolve all development problems. Needless to say, using simulations during the design stage makes it possible to confirm functions and performance, so it helps make development more efficient. However, in the manufacturing industry, such as the automobile industry, supply chains are almost always multi-level. Even if Model-Based Development improves the efficiency of development for individual component manufacturers, such as the manufacturers of engines, brakes, or steering systems, it still takes multiple rounds of prototyping and evaluation to perform evaluation on physical vehicles by assembling all of those components into prototypes. This process makes overall automobile development both time-consuming and costly.

Toshiba Digital Solutions is working to improve the efficiency of the entire development process by modeling everything, even the final physical prototypes of product manufacturers, so that evaluation can be performed in advance. Models are collected from each component manufacturer (model distribution), and multiple models are linked to perform simulations. This is known as "co-simulation" or "collaborative simulation." (We call it "co-simulation.")

In physical sites, the connecting of individual components and the assembly of systems usually involves manual work performed by engineers. Like fitting and embedding, the models used in co-simulation must be converted into a tool-usable form, as necessary, and arranged.

One of the difficulties involved in performing co-simulation is linking models from different companies. This is because simulators and their versions vary by company and organization, so models cannot all be handled in the same way. This makes it vital to use technologies that accurately connect the inputs and outputs of each model, without error. We do not directly connect the simulators of each company, but instead connect them indirectly (using a distributed approach). This makes it easy to link the data of the different models.

To indirectly connect models, we use mechanisms called "bus connectors." Bus connectors function as interfaces for the inputs and outputs of each simulator. They match the differing formats of simulators (converting them), and are automatically generated based on the input and output specifications defined by the company that manages the co-simulation. This makes it possible to perform simulations over network connections to other companies' models simply by having each company embed the provided bus connectors in their simulators.

Currently, bus connectors support both Simulink's S-Function format, which is the standard for Model-Based Development tools, and the FMI format, an international standard. Furthermore, we provide an application programming interface (API) to support proprietary models.

FMI, or Functional Mock-Up Interface, is an international standard that was formulated as a shared interface standard for model distribution. Many simulation tools offer FMI support. Our bus connectors support both FMI2.0 and FMI3.0, so they can connect to a large number of simulation tools. The API can be used to connect to special-purpose hardware, Hardware In the Loop Simulations (HILS), and the like.

We provide the VenetDCP Distributed Co-simulation Platform as an environment for joint digital prototyping through the use of simulations that indirectly integrate multiple (distributed) models (co-simulation) (Fig. 2).

With VenetDCP, companies can perform integrated simulations with other companies via a network. However, each company's models may contain various confidential information, such as detailed product data or know-how. Confidentiality and security are therefore pressing issues in these simulations. That is why we developed a mechanism that does not require valuable models to be handed over to other companies. When performing co-simulation, the only data that is provided to other companies are the results of model calculations, which are shared by the bus connectors. Each company can therefore feel secure taking part in co-simulation, and the platform makes them more willing to collaborate. We believe this is important to maintaining smooth collaborative relationships with clients and partners.

Connections and simulation configurations can be set up in advance, so each company can call up and execute its models when it wishes. During the simulation process, there is no need for the companies supplying models to stand by while a physical prototype is tested. In this way, the platform is designed to provide users with even greater ease of use. It also takes security into consideration. The platform has mechanisms that allow model providers to apply limits to the calling up of models. This prevents other companies from being able to perform processing that the model providers would rather avoid.

It is becoming increasingly common for co-simulation to be performed at remote sites, using different tools (Fig. 3).

VenetDCP is already being used by major automakers in Japan. Distributed Co-simulation environments have been praised for the benefits they offer to both suppliers and automakers in the automotive engineer chain. They are being effectively utilized by performing actual simulations at domestic and overseas sites and by integrating multiple simulations.

Model-Based Development has drawn a great deal of attention as a major future development scheme, and Toshiba expects it to see widespread deployment throughout the industrial world in the future. We are confident that VenetDCP can make development even more efficient for companies that use Model-Based Development. Even within companies, it can assist with simulations that involve multiple sites or different tools. Of course, in the Japanese manufacturing industry, which has complex supply chains, it is also extremely effective for performing coordinated product development that spans multiple companies or organizations. We are also coordinating with partners in the U.S. and China using VenetDCP, and recently we have been receiving inquiries from a growing number of companies involved in social infrastructure, such as energy companies, about the use of this platform. VenetDCP continues to evolve.

Various organizations have been established in Japan and overseas to promote the use of model distribution. As a company that promotes both Model-Based Development and model distribution, Toshiba is taking part in activities by groups in Japan and Europe, serving as a bridge between the two. We are contributing to the widespread use of model distribution on a global scale.

VenetDCP is a global pioneer, a platform that makes it possible to build remote, integrated simulation environments without sharing models with partners. Toshiba Digital Solutions will continue to support the digital transformation of the manufacturing industry and contribute to its competitiveness through distributed co-simulation that helps solve the manufacturing problems that companies face.

• Distributed Co-Simulation Platform VenetDCP