Innovative Software and AI: “Invented for life”
In our modern world, software is at the heart of practically everything we use daily. It enables us to connect, automate tasks, and create powerful devices, tools, and services that help us achieve great things. That’s why at Bosch Research, we tackle the challenges head on and make software and artificial intelligence engineering faster, more efficient, and better than ever.
Bosch’s mission is “Invented for life” — and software and artificial intelligence (AI) have long been an integral part of this mission. In fact, since 2022, all Bosch's product classes now contain software or AI, whether to control an electric motor, predict maintenance intervals for industrial equipment, or ensure the quiet and comfort of your home. The rapid advance of software and artificial intelligence inspires us as Bosch Research to push the boundaries of what is possible with both.
As the importance of software has grown, the way software is designed, built, and evolved has been changing over the years. Gone are the days of static, embedded software that was ready at the end of the factory line and never changed afterward. Today, all software is constantly being updated, improved, and deployed to devices in the field. Thanks to software updates, devices’ usability and feature set improve throughout their lifecycle.
Ahead of time: Adapting software and new programming languages for development
One area we are particularly interested in is software that adapts to the field. In other words, software that is not only smart, but also responds to changing conditions in real time, for example, changing weather patterns, unexpected workloads, or even normal wear and tear. We explore modern control theory approaches, such as online, reinforcement, and federated machine learning, and make them accessible to Bosch's business units. This way, as vehicles such as excavators deteriorate in the field, they can adapt without having this behavior explicitly modeled, and perform complex maneuvers as if they were brand new. We also explore how automation helps us efficiently safeguard, verify, and validate these self-adaptive systems.
Another area of focus is automating development and certification processes for software, i.e., tools and techniques to streamline software development, testing, preparing certifications, deployment, and operations. As frequent deployment into the field becomes the norm, development cycles need to reduce the time, cost, and number of specialists necessary to bring a product to market. We at Bosch Research also explore how to ensure the security and safety of these quickly changing cyber-physical systems in an automated way. This means, for example, that instead of manually rewriting the documents with every change in the software, the required documentation for safety and security certification can be updated automatically with every release.
We also explore new programming and modeling languages. Current programming languages are often hard to use, and it’s easy to make mistakes. Instead, we explore custom, domain-specific modeling and programming languages (e.g., Blech) that make software development faster, more secure, and more efficient. They will help Bosch business units reduce time and cost in evolving their software while preserving Bosch's high quality and safety standards.
Faster, higher, further: Innovations in software design and the crucial role of AI
Code quality and development velocity are also essential areas of our applied research. For example, suitable software abstractions allow for hardware-agnostic deployments, for code that's easier to understand and modify, and for engineers who focus on the higher-level design of the software rather than the implementation details. In addition, innovation accelerates both the software development process and its implementation.
But the most exciting development in recent years has been the rise of artificial intelligence (AI). AI has the potential to completely change the way we add behavior and intelligence to devices, systems, and services. Instead of programming them explicitly, AI allows us to train systems to perform tasks. This opens up new possibilities for creating smarter, more adaptive products quicker – and adapting and improving them over time. Most recently, developments in this field have seen foundation models revolutionize the way we use AI. Examples include combining domain knowledge with generic foundation models to build querying systems on our own data. We also use these in our software development and tackle some of the most challenging questions in this area:
- How do we efficiently collect, manage, and prepare high-quality data to train AI systems?
- How can we train AI systems with limited data when large amounts of data are traditionally required?
- How can we bring AI into both hardware and software engineering – in a reliable, safe, secure, effective, and transparent way?
- How do we create explainable and transparent AI systems in safety-critical industries like automotive, home, and industrial processing?
Last but not least, we at Bosch Research bridge academia and practical applications. We publish at top-tier conferences and consistently look for collaborations across academia, associations, and potential future business partners that share our mission to invent applications and products for life.