Software for sensors has evolved from simply reading out and evaluating sensor data to making intelligent decisions based on that data, a transformation enabled by new software synthesis and artificial intelligence (AI) technologies. Together, they make consumer devices smarter, dramatically improving the user experience through greater interactivity and higher levels of automated personalization.
SEMI’s Nishita Rao spoke with Stefan Finkbeiner, CEO and General Manager at Bosch Sensortec, who will explore the topic in his October 23 keynote, How Software Makes MEMS Sensors into Smart Systems, at MEMS & Sensors Executive Congress (MSEC), October 22-24, 2019, at the Coronado Island Marriott Resort & Spa in Coronado, Calif.
Join us at MSEC to meet Bosch Sensortec and other industry influencers driving MEMS and sensors innovations. Registration is open.
SEMI: What is the relationship between MEMS sensors suppliers and specialized software synthesis providers?
Finkbeiner: Collaboration is a key driver for innovation in sensor software. There are already several fruitful collaborations between MEMS sensors suppliers and specialized software providers, which are mostly startups. Collaborations with providers of simulation and evaluation tools as well as with well-known universities in the field of AI are starting to show positive results.
Domain expertise is also critical for developing smart sensor software, making it essential to future sensing solutions.
SEMI: How does software synthesis relate to sensor fusion?
Finkbeiner: Put simply, software synthesis refers to ways of automatically generating code based on domain knowledge and given constraints for specific product versions. Sensor fusion combines sensor data from different kinds of sources in order to improve the results.
Software synthesis techniques enable a level of automation that creates new opportunities for more complex sensor fusion, which was formerly out of reach when using traditional approaches that involved, for example, big data and a large number of potential data sources.
The traditional sensor fusion toolset can now be further extended by machine learning techniques that help to determine which sources are more reliable than others and how to combine data streams. This topic and others are still active areas of research. A wearable device with motion detection is a case in point. With unsupervised learning, the device could identify short versus long cyclically repeating motions and treat them differently from other types of motion.
SEMI: How is the new software synthesis-AI approach different from previous approaches? To what degree will the new approach open up new applications?
Finkbeiner: Traditionally, technology companies have used cloud computing for data storage and machine learning on aggregated user data. In that model, MEMS sensors generate large amounts of data that power-hungry hardware (such as digital signal processors) must process. In addition, machine learning generally requires lots of power-hungry cloud nodes with GPUs. This model, however, is not the best option for many users. Just think for a moment about all the scenarios in battery-powered devices where frequent battery charging frustrates users.
Leveraging both software synthesis and AI techniques in MEMS sensors is therefore a very promising approach because it supports improved recognition and learning inside the sensor. This means that user-specific data isn’t transferred to the cloud. Instead, it remains private inside the sensor. This improves existing applications that learn all the time and opens up new opportunities for applications such as smart clothing, predicting a product’s lifespan, detecting whether a window or door is open or closed – all without server connectivity.
SEMI: How will such software adapt to the individual user?
Finkbeiner: Devices will offer much more personalized information to users. For example, optimizing a step counter to match the height, age or Body Mass Index (BMI) of a user – or to adapt to a user’s environment (is the person running on a beach, hiking up a mountain or strolling in a park?) – will provide more accurate information on calories burned. Not every step is created equal, and both pre-loaded personal data as well as real-world environmental data will prove that some steps consume a lot more energy than others.
SEMI: What would you like MSEC attendees to take away from your presentation?
Finkbeiner: I want to introduce the journey of software development by illustrating specific use case examples. I would also like to offer my outlook on the role of software and AI in MEMS sensors to help increase their adoption in current and new applications. Ultimately, I think it’s important to raise awareness in our industry on why we should embrace the use of software and AI.
Stefan Finkbeiner, Ph.D., CEO and General Manager, Bosch Sensortec, was appointed CEO of Bosch Sensortec in 2012. He joined the Robert Bosch GmbH in 1995 and has been working in different positions related to the research, development, manufacturing, and marketing of sensors for more than 20 years. His senior positions at Bosch have included director of marketing for sensors, director of corporate research in microsystems technology, and vice president of engineering for sensors.
Finkbeiner received his Diploma in Physics from the University of Karlsruhe in 1992 before studying at the Max-Planck-Institute in Stuttgart, where he earned his Ph.D. in Physics in 1995. In 2015, Finkbeiner received the prestigious lifetime achievement award from the MEMS & Sensors Industry Group (MSIG), a SEMI technology community.
Bosch Sensortec is a member of MEMS & Sensors Industry Group, the industry association representing the global MEMS and sensors supply chain. To learn more about how MSIG enables professionals in the MEMS and sensors industry to innovate, address common challenges and accelerate business results, visit us today.
Nishita Rao is marketing manager for technology communities at SEMI.