At LiU, there are several research projects on AI, autonomous systems and software and, not least of all within the framework for WASP, Sweden’s largest single research programme. The programme has a budget of SEK 1 billion, and LiU is the host university. Last year, WASP started the NEST (Novelty, Excellence, Syngergy, Teams) initiative, and took the decision to start a total of nine NEST projects. These will be carried out over subject boundaries, and all WASP universities will be involved in NEST.
Several LiU researchers
One of the NEST projects will be based at LiU and will involve several LiU researchers, such as Jonas Unger, professor of media and information technology, who will coordinate the project.
“The idea behind this is that we who work with it are world leading in our disciplines. This isn't just a single project – it's about building a new research environment”, he says.
At the turn of the year, it was announced that the nine NEST projects were to receive a grant of SEK 20 million each during a five year. The funding source is the Knut and Alice Wallenberg Foundation. This makes NEST a bigger initiative then the earlier initiatives within WASP, says Jonas Unger.
“The goal with NEST is to take on some of the biggest challenges within AI, which is about developing systems that can integrate data that comes from both the real and virtual worlds.”
Vulnerable systems
One area which would benefit greatly from integrating real data with virtual data, for example, is the automotive industry. Today, there are self-driving cars which can understand their surroundings with the help of sensors and cameras. Using these, the vehicle can understand how to navigate its surroundings. But there is a lot of manual work that goes on behind the technology in order to gather the necessary data. Furthermore, the technology has a certain vulnerability, as the data gathered often represent normal situations without disturbances.
“This leads to problems if you want to train the algorithm so that it can force the car to make an emergency stop if something pops up in front it. In the real-world data which is gathered, there probably aren't so many examples of that”, says Jonas Unger.
Using simulation, the researchers want to generate synthetic data which complements the data from real life.
Put into practice
To give the research results a practical application, the researchers are collaborating with several different actors: Ikea Communications (who work with product visualisiation), the company Arrivers (vehicle safety), the Swedish Transport Administration, AI Sweden, WASP WARA Media and Language.
For example, in their collaboration with the Swedish Transport Administration, the researchers want to see if it is possible to use other methods than pictures and sensor data to combine simulation with AI models.
"For example, this might involve combining input data that describe the flow of traffic in a city with data about the weather. If it rains or is icy, then the simulation can help determine where the risks for accidents are largest”, says Johan Unger.
Another example that the researchers want to study is simulation in product visualisation, and the technology called augmented reality, which is about laying digial pictures over reality. In collaboration with Ikea Communications, the researchers want to develop technology that allows customers to visualise how furniture will fit into their homes.
“This technology already exists, but there are many ways we can make significant improvements. Today, the technology is limited in terms of how realistic it looks when, for example, putting a sofa in your home. We also want to make it go faster in real time.”
Collaboration with Örebro
The NEST project at LiU is being run in collaboration with Örebro University. The collaborating researcher there is Amy Loutfi, professor of information technology. She says that developing AI technology that can adapt using data from both the virtual and physical worlds is at the heart of researchers’ methods.
“If we can bridge the gap between information from these two worlds, we will be able to make better visualisations, better predictions and, not least of all, develop better AI systems when it comes to interaction with people. We are seeing more and more how virtual environments and physical environments can be combined in order to get the best of both worlds”, says Amy Loutfi.
The initial work
The work is to begin in April 2022, and the first step for the NEST project – which is based at LiU – will be to establish a research environment for the long term.
“We hope that it will become a world-class distributed research environment here in eastern Sweden. We will work with the whole range of WASP instruments and international actors available to us”, says Amy Loutfi.
The researchers who are part of the NEST project at LiU are: Jonas Unger, professor of media and information technology, who will coordinate the project, as well as Michael Felsberg, professor of computer vision, and Fredrik Lindsten, associate professor in statistics and machine learning.