If Linköping University is to be a leading player in self-driving vehicles, we need to gear up. This is according to Theodor Westny, doctor at the Division of Vehicular Systems, who is the initiator of the new research and teaching platform.
“Much of our research is currently being developed and evaluated in virtual environments, but simulations can’t fully capture the real-time demands and limitations that arise in physical interaction with the outside world,” says Theodor Westny.
Sim-to-real
The goal of the project From Virtual Agent to Real Robots was to develop a so-called sim-to-real platform – a bridge between simulation and reality. For such a bridge to work, the simulation must be sufficiently realistic. The cars in the simulated world must mimic their real behaviour, both in terms of how things move and how the sensors on the cars react to the surroundings. The platform must also provide traffic scenarios with traffic rules, road signs and other agents. This is necessary to create a credible and interesting development environment.
“I’ve always worked with these cars, but from a training perspective. This is so the students have a platform that is easy to work with, and then with just a single car that will perform some kind of task. But putting everything together with several cars that interact with each other is really new for us. I started seriously working on this after defending my doctoral thesis.”
From Boston to Linköping
The inspiration for the platform came from a lecture by Karl Berntorp, who then worked at Mitsubishi Electric Research Laboratories (MERL) in Boston. There they had a similar set-up with small cars, and it was after that lecture that Theodor Westny started putting his platform into practice at Linköping University.
“There are many more simulators that are much more complex, but our platform is designed for the spaces we have – that is, Visionen. We have made the best use of this resource as possible.”
Visionen is a research arena for experimental research and teaching that provides a large and open research infrastructure. The arena is run by the Department of Electrical Engineering (ISY) and is funded by the research environment ELLIIT (Excellence Center at Linköping – Lund in Information Technology).
“The important thing was never that the platform would be something extraordinary, but rather to make sure that this resource exists. To have something that we can use to experiment with. This is important both for more interesting teaching and for research where we can put theories into practice.”
The importance of industrial collaboration
At the Department of Electrical Engineering at Linköping University, there is a focus on cooperation and large industrial collaborations. The platform is therefore built based on today's industry standards so that it may be passed on.
“When I talk about the connection to the industry, I’m mainly referring to the modular architecture. There is a built-in verifiability in it. If you discover that something is behaving strangely, it’s easy to identify where in the information flow something has gone wrong. That’s really how I’ve seen the platform the whole time – that you should be able to pass it on, for example to the passenger car industry.”
The future of Theodor Westny’s platform looks bright, as it has behaved as he intended. But more remains to be done.
“The next step is to start using Lidar instead of the sensors we use today.”
New 3D printed cars
The platform is already widely used for teaching purposes. For example, in the Autonomous Vehicles - Planning, Control and Learning Systems course, on which Theodor Westny teaches, where a new version of cars will hopefully be available in the autumn. The new cars are still under construction, but have great potential.
Regarding the course, Theodor Westny concludes as follows:
“The students think it’s great fun!”
