With a new system AI vehicles get even smarter

Jun 19, 2017 | Research/Cooperation

The human factor lies behind 90 percent of all traffic accidents. But if vehicles can learn to identify risk behaviours in pedestrians, cyclists and drivers, a large number of these accidents can be prevented. This is what researchers in the new project SimuSafe, a major international venture in which Mälardalen University (MDH) is taking part, believe, and which will take three and a half years to develop such a system.

SimuSafe consists of sixteen partners, universities as well as companies and associations, from eleven countries around the world. The project is managed from Instituto Tecnológico de Castilla y León in Spain and is funded with close on eight million Euros by the European Commission. MDH is responsible for one of nine work packages in the project, and deals with intelligence data analysis on a high volume of dataset with large variations. Huge amounts of information, which in different ways is collected from road users via sensors, internet of things, simulators etcetera is to be analysed to be able to see contexts and patterns that can indicate increased risk behaviour.

Mobyen Uddin Ahmed and Shahina Begum both work as researchers in embedded systems at MDH, and are participating together in the project with the project manager Gunnar Widforss.

- The ultimate objective of the whole project is to understand and develop realistic human behavioural computer system based on simulation, artificial intelligence, virtual reality and data science methodologies, and by means of these predict accidents. The system should be able to be used in different types of vehicles to increase the representation of driving patterns and computation of risk metrics. If vehicles with advanced AI and machine learning technology can work out how people react, accidents can be prevented to a greater extent and thus be reduced. To develop such a complex system on your own is very difficult, but in a big project like SimuSafe we should be able to achieve it, says Mobyen Uddin Ahmed. 

To collect all the data, tests are being made on a total of 801 people of all ages and countries, of which 90 people are in Sweden. The tests are carried out in reality on test tracks, as well as by using so-called virtual reality and with simulators. In these tests, real-life traffic situations are simulated to be able to collect data from these about how the road users behave, and which are then analysed by the work team at MDH.

- We have extensive experience of analysing huge amounts of data. We have also been working a lot and for a long time with monitoring people, that is to say collecting information about physical and mental states of health by means of sensors. In this project, we can combine these two areas of expertise, with the aim of developing a system that can make a real difference to road users all over the world, says Shahina Begum.   

The SimuSafe-project has received funding from the European Union's Horizon 2020 research and innovation programme.