Car radar spots the danger

Radars found in many cars collect a lot of data that is never used. In his thesis, Christian Lundquist shows that this type of information can be of use in designing an automatic right-of-way system.

Today several models already include braking functions that halt a car if it gets too close to another car. At the front of the car, radar picks up these movements but this is only a small part of all the measurement data from the radar that is put to use; the rest is discarded.

In his thesis on sensor informatics at the Department of Electrical Engineering, Lundquist examines the usability of the squandered information. With the help of various sensors would the car be able to monitor what’s going on around it, avoid oncoming traffic, keep from ramming a guard rail, or notice that the road takes a sharp turn?

Lundquist explains, “The radar can differentiate between stationary and moving objects, it identifies an object in close proximity and that it’s mobile. However the radar is unable to differentiate between a truck and a bicycle.”

Here a camera has the advantage, however it does not work in bad weather. In good weather, a camera can easily direct the car to follow the white lines on the road, but in snow or fog it fails. Here the radar works better, regardless of the weather and it can identify and measure the distance to the posts on the shoulder as well as to the car in front.

Lundquist studied how data from several different sensors like cameras, radars, gyros, and so on are put together so as to estimate signals more accurately and achieve more robust systems. He also looked at how filters, which screen noise from the signals so that they can be interpreted more clearly and they can be made more efficient so that they use less computing power. This is so the car can, in the best way possible, identify how the road curves and whether any obstacles exist ahead.

In the event that an obstacle arises, the car will identify their location and any available surfaces to use for evasive manoeuvres.

“Sometimes it can be safer to drive around than to brake sharply,” he argues.

Lundquist received his civil engineering degree at Chalmers and worked for several years as a subcontractor in Germany; he is now about to get his doctorate from LiU. And he’s thinking of staying.

“I want to study methods of determining the size of the obstacles encountered, for example to be able to determine whether it’s a truck or a bicycle approaching my car. This is a research area – extended target tracking – that is getting hotter and hotter, especially in the automotive industry, where the distance between the radar and the target is quite small,” he says.

He is also tempted to commercialise the part of the research that deals with locating people or places.

“We started a company to address this. It’s the same algorithms but with an entirely different application that deals with locating and helping people navigate indoors. It could be finding a gate at an airport, a certain store in a shopping centre, or a building here at LiU,” he says.

Lundquist defended his thesis on Friday, November 25; his supervisor was Fredrik Gustafsson, professor of Sensor Informatics.
Thesis Sensor Fusion for Automotive Applications, by Christian Lundquist, Automatic Control, Department of Electrical Engineering, Linköping University, 2011.

Monica Westman Svenselius 2011-11-22