A lot of vehicle testing takes place where the car is driven around a circuit by real people. There are times when this is absolutely necessary. People with experience in such things need to ‘feel’ the car as it develops. In some instances, however, the testing process can be hampered by human error because humans are ‘variable’. That is, they won’t necessarily repeat the exact same action time after time after time.
If we can develop a machine to do some of these tests, you get the same circuit driven the same way every time and the vehicle data retrieved from such a test should be based on consistent inputs in every respect. More than that, steering robots are actually able to give more precise and more dramatic (i.e. faster) steering inputs that humans can, and they don’t get tired either.
The simple version of “How” involves a track modelled on GPS data and some sophisticated hardware being installed into the car to steer it according to this pre-programmed course. The hardware used to control the steering is a steering robot from Vehico and the work is done in cooperation with them. The steering robot is currently only able to perform so-called ‘open loop’ tests (steering as function of time) such as step-steer, sine-sweep steer, and sine-with-dwell steering, which are used for vehicle dynamics characterisation. The task for Klas and Carl, which even the experts say is very challenging one, was to create the control software necessary to control the steering such that the vehicle follows a pre-defined path using advanced GPS and motion sensors as input data.
The guys responsible for this job were Klas and Carl, two students from Chalmers University who completed this project in conjunction with Saab as their Masters thesis.
Their thesis presentation was made back in June and I’ve just been forwarded a copy of the video they made as part of that presentation. In the video, you can see the robotic steering system in action, a vehicle’s eye view of the road they’re driving as well as speed and tracking information.
It’s all fascinating stuff and indicative of just some of the technical work that goes on behind the scenes here at Saab.
I know what you’re thinking when you read that title….
You’re thinking this might be some sort of apocalyptic horror story where we aren’t allowed to drive anymore and robots from Sweden take over the driving, plugged into some evil worldwide network that monitors traffic, your speed, where you’ve been and so on and so forth. It sounds like Big Brother moving right into your garage.
Not so.
The story does involve robotics and driving, though, and it’s both fun to watch and incredible to experience first hand.
Klas Olsson and Carl Sandberg are engineering students from Chalmers University and what they’ve developed is a system whereby a car can be driven by a robotic steering unit according to pre-programmed parameters (circuit, speeds, etc). They were offered a chance to develop this closed-loop circuit system as part of their Masters Thesis project and in partnership with Saab. They did their work at the Chassis department, headed by Martin Öman, and were supervised by Dr. Matthijs Klomp, who is a development engineer in the vehicle dynamics simulation group at Saab.
Why and how?
This is the simple version. The details are something I can’t go into (even if I understood them fully) as they involve software and technical IP that belong to Klas, Carl and Saab. But here goes….
A lot of vehicle testing takes place where the car is driven around a circuit by real people. There are times when this is absolutely necessary. People with experience in such things need to ‘feel’ the car as it develops. In some instances, however, the testing process can be hampered by human error because humans are ‘variable’. That is, they won’t necessarily repeat the exact same action time after time after time.
If we can develop a machine to do some of these tests, you get the same circuit driven the same way every time and the vehicle data retrieved from such a test should be based on consistent inputs in every respect. More than that, steering robots are actually able to give more precise and more dramatic (i.e. faster) steering inputs that humans can, and they don’t get tired either.
The simple version of “How” involves a track modelled on GPS data and some sophisticated hardware being installed into the car to steer it according to this pre-programmed course. The hardware used to control the steering is a steering robot from Vehico and the work is done in cooperation with them. The steering robot is currently only able to perform so-called ‘open loop’ tests (steering as function of time) such as step-steer, sine-sweep steer, and sine-with-dwell steering, which are used for vehicle dynamics characterisation. The task for Klas and Carl, which even the experts say is very challenging one, was to create the control software necessary to control the steering such that the vehicle follows a pre-defined path using advanced GPS and motion sensors as input data.
I’ll let Klas and Carl explain more….. with a demonstration as well.
—
So, aside from the challenge and the associated fun of building a robotic vehicle, why are Klas and Carl doing this and why are Saab happy to provide the tools?
There are four main benefits with this thesis work from Saab’s point of view:
Evaluation of the Vehico steering robot
‘Closed-loop’ path control (autonomous driving) enables Saab to perform more vehicle dynamics testing in Sweden, saving travel and transportation costs and test-track rental.
Since the closed-loop control is co-developed with Saab we will have full access to the work and will be able to use the developed algorithms for future research into autonomous driving for lane-keeping tasks, platooning an other future functions.
This is a very advanced vehicle dynamics control task attracting top students, which will enhance our mutual relationship with Chalmers.
The immediate benefit of autonomous driving is to enable vehicle dynamics characterisation testing as well as legal compliance testing that is not normally possible on relatively narrow surfaces such those that Saab have access to locally in Sweden.
One example of a vehicle dynamics surface normally used by Saab is the one at IDIADA in Spain, a facility that costs significant money to use and is in constant demand. If Saab can do the same work at lower cost and in their own time, all the better.
Dr. Matthijs Klomp summarizes:
“As mentioned, the problem of closed-loop path control is a very challenging one. This in particular since the control must be accurate and robust in a wide range of conditions, including and up to the handling limits of the vehicle. Additionally the controller must be easy to adapt to changes in the vehicle and the road surface (i.e. snow, gravel, asphalt).
Klas and Carl started to implement their control strategy in our simulation environment using IPG CarMaker, who also sponsored this project. Subsequently they moved to implement their controller in the real-time computer of the steering robot and to get it all working in the real-world.
The end-result is truly amazing, the vehicle both follows the path very well in both the linear and non-linear handling range without the steering becoming nervous or erratic, a common problem otherwise, yet is very simple to operate for the test driver. I congratulate Klas and Carl to a job well done!”
