Professor Lemmen, before your current position at the university, you worked at Ford in Cologne on the development of a whole range of steering systems. Which ones specifically?

Particularly noteworthy are, for example, the first electromechanical steering system for Ford in Europe, for which I was responsible for the Ford Fiesta model B299, a so-called column EPAS, an electric power steering system in the vehicle interior on the steering column, and also the so-called belt-drive EPAS, an electromechanical steering system in the engine compartment with a recirculating ball gear. I was also involved in the initial developments of the pre-series phase for the Ford Mondeo and the Ford Edge.

Based on your many years of experience in steering optimization, what are the most important characteristics of electrically assisted steering? 

Ideally, the driver should not even notice that the steering is electromechanical. However, due to its specific characteristics, including friction, inertia, and damping behavior, it is a major challenge to achieve this through the control of the steering actuator. Incidentally, the original motivation for developing this type of steering system years ago was to reduce energy consumption. The resulting reduction in steering feel was a challenge that had to be overcome. In addition, access to a steering torque interface was also developed, which is crucial for the development of autonomous vehicles. These systems are also the basis for many of today's steer-by-wire systems.  
 

What are the main aspects to consider for reliable steering in autonomous vehicles?

Several points come to mind. Among other things, the necessarily higher torque of the EPS motor because the driver is no longer constantly operating the steering wheel. Other key aspects in this context are changing track dynamics, the natural frequencies and damping of the steering, and, in the case of steer-by-wire steering systems, the communication latency between the actual steering actuator and the actuator for driver feedback. 

The key points you have just mentioned hint at a lot of new challenges. Can you give an example to illustrate this?

Already 15 years ago, in a joint paper with ZF – then TRW – we pointed out that tuning steering systems for a so-called tune, i.e., the tuning of a vehicle variant for different regions of the world, requires consideration of around 2,500 parameters. To a very large extent, these parameters are compiled manually or empirically by development engineers from a wide range of departments, some of which have different software versions. The parameters must be combined, checked for plausibility, corrected for errors, and tested, which can only be done reproducibly in the laboratory. And there is no end in sight to the growth in complexity when you consider developments such as Car2x communication.  

How can a tool provider like dSPACE help?

dSPACE offers the right tools for all work steps to meet the extremely high requirements in terms of safety, availability, and validation. Among other things, we used the dSPACE MicroAutoBox for function prototyping, the production code generation was model-based with the help of dSPACE TargetLink, and we implemented the ASIL-D validation required for steering systems with HIL systems from dSPACE. In this way, for example, preparation times and the number and duration of real test drives are reduced to a minimum.

What were the decisive factors for the purchase of the dSPACE steering test bench?

Because the reproducibility of real test drives is limited, yet the subjective impression in the vehicle is indispensable, I was looking for a steering system test bench for HIL-based validation with real forces and torques. Specifically, the aim was to generate both static loads of up to 20 kN and highly dynamic loads in a closed control loop up to over 35 Hz. This then enables tests of the dynamic vibration effects of a power steering system and a subjective assessment of the steering on the test bench with test drivers. In addition, for reasons of maintainability, energy balance, and laboratory infrastructure, the actuators should preferably be electric and not hydraulic. Another very important point is the open-source code of the test bench software and the dSPACE simulation models (Automotive Simulation Models, ASM) implemented with MATLAB®/Simulink®. This level of openness allows the user to make flexible adjustments. All in all, these are very high requirements, but the dSPACE test bench fulfills them all.  
 

For which use cases do you use the test bench?

Mostly for research and teaching purposes, but sometimes also for industrial measurements. At the moment, we are focusing on possible automated error detection of devices under test, for example, due to faulty assembly, during the execution of standard measurements. But we are also interested in determining the parameters of steering models and in performing regulatory analyses.

To what extent can the test bench replace road tests? 

To a high degree. It is generally possible to generate synthetic signals, integrate measurement signals, or use a vehicle simulated with the dSPACE Automotive Simulation Models (ASM) in real time and with a cycle time of 125 ms, for example, to steer the rod actuators according to the simulated track rod forces. Of course, the quality of the HIL test depends on the quality of the vehicle model.

Another special feature is a linear motor specially developed by dSPACE for this test bench. Which tests are possible?

This actuator opens up unique possibilities. Thanks to its reduced power, it allows subjective haptic investigations of the steering by test drivers, i.e., in a human-in-the-loop scenario. To do this, we replace the wheel actuator with a steering wheel so that a test person can perform and evaluate the steering movement themselves. The reduction in loads also allows the installation of precise torque and force sensors, which could otherwise be destroyed. 
The modified dynamics of the test bench control system and its actuators are also an advantage. When comparing the results of tests carried out first with the small actuator with low inertia and then with the large actuators with high inertia, a clear distinction can be made between the dynamics of the device under test and those of the test bench. 

Where do you see further potential applications for such highly dynamic test benches in the future? 

In trucks, too, current developments in the steering context are moving toward partially and fully electric steering systems. I could imagine that similar or even more complex issues will arise here, since the level of automation in trucks is higher than in today's passenger cars. dSPACE has a portfolio that can be used to test both the components and the networking with other control units. In this case, however, the issue is not limited to the road but will certainly also affect construction and agricultural machinery. 

 

Thank you for talking with us.