GAZELE - Generation of radar-echos from synthetic scenarios for automotive applications
The project GAZELE is a cooperation with the Institute of Automation and Control of Graz, University of Technology (IRT) and AVL List GmbH. durchgeführt.
Our aim is to develop a radar target stimulator, which creates complex virtual scenarios for an automotive radar sensor, in order to test modern driver assistance systems and autonomous driving software on a testbed instead of driving hundred of thousands of kilometres on real streets. While the IRT develops the software to calculate the necessary stimulation parameters based on an environment simulation (distance to target, velocity, radar cross section), our institute is designing and building the necessary microwave hardware.
The 77 GHz signal frequency of modern automotive radar sensors makes it practically impossible to build a radar target stimulator without frequency down-conversion. We decided to use an intermediate frequency (IF) at about 2 GHz. At this frequency the radar signal is delayed, shifted in frequency and attenuated according to the parameters of the target to be simulated. Afterwards its up-converted back to 77 GHz and transmitted back to the radar. This design is shown in the figure below.
The requirements that on the one hand targets down to about 5 meters have to be synthesized and on the other hand no a-priori knowledge about the radar sensor - and therefore also about its waveform - shall be used make a analog signal processing necessary, since the delays inherent in digital processing (eg. AD-/DA-conversion) are too large to allow such short target distances.
This unfortunately means a high hardware effort especially due to the analog delay-lines and target emulators. To circumvent this problem we developed a highly flexible and scalable design which allows to change both the number of emulated targets and the target range.
The photo shows a setup for demonstration purposes with the delay-lines and the switching-matrix in the foreground and the target emulator modules in the back.
One challenge of the design was the rather large - about 1 GHz - bandwidth for which the frequency response (amplitude and group-delay) should be as flat as possible. If this would not be the case the simulated target reflectivity would be heavily frequency dependent causing problems in the radar signal processing.
In this figure the measured - sufficiently flat - frequency response is shown.
At the end what counts is that the simulated targets are correctly detected by the radar sensor. As an example the figure shows the display of the radar used by us for testing purposes with 2 correctly identified targets at about 5 m and 30 m distance with different velocities.
These research topics will be extended in the framework of the ENABLE-S3 project.
FFG / Bridge