In this ISJ exclusive, Rohde & Schwarz discusses its strategic partnership with fiveD and how the two companies are enhancing the R&S QPS security scanners using customised AI training data.
Since July 2024, Rohde & Schwarz has been a strategic partner of fiveD, a deep-tech startup specialising in hyper-realistic radar simulation.
The Erlangen based startup’s Radar Simulation Suite uses a newly developed simulation approach.
It can simulate raw data from any radar sensor, regardless of application and scenario, with an unprecedented level of realism.
Rohde & Schwarz is using this data to enhance its R&S QPS security scanners, which are used around the world for contactless security checks at airports, stadiums and other critical infrastructure facilities.

Synthetic training data for AI
The security scanners use millimeterwave echoes on the human body and artificial intelligence (AI), to assess whether someone is carrying dangerous or unauthorised objects on their person.
The larger the training database used to teach the AI and the higher its quality, the more powerful the scanners become.
Up to now, human test subjects have been required to gather training data, making the process time-consuming.
Simulated radar signals, however, allow the database to be expanded almost indefinitely – even with scenarios that cannot or should not be recreated with human test subjects.
As a result, security scanners can be adapted to changing security requirements more quickly and trained for completely new use cases more easily.
The range of applications for the R&S QPS now goes far beyond simply screening for dangerous objects.
For entry and exit checks in data centers, for example, the AI is trained to recognise USB flash drives and even slips of paper that someone is carrying on their person.
This prevents malware from being smuggled in and confidential data from being stolen.
High simulation data quality from fiveD
Simulation data should have the maximum level of realism while minimising computational effort. Developers use the Radar Simulation Suite to model the entire chain of effects – an application scenario with a radar sensor, signal propagation, environment and subsequent signal processing.
A ray tracing method adapted to radar signals is used to calculate realistic signal propagation.
This method takes into account detailed wave characteristics such as phase noise.

No time-consuming labelling of training data
Training data has to be labeled before an AI system can learn from it. For the R&S QPS, this means that the database must contain information about which test person with which object on their body corresponds to a given data set of millimeterwave echoes.
This process has traditionally been carried out manually, making it error-prone.
Fortunately, it is no longer needed when working with simulation data, since the simulation model already contains information about which object generated a particular echo signal.
The Radar Simulation Suite automatically labels simulated data sets, saving time and eliminating errors.
Thanks to the software, signal components – which are invisible to the human eye and cannot be labeled manually – are included in the labeling process for the first time.
In addition to the standard three location coordinates and velocity information, labeling gives the radar sensor data a fifth dimension, which is where fiveD got its name.
“A perfect strategic synergy”
“Collaborating with Rohde & Schwarz is a perfect strategic synergy for us that goes far beyond security scanner applications,” explains Marcel Hoffmann, Managing Director of fiveD.
The Radar Simulation Suite can be purchased directly from fiveD and is also used to help develop vehicle radar systems and industrial radar applications.
“When customers use Rohde & Schwarz test and measurement equipment for their product development, they usually do so from the prototype development stage onward,” says Andreas Hägele, Vice President Microwave Imaging at Rohde & Schwarz.
“By partnering with fiveD, we can discuss with them right from the simulation phase how they can most efficiently validate their simulation models with Rohde & Schwarz test and measurement equipment.”

