This is the multi-page printable view of this section. Click here to print.

Return to the regular view of this page.

News

First building workshop at Fablab Cottbus

Our sensor is ready to be produced in large quantities. Together with members of the Fablab Cottbus e.V. - where the idea was born - we want to hold a first building workshop in January and put our building instructions to the test. After this trial run, we will also offer a building workshop for anyone else who is interested.
By Bene |

After four months of development, the CitRad sensor is now ready to count vehicles and measure their speed. At the same time, we have largely finalized the building instructions for the sensor and also pushed ahead with the development of the data platform. Now it’s time to check whether our documentation is also comprehensible to outsiders. In a first construction workshop, we want to invite members of Fablab Cottbus e.V., where we developed the sensor unit, to a first test build. We will use the findings from this event to run another optimization loop and then offer further events for all interested parties.

Once the required electronics have been assembled, the result is a compact package that only needs to be installed in the housing and connected to the battery.
Photo: Benedikt Stahl / CC-BY-CA

Focus and prerequisites of the first construction workshop

The central basis for data collection is a functioning sensor unit. Therefore, this first event serves primarily to check our building instructions for comprehensibility. The electronics are quite compact in their present form and the necessary soldering work is probably a little too demanding for absolute beginners. We are therefore particularly interested in those who want to help improve the instructions and have soldering experience. Participants may of course use the sensor units produced for initial test measurements. However, they will remain in our possession so that they are available for future measurement campaigns.

Other ways to obtain a CitRad sensor

We will use the experience gained from the first test run to improve our instructions and then offer further construction workshops and announce them accordingly. If you want to have your own CitRad sensor, you can either order the parts yourself (and assemble them with our help) or come along to one of the future workshops and discuss possible options with us.

Smart radar unit

Implementation of data evaluation on the radar unit
By Nanu |

After we have developed an algorithm on the computer to analyze the radar data, the whole thing now has to be implemented on the device. Our goal is that all the data is already analyzed in the box and we only need to store the number of cars and their speed.

Compared to the evaluation on the computer, there are a few things that need to be done differently on the device: On the computer, we worked with recorded data. On the device, we do the analysis live. This means that we have to buffer values in some places. For example, for smoothing the noisy raw signal. In one direction of travel, we have to buffer the speeds until the trigger of a passing car arrives. Only then can we save the data. For cars traveling in the other direction, we receive a trigger and then have to wait for the speed data.

The evaluation should later run completely on our sensor unit.

In addition, there is still a lot to optimize in order to evaluate the data efficiently on the small computing core.

You can follow our development in the SensorUnit repository on github.

Data analysis

How to count cars with a radar sensor?
By Nanu |

After all our experiments with amplifier circuits and a low-noise power supply, our sensor now seems to be working well. At least you can clearly see different vehicles in the spectral images of the radar data and recognize how fast they are driving. But how can the whole thing be evaluated with an algorithm? In the end, we just want to count vehicles and save the speed of each vehicle.

Data analysis on the computer. The algorithm counts cars and measures speeds.
Screenshot: Nanu Frechen / CC-BY-CA

The eye can easily distinguish the actual signal from the noise. In data analysis, this requires some smoothing and well-chosen limit values. The noise also changes over time. When the battery discharges. We also had to find a solution for this.

The signal analysis is based on the fact that certain frequencies stand out from the spectrum. This can be seen as lines over time. Each line is a moving object. A moving car, for example. The speed can then be calculated from the frequency.

We use a special trick to recognize when a car passes by: As the car passes, the angle to the Rarsensor changes. This causes the signal to bend downwards (towards 0 km/h). We recognize this bend and know that a car has just driven past the sensor. Whether this bend is to the left or to the right tells us in which direction the car was traveling. Now we just have to determine the detected speed before or after this trigger and save it.

What sounds simple is the result of a lot of fiddling around. And of course we first developed the whole thing on the computer using recorded data. This now has to be programmed and tested on the device itself. Only then will we be ready to give the device to interested parties for their own measurements. But that will be soon!

If you are interested in the technical details, you can take a look at our repository on github.

Data protection compliant traffic counting

Camera-based projects for analyzing traffic data exist as commercial and open source projects. We have deliberately decided not to build on these projects, as they bring with them various problems with data protection. We made a conscious decision to use radar as the basic technology. Radar allows us to collect traffic data without violating personal rights.
By Bene |

When we realized that we wanted to collect open traffic data and make it available for analysis, we quickly came to the question of how. Existing projects for the automated analysis of camera images were tempting, as they would have saved us a lot of basic development work. However, data protection-compliant traffic counting would not be possible. Because by recording people, license plates, etc., we would be operating in a legal grey area - even if we deleted this data immediately after evaluation. Radar does not have any of these data protection problems, but it does present additional challenges in terms of analysis.

The radar only recognizes individual objects without capturing details. As if a camera could only record blurred road users.

Anonymity of the data

Radar technology works by emitting electromagnetic waves that are reflected by vehicles. By analyzing the reflected waves, the system can collect information about the traffic flow without collecting personal data. This ensures the anonymity of road users. In contrast, image-based systems can collect potentially sensitive information about the identity of drivers and passengers, which is not the case with radar technology.

As we are of the fundamental opinion that personal data should be kept out of data sets where not absolutely necessary, Radar was the ideal solution for us. Of course, a pleasant side effect was that we did not have to concern ourselves with GDPR compliance. Because even if the image data had only been analyzed on the microcontroller and no one had seen it, numerous problems would have remained unresolved. For example, how do you ensure that third parties cannot gain access to the devices and access raw data that has not yet been deleted? Will we have problems with acceptance of the system if people have the feeling that they are being recorded on camera just because they are driving along a road in accordance with the rules? We were able to put all these considerations aside with a clear conscience by using radar.

Technical requirements and advantages

As far as radar technology itself is concerned, there are some advantages but also many challenges. Camera images are easy for people to grasp. This makes it much easier and quicker to check the evaluation mechanisms during development. Radar, on the other hand, provides us with a spectrum for evaluation that is very difficult for inexperienced viewers to decipher. The initial effort and interpretation of the results will therefore tend to take more time. In the long term, however, radar works more reliably than image analysis. Radar does not care what light conditions prevail. The reflected waves look the same in pitch darkness as they do in broad daylight. Different weather conditions are also no problem for radar. Fog, rain and snow should, at least in theory, be manageable as soon as the evaluation algorithm has been properly adjusted. We will describe how successful we are in doing this in the coming blog posts.

Conclusion

CitRad collects traffic data in compliance with data protection regulations and makes it available for further analysis. Nobody who drives past a CitRad sensor has to fear that data records will be created somewhere that will be associated with them. And there is certainly no need to fear points in Flensburg. Even if we detect speeding violations, we can only record them statistically and cannot draw any conclusions about the exact car or driver. Misuse of data is therefore also ruled out. As far as the accuracy of the evaluation is concerned, it will certainly take some time before we can include all eventualities in the evaluation. Until then, however, the CitRad sensor already provides good trends and quantitative overviews.

CitRad starts at the Prototype Fund

CitRad was one of 25 projects to be recommended for funding within the Prototype Fund and will now be supported by the BMBF for the next six months for the development of the evaluation and data platform.

After the hardware was developed the year before, CitRad is one of the funded projects of the 16th round of the Prototype Fund Software. The funding from the Federal Ministry of Education and Research enables us to improve the sensor’s evaluation routines and work on a data platform. The funding period runs until February 2025.

Following the development of the hardware and the initial tests, we took the opportunity in 2024 to present CitRad at district festivals and events such as the car-free university day and Parking Day. The response was consistently positive. We also presented CitRad to the city of Cottbus in spring 2024. It quickly became clear that there was a basic level of interest from a wide range of parties - from private individuals to mobility research and urban development.

Successfull application to the Prototype Fund

It quickly became clear that we wanted to continue working on CitRad. The fact that the application phase for the 16th round of the Prototype Fund had just started played into our hands. So we had to write an application and hope. But not for too long. Soon the invitation to the kick-off event on September 1 in Berlin fluttered into our mailbox.

CitRad is now one of 25 open source projects that will be funded for six months. Six other projects can also be assigned to the topic of city/mobility. We are excited to see where we will find intersections and look forward to an exciting next six months