As autonomous driving improves, so too must the real-time evaluation environmental scenarios. Thomas Doerfler, managing direc explores how intelligent radar sensors could improve the s


For the development and validation of systems for autonomous vehicle management, real traffic scenarios are usually simulated and tested in the laboratory. Real data injection combines the advantages of test drives with simulation, bringing scenarios of the

real environment into the lab. For this validation method, the raw data of the sensor front ends are recorded bit-

precisely during one-time journeys. This raw data collection is then fed back to the sensors in the laboratory and can be used to validate newer software releases. However, the recorded raw data of a sensor syst em is never suitable for vehicles int erfaces. Therefore, a system recording real data must be coupled

directly to the respective sensor sy stem. In addition, it should be compact and

robust, because it must remain functional in a wide range of weather condit ions.

ACQUISITION OF REAL DATA A split structure, as developed for the gigabit data logger sy stem DP²4R from

embedded brains GmbH, provides one solution. A central unit (controller) is inst alled in the vehicle, managing the initialisat ion, control and data acquisition of up to four remot e head units. This

controller is connect ed to the head units via gigabit-compat ible cables. In turn,

each head unit is directly coupled to the sensor electronics.

The head unit, with it s FPGA, is

responsible for t he data transfer from the respective sensor electronics. Furt hermore, it acquires, formats and sends the dat a from the sensor electronics to t he controller. In turn, the controller aggregates t he data of all head units and st ores all of them.


T his data is then archived by using commercially available NAS systems at t he development site. However, very high capacities are needed. For example, one y ear of complete recording requires storage capacity in the single digit

petabyte range. The elegant way to transport the data from the test vehicle t o the NAS would be a network interface. Yet, this limits the practically

achievable data rat e to approximately 10Gbit/sec. T herefore, the SSDs should be manually removed from t he central processing unit and inserted directly

into the NAS. Combining a five-minute walk from the vehicle to the NAS and two 8-TByt e SSDs, this corresponds to a data rate of about 430 Gbit/sec, which is difficult to outdo wit h network cables. The data collected during the t est drives is subsequently ready for further use, preparat ion and evaluation.

me evaluation of traffic an ctor of embedded brains GmBH safety of autonomous

ic and

ins GmBH, drivin g


The collected raw data can be evaluated in many ways. The main benefit

lies in the ability to feed the data back into the sensors and evaluate their

Similar components are once again

used in this design. The head units are connected to the sensor systems to be validated; they do not record the raw data but feed it into the sensor system, instead of the front end. In turn, the head units receive the raw data synchronously from the controller, which in turn picks them up directly from the NAS.

In this kind of setup, the sensor systems pass through the previously

The hardware-in-the- loop (HiL) structure for sensor va

raw data injection

ware-in-the- ) structure for lidation with injection

recorded sequences in synchronisation. If their output towards the vehicle bus system is monitored and compared with the relevant objects, the decisive capabilities of the software can be verifiably quantified.


The intelligent sensor sy basis of tomorrow’s

challenges for validation and quality assurance. System validation based on real driving data is inevitable for O as it allows for reliable conclusions on the robustness of th

before the market launch. Methods and systems for t

or systems, as the vehicles, pose new ation and quality alidation based on nevitable for OEMs, ble conclusions on e sensor technology aunch.

are available and should be considered early on in development to achieve the project

safely and with the required care.

embedded brains Gm



ems for this purpose ould be considered

ment projects, in order ct goals on time, required care.

ata can be evaluated main benefit, though, feed the data back d evaluate th i

responses in a classic hardware-in-the- loop (HiL) setup.

Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44