Smart Headlight

As the degree of automation in mobility increases, so do the demands on the sensors used and their number in the vehicle. This, however, is partly at odds with vehicle design. This environment, a competition for space, will intensify further with growing electrification and automation. Electrification will turn the automobile’s front panel into a new, prominent surface and put the traditional sensor placement, behind the radiator grille, to the test.

Semi-autonomous and autonomous driving requires not only more sensors, but different kinds as well – such as lidar, radar, ultrasound and cameras – to ensure that the overall system performs best regardless of interference. In other words, the environment must be monitored as quickly and reliably as possible. Furthermore, the sensor fields must be congruent with each other to enable data fusion.

This is where the headlights come into play: They play a strategic role in how sensors can be integrated. Cameras and lidar, which operate in the optically visible or near infrared range (NIR), can be perfectly accommodated in the headlight and are, thus, protected from environmental factors.

The Smart Headlight project, which brings together researchers from different disciplines, addresses key questions regarding how feasible it is to coaxially integrate radar and lidar into the headlight itself. The project takes into account illumination, the measuring range, interference immunity, the field of vision, temperature development and soiling.

The wavelengths of the three systems span a total of four orders of magnitude. The researchers superimpose them coaxially via an optical system and couple them together in the direction that the vehicle travels; they combine the different light spectra using bi- and multispectral combiners.

Both optical systems require thin-film systems, which have been developed and applied by Fraunhofer FEP. The coating of the bispectral combiner is optimized so that it is highly reflective for the lidar wavelength of 905 nm and at the same time highly transparent in the visible spectral range from 400 nm to 650 nm without shifting the chromaticity coordinates. The multispectral combiner also couples the radar radiation coaxially.

Both lidar and light require a coating with anti-reflective (AR) properties. To enable radar reflection, the researchers use a transparent conductive oxide – in this case indium tin oxide (ITO) – for the highly refractive component of the AR thin-film system.

The scientists have not only optimized the thin-film system: To ensure that radar, lidar and illumination work together optimally, the Fraunhofer ILT team used ultrashort pulse laser structuring to partially remove the AR multilayer system on the radar-facing side of the multispectral combiner.

The resulting structures are significantly smaller than the radar wavelength used, so that the multispectral combiner can be regarded as a homogeneous specular reflector that supports both horizontal and vertical polarizations.

The structuring reduces the range of the radar by only 15 percent, but increases that of the lidar. The radar sensors used in the frequency range from 76 to 81 GHz are standard components and have been established in the automotive sector for several years.

Lidar systems pose a challenge, however, as they cannot directly identify the origin of the photons: The systems not only record photons from the laser systems in daylight, but also those from the sun. In order to filter out the sunlight, the systems therefore send and record several hundred laser pulses for a single distance measurement, but this generates a large amount of data.

To solve this issue, Fraunhofer IMS has developed an intelligent algorithm to efficiently reduce the large amount of data. It reduces the memory needed by a factor of 200: from 2 MB to 10 kB. Compared to conventional lidar systems, which have frame rates of 10 to 25 Hz, the newly developed algorithm achieves a frame rate of up to 144 Hz.

The fast processing of the lidar data makes it possible to fuse the radar and lidar data for the joint detection of objects. The Smart Headlight concept enables coaxial emission of the radar and lidar system so that both systems can observe an object from exactly the same angle. Since the data are fused, each technology can optimally play out its advantages to determine distances and recognize objects. Only data fusion ensures that the advantages of both technologies are effectively utilized for determining distance precisely and recognizing objects accurately.

In order to further reduce the space needed to install the fully integrated smart headlights, Fraunhofer IOF has replaced conventional beam-shaping elements such as free-form mirrors and aspheres with tandem microlens arrays. These enable high system transmission and offer new freedom, for example when designers choose the outer contour of the lights and their visual appearance when switched off.

Until now, tandem microlens arrays have been produced with buried metal apertures to do the actual beam shaping. However, as the surface area increases, so do the manufacturing costs. The aim of the newly developed micro-optical system is, therefore, to dispense with buried apertures in order to reduce manufacturing costs. The basic approach is to use irregular-edged microlenslets instead of the metal apertures previously used for beam shaping.

The Fraunhofer Institutes involved and the industrial advisory committee – consisting of leading companies such as Hella, ams-Osram and Trumpf – have successfully carried out preliminary work. In a demonstrator for the Smart Headlight, the research team is currently verifying that the new microlens optics design functions correctly. The demonstrator will use an ECE-compliant low beam and a 23x segmented high beam to demonstrate the new microlens optics approach. In addition, the team will determine the optical losses of the lighting, radar and lidar. Further tests for combined, coaxial distance measurement and object detection of lidar and radar will examine the interference immunity and the detection distance.