Going beyond typical network backbone discussions, Gennady Benderman discussed the future needs of high-resolution adaptive headlights, presenting the company’s open thinking about optimizing the car’s headlights E/E architecture to ensure real-time beam adaptability.
Carrying out his research on behalf of Porsche as an external PhD student at the Karlsruhe Institute of Technology, Benderman first introduced the high-resolution adaptive headlight concept in a video whereby a camera-based ADAS system identifies and tracks all users on the road ahead, and feeds the data back to a controller itself driving a LED matrix.
“We’ve been working with 84 tracking segments, each object gets its own light distribution”, Benderman explained before looking at the challenges ahead.
For high resolution headlights, technology is moving from LED-matrices offering up to 1000 pixels, to LCD matrices capable of more pixels (up to 30,000) and ultimately, the adoption of MEMS-based Digital Mirror Devices (DMDs) with hundreds of thousands of movable pixels capable of reflecting and multiplexing an intense light source such as a laser into many different sectors and at different levels of intensity.
The benefits of moving to denser light arrays include the possibility to create smaller masked regions (more discretization of the non-illuminated areas on the road), without relying on mechanical actuators, but such adaptive headlights are less energy-efficient and require more complex algorithms to run, with yet more data and control commands to process in real-time.
As well as improving night driving conditions, high-resolution headlights featuring in the range of 100,000 pixels could bring up novel light functions, the PhD student said, showing a Porsche logotype projected in grey shades from a 30,000 pixels prototype.
If you follow the logic, I guess it is only a matter of time before drive-in movie goers can run a black & white movie through their car’s headlights onto a wall.
Back to the presentation, Benderman evaluated different network architectures, understanding that the high-resolution algorithms required to drive the individual left and right headlights ought to run close-by, his preference was for a centralized system structure. His assumptions were that each headlamp with 100,000 addressable pixels driven to 8-bit grey levels (256 shades) at a 50Hz frame rate would require 40Mbit/s data rates. This would call for much higher bandwidths than what today’s CAN with Flexible Data Rate (CAN-FD) could accommodate.
Then, Benderman exposed various Ethernet topologies, each one with trades-off in bandwidth, number of ports required from the master ECU, and possible latency issues between left and right headlight signals.
“Synchronization is needed, as well as a real-time transport protocol. We must also ensure data integrity in front of the lighting systems”, he stressed, adding that Porsche was also evaluating different image JPG compression ratio to lighten the data load.
Visit Porsche at www.porsche.com