New demands for on-vehicle data processing, and over-the-air updating, are expanding the use of these programmable semiconductors in production vehicles. The recent Daimler-Xilinx linkup shows the way forward.
The increasingly varied nature of data tied to safety systems and connected cars and trucks is altering electronic architectures, putting more emphasis on adaptability during design phases and after new vehicles enter the field. Field Programmable Gate Arrays (FPGA) are increasingly seeing use in production vehicles, with expectations that usage could grow as artificial intelligence (AI) and over-the-air updating become more commonplace.
FPGAs are semiconductor devices that are based around a matrix of configurable logic blocks, connected via programmable interconnects. They can be reprogrammed to desired application or functionality requirements after manufacturing. In this way FPGAs differ from Application Specific Integrated Circuits (ASICs), which are custom manufactured for specific design tasks.
In new vehicles going forward, inputs now come from multiple sensors and wireless links, areas where changes occur far more regularly than in conventional automotive systems. AI also requires the ability to adapt to changing patterns. These shifting demands for data processing are helping FPGAs expand their role in production vehicles.
Programmable devices from Xilinx and Intel/Altera migrated beyond prototyping a few years ago, largely in rapidly-changing infotainment systems. Now, the image processing requirements of cameras, radar and LiDAR provide a boost for FPGAs, as does the looming implementation of AI.
According to Grand View Research, automotive is now the third largest global market for FPGAs, after industrial and telecom. Another analysis firm, Markets and Markets, predicts FPGA revenues will rise from $5.83 billion in 2017 to $9.5 billion in 2023, noting that rising vehicle volumes in the Asia-Pacific region will drive rapid FPGA growth in automotive.
Xilinx, which has shipped over 40 million parts to OEMs and Tier 1s, is claiming significant progress in full-run vehicle shipments. In 2013, its chips were in 29 production models made by 14 OEMs. This year, they are in 111 production models from 29 OEMs.Recently Daimler announced it is teaming up with Xilinx so its deep-learning experts at the Mercedes-Benz Research and Development centers in Germany and India can develop AI algorithms on an adaptable Xilinx platform.
“Through this strategic collaboration, Xilinx is providing technology that will enable us to deliver very low latency and power-efficient solutions for vehicle systems which must operate in thermally constrained environments,” said Georges Massing, Director User Interaction and Software, Daimler AG.
Xilinx is competing with Nvidia graphical processing units (GPUs), Intel’s Mobileye vision processing devices and the FPGAs Intel gained by acquiring Altera. Willard Tu, Senior Automotive Director at Xilinx, said Xilinx devices provide more transparency than Mobileye’s black box approach. If there are problems, that makes it easier to debug. He added that FPGAs can be faster than GPUs.
“GPUs batch parallel tasks, holding some until a set number arrive. That introduces latency,” Tu explained. “We do parallelism, running batchless processes where each input is an independent piece of data. There’s no queueing, so all elements have the same latency.”
He noted that as connectivity brings security concerns, FPGAs provide an extra layer to defense-in-depth protection schemes. Tu compared silicon to a door lock, saying that once hackers find an opening, they can continue to exploit it even after software has been updated.
“Hardware is the lock, once hackers figure out how to defeat that lock, they know how to get in. You can change the software, but they can still get in. With FPGAs, you can change the lock, closing that vulnerability for good,” he asserted.
While conventional processors scale by moving to higher clock rates or adding cores, FPGAs can be upgraded without major redesigns. When alterations are needed, programmable logic can be upscaled by adding more fabric, which is simpler than redesigning a processing unit or waiting for faster parts. That is important as more factors change as OEMs move towards autonomy.
“When you look at data aggregation and pre-processing and distribution, it’s hard to predict how many cameras, what type or radar and the style of LiDAR will be used,” Tu said. “There are a lot of variabilities in sensors, and they may link to CAN or Ethernet, so there’s a real need for programmability.”
- Subscribe to SAE MOBILUS for access to more than 200,000 resources, including autonomous vehicle technical papers, eBooks, magazines, and videos.
- Bookmark SAE MOBILUS training for information about professional development and training opportunities.
- Bookmark SAE MOBILUS events to stay up to date with upcoming events about automated and connected vehicles.
- Bookmark SAE MOBILUS news to keep pace with the latest automated and connected vehicle news and information.
- Get connected and join the conversation about automated and connected vehicles at the Automated-Connected Wiki.
Terry Costlow is a contributing writer to SAE International.Original Article