IARPA bets $4.8M on electro-photonic system developed by Lightmatter, Harvard, and Boston University.
Photonics company Lightmatter announced in a press release early last week that it has teamed up with Harvard and Boston University to create an “Electro-Photonic Computing (EPiC) solution for autonomous vehicles.” The project has received $4.8 million in funding from the Intelligence Advanced Research Projects Activity (IARPA).
The EPiC project aims to address one of the biggest roadblocks to realizing autonomous vehicle (AV) technology: the need for high performance computing that is low in both latency and power consumption. These requirements are difficult to juggle with conventional electronic processors, but Lightmatter believes an electro-photonic approach can provide better balance.
The proposed EPiC system employs photonics—chips that compute with photons rather than electrons—for large matrix calculations, and uses conventional electronics for data storage and non-linear operations. The system will be fully integrated with autonomous vehicle sensors, according to Lightmatter.
Autonomous vehicles require numerous sensors such as radars, LiDARs, and cameras, which consume significant power and generate massive amounts of data. Onboard computers must churn through trillions of calculations per second to process the data, and this too requires significant power. The more sensors are added, the more data is available to the machine learning algorithms, but at the cost of higher power consumption and longer processing times.
Achieving acceptable driving performance with enough power left over for acceptable vehicle range is ever an obstacle for AV engineers.
“The autonomous vehicle industry needs to overcome this major technical hurdle in order to truly improve driving range—having powerful enough on-board computing power to support trillions of calculations per second, without consuming extreme amounts of energy,” said Lightmatter Chief Scientist Darius Bunandar in the press release.
“We’re thrilled to be teaming up with Harvard, Boston University, and IARPA to fix that issue with Lightmatter’s Electro-Photonic Computing solutions,” Bunandar added.
Another Solution for AV Sensors
The quest to improve autonomous vehicles is multipronged, and other teams are trying different approaches. Last year, researchers at Western Michigan University (WMU) won a $2.5 million grant from the U.S. Department of Energy for a three-year project to develop infrastructure-based technology to improve AV efficiency. Instead of improving technology within the vehicle, they plan to embed technology within roadways or other external infrastructure where sensors could transfer data through a wireless network usable by the cars.
“You have all of these sensors that are drawing tons of power being fed into a supercomputer that processes all those sensor outputs and runs cutting-edge artificial intelligence. It sounds cool, but it turns out that it uses a lot of energy,” said Zachary Asher, the grant’s principal investigator, in WMU News last July.
Asher continued: “If we removed that and instead just had an electronic barrier on the side of the road where the car can basically download that information, we don’t need a high-performance supercomputer (in the vehicle) or high power consumption artificial intelligence equipment. It’s an approach that no one has really pursued too heavily at this point, and we’re the lucky ones who get to do it.”
