Quebec-based technology company LeddarTech is stepping away from the lidar-components business to focus on developing software to make autonomous-vehicle and advanced driver-assist systems (ADAS) more accurate and cheaper.
LeddarTech is shifting to what it sees as a major growth area in automotive, but the move pushes it largely out of a segment it has served for years. The move to a software-focused business model also resulted in dozens of layoffs at company offices worldwide.
For the past two years, LeddarTech had one division focused on lidar components and another on software that fuses together “raw” data from vehicle sensors, CEO Charles Boulanger said. Having dual priorities with limited investment resources, he said, was “putting more at risk on both sides.”
Lidar, an abbreviation for light detection and ranging, is key technology used by autonomous vehicles to map their physical proximity to the environment and vehicles around them.
“We came to the conclusion, given that the industry is now seeing raw data fusion as the key way forward, … we should put all of our effort on it.”
LeddarTech laid off about 50 people in early October as it formalized the transition, Boulanger told Automotive News Canada.
About half the layoffs were at the company’s Quebec City hub, with others at offices in Montreal, Ottawa, Germany and China. LeddarTech transferred about 30 employees from its lidar components unit, which designed systems on chips for lidar makers and Tier 1 and 2 suppliers, to its data-processing segment. It could not reallocate the employees who specialized in hardware, however.
The company now employs about 220 people at its offices worldwide.
LeddarTech is not completely out of the lidar-components business, however. It’s currently in discussions with partners looking to license the technology, which will enable the business line to continue, Boulanger said. The company also aims to spin off its “profitable and growing” module business, which designs lidar units for nonautomotive applications such as traffic management and mobility.
SHIFT TO SOFTWARE
LeddarTech’s expertise is in fusing sensor data from lidar, radar and cameras into digital models that AVs can use to navigate safely. The company has been doing in-house development on such software for years but accelerated its work through a 2020 acquisition of Israel-based VayaVision.
“For the OEM, it’s really about how we combine these sensors to reconstruct the environment around the car in real time,” Boulanger said.
In ADAS systems today, he said, each sensor perceives its environment independently from other sensors on the vehicle, which can lead to partial or contradictory readings. LeddarTech’s LeddarVision software uses a process the company calls raw data fusion. It relies on artificial intelligence (AI) and computer-vision algorithms to blend data from separate sensors in a way that gives AVs and ADAS systems a more wholistic view of their environment.
“Instead of trying to perceive what the sensor sees and then fuse the object, we do the opposite,” Boulanger said. “We fuse the raw signals of all these sensors, and we re-create … an abstracted relative model or a digital twin.”
LeddarTech’s software lets automakers reduce the number of sensors they put on a vehicle while still creating a more accurate model of the environment it than current systems, he said.
The process addresses a key scaling issue that automakers are running into as they outfit more models with ADAS technology, Boulanger said.
The traditional route of operating each sensor independently is a “dead-end street,” because automakers are forced to “start all over again” to retrain the sensors to work with every new model and every new generation of vehicle. This retraining can take millions of test kilometres to ensure that autonomous systems make the right decision in even the unlikeliest of situations, which are known in the industry as edge cases.
LeddarVision requires no such validation, Boulanger said.
“You can use the same architecture and change behind it, the sensor set, the configuration, the positioning,” he said. “You have some retraining to do, but it’s minimal. You don’t lose all the edge cases that your millions of miles have given you.”
LeddarTech would not identify the automotive companies it is working with but said it is collaborating with automakers and Tier 1 suppliers to replace older automated systems with its raw-data fusion technology. An example of what is known as middleware, LeddarVision is designed to live on an automaker’s wider software platform and work with sensors from all companies.
The software, which the company plans to license on a per-vehicle basis, is scheduled to roll out in off-road vehicles next year and in production passenger vehicles in 2024.