Terranet self-driving sensor tests show swift reaction time
LONDON, July 21 (Reuters) - Self-driving car technology developer Terranet (TERRNTb.ST) said on Wednesday that tests of its sensor system have shown a dramatic improvement in reaction time to objects in the road, which should help cars brake faster and reduce accidents.
The startup has been conducting tests on a Mercedes-Benz car in partnership with Daimler AG (DAIGn.DE).
Terranet said the tests showed its 3D motion awareness technology, called VoxelFlow, could identify and react to objects in or alongside the road in less than 20 milliseconds, far faster than the 300 milliseconds common for the Advanced Driver Assistance Systems (ADAS) available in systems today.
Chief Technology Officer Nihat Küçük, a former Mercedes-Benz executive, said the system used artificial intelligence to quickly identify an object, whether a plastic bag blowing in the wind or a child running between parked cars, and prepared to brake if there was a safety risk.
"We can't cheat physics and reduce braking distance at speeds, but we can reduce the reaction time before braking," Küçük said, adding that the company was talking to other carmakers and auto suppliers about building prototypes to test the technology.
Automakers are pushing to develop self-driving cars, but this poses significant challenges, especially getting vehicles to mimic the complexity of the human brain in recognising and reacting to objects in the road at high speeds.
Many automakers have focused on light detection and ranging (lidar) sensors, using laser light pulses to render precise images of the environment around the car, to develop higher levels of driver assistance on the fully self-driving vehicles.
VoxelFlow consists of three event cameras and a continuous laser scanner to create a 3D image of the road ahead.
Küçük said lidar was better suited to highway driving but Terranet's system was designed for more urban, crowded areas where a swift reaction time is needed to avoid pedestrians and cars.
Our Standards: The Thomson Reuters Trust Principles.