Chinese autonomous driving startup DeepRoute.ai has introduced its second-generation inference engine technology designed to accelerate the path to Level 4 self-driving cars.
The DeepRoute-Engine, which debuted at CES 2021, accelerates neural network computation, offering a six times faster inference process than other open-source deep learning frameworks, the company said. The technology is compatible with graphics processing units (GPUs) of various brands including Nvidia, Intel and AMD to minimize the computation needed for Level 4 self-driving technology.
(Learn more about the levels of self-driving with “A Primer to the 6 Levels of Autonomous Driving.”)
The technology can run on DeepRoute’s 45 W computing platform, called DeepRoute-Tite, which consumes only one-ninth of the power compared with other neural networks, DeepRoute said.
Additionally, the company is showcasing its all-in-one sensing solution, DeepRoute-Sense II, which houses three lidar sensors, seven HDR cameras, a global navigation satellite system (GNSS), an inertial measurement unit (IMU) and a synchronization controller. This allows vehicles to accurately detect objects at a range of 153 yards and comes in a roof-box design.
In the video, DeepRoute shows the importance of inference engine and how it inputs the data from different sensors from deep learning models to make a prediction for the vehicle about the road in front of it. The faster the inference is, the more time is left for the planning model to increase safety. In a neural network, there could be hundreds of nodes, with thousands of ways used for calculation by the vehicle. The calculation speed varies and the artificial intelligence used by DeepRoute’s technology is designed to reduce the computation needed to make a prediction.
CES 2021 is all-digital and takes place Jan. 11-14.
