On January 7th, at CES 2020, Chinese self-driving startup DEEPROUTE.Ai (Chinese:” 元戎启行”) announced the launch of a new autonomous driving computing platform – DeepRoute Tite. The computing platform is based on the NVIDIA Xavier, a vehicle-grade processor platform, and is embedded with the algorithms required for L4-level autonomous driving, the overall power consumption has been reduced to 45 watts, the cost and volume have been significantly reduced.
NVIDIA launched Xavier in 2016, an artificial intelligence supercomputer designed for autonomous driving vehicles. In terms of computing power, Xavier has a computing power of 30TOPS (trillion times per second) and a power consumption of 30W. DEEPROUTE.Ai believes that Xavier has more advantages in cost and meets mass production requirements.
DEEPROUTE.Ai successfully transplanted the autonomous driving related algorithms to Xavier through the self-developed inference engine, reducing the cost of the overall computing platform solution to about half of the traditional solution, and reducing the power consumption to nearly nine points of the traditional solution. one. Zhuang Qi, the person in charge of the inference engine and hardware acceleration of DEEPROUTE.Ai said that the self-developed inference engine has made many optimizations in computing, making it possible to use a Xavier to meet the computing requirements of L4 autonomous driving. At the same time, the self-developed inference engine is also easy to port and expand, and can support computing platforms such as Nvidia, Intel, and AMD, and is suitable for deep learning frameworks such as PyTorch, Caffe, and TensorFlow.
According to info from DEEPROUTE.Ai, the company also developed a DeepRoute-Sense sensor solution at the autonomous driving perception layer. The on-vehicle camera DeepRoute-Vision and the synchronization controller DeepRoute-Syntric in this solution were independently developed by DEEPROUTE.Ai. At present, the DeepRoute-Sense sensor solution has been adopted by Dongfeng Motor and applied to L4 Robotaxi (autonomous taxi).