Startup raises $25 million to develop energy-efficient autonomous driving reasoning system

According to foreign media reports, Silicon Valley startup Recogni announced that it has raised $25 million (about 170 million yuan) in Series A financing, which was led by GreatPoint Ventures, BMW iVentures and Toyota AI Ventures, the venture capital arm of automakers BMW and Toyota, automotive technology company Faurecia, Fluxunit, the venture capital arm of automotive lighting technology company Osram, and DNS Capital. Recogni is developing a visual artificial intelligence platform for self-driving cars.

Recogni is focused on enabling high performance and low power consumption for AI edge processing in self-driving cars. Founded in 2017, the company is headquartered in San Jose, California, USA, with a branch in Munich, Germany. Its founders have extensive experience in system design, artificial intelligence (AI), computer vision, and custom silicon design.

Recogni aims to revolutionize the perception processing of L2+ autonomous vehicles, giving it superior real-time processing performance while consuming virtually no energy. The company says its perception technology can not only identify cyclists and pedestrians, but also their exact location, speed, and distance, better than any competing product on the market. Recogni has designed a "visual cognition processor," an artificial intelligence platform for autonomous vehicles.

One of the challenges facing self-driving cars is the ability to process AI and machine learning algorithms without filling the trunk with expensive and power-hungry hardware. Although most AI-based machine learning systems are currently trained offline, self-driving cars still need to process sensor data in real time while navigating, which requires powerful computing power and hardware. Therefore, Recogni creates high-performance, low-power AI processing solutions for self-driving cars by more efficiently solving the problem of terminal reasoning in self-driving vehicles.

Recogni believes that today’s self-driving cars have reached a bottleneck in processing data, and that they will not be able to transition to L3 and L4 without using edge processing to implement AI-based perception software. The company says its system can provide superior reasoning performance, with more than 500 times higher energy efficiency than other solutions, and that improved performance enables better edge processing.

Most technology companies such as Google, Nvidia, Microsoft and Amazon use powerful and energy-intensive processors to perform AI reasoning in the cloud. Using AI to process autonomous driving decisions in the cloud will cause speed and latency issues, while using edge processing technology can allow self-driving cars to make driving decisions faster than humans while consuming the least energy.

Unlike cloud computing, where data is uploaded directly from the self-driving car to the cloud for additional processing, edge processing is closer to where the sensor data is generated. In addition, the amount of data generated by self-driving car sensors is often too large to be sent to the cloud for effective processing.

Recogni plans to use the funds to improve its reasoning system, enabling state-of-the-art sensors to fuse data from vision and depth sensors for use in Level 2 autonomous driving technology already being adopted by automakers, and to expand its engineering team.