The automotive industry has been undergoing an unprecedented revolution in recent years with the rapid entry of new players and technologies to the market. Among other issues, future autonomous cars will require considerable processing power, to the extent that there are those who refer to today’s autonomous test vehicles as “a data center on wheels.”
In order to provide a solution to this challenge and others, the automotive industry is rapidly adopting deep learning methods. These methods will enable the development of ADAS (Advanced Driver Assistance Systems) and autonomous driving applications that require sensing of surroundings. According to ABI Research, by 2023, some 38 million cars marketed worldwide will be equipped with Level 2 to Level 5 ADAS systems. More than half of the 34.4 million cars incorporated with ADAS Level 2 are expected to use machine vision based on deep learning. IC Insights notes that the volume of chip sales for cars and other vehicles is predicted to reach $42.9 billion in 2021, compared to $22.9 billion in 2016. The existing infrastructure of general purpose processors is incapable of the deep learning intensive computation necessary for these cars’ applications.
Here’s where the Israeli startup Hailo (www.hailotech.com) enters the picture. Hailo, which was named one of the 20 most promising startups by “The Marker” Magazine in July of this year, is currently accelerating its development of a deep learning chip which will provide data center processing capabilities to vehicles and other edge devices. In order to support the development efforts, the company recently announced the completion of A round capital raising amounting to $12.5 million. Since its inception, Hailo has raised a total of close to $16 million. Hailo’s processor is expected to provide the vital breakthrough necessary for processing consumption, heat dissipation, as well as area budget and cost. It will be suitable for mass production according to the safety qualifications required by the automotive industry.
With the help of deep learning capabilities embedded in the vehicle, cars will be able to gather data from sensors and analyze it in real time so that the autonomous car can make the right driving decisions despite what may be challenging surrounding conditions. Deep learning capabilities, when combined with other technologies, will enable autonomous cars with a wide range of capabilities including creation of a three-dimensional model of surrounding vehicles and prediction of their expected driving route, recognition of obstacles and pedestrians located in the vicinity, and identification of the road, lanes and routes where driving is permissible.
Hailo’s leading team, which was established in 2017, includes CEO Orr Danon, CTO Avi Baum, and Hadar Zeitlin, who serves as the company’s Chief Business Development Officer. Danon served in a elite Israel Defense Forces technology unit and is a recipient of the Israel Defense Prize. Baum held several senior engineering management positions at Texas Instruments, including serving as CTO for the Wireless Connectivity Group. Zeitlin also served in the same IDF technology unit for nine years and was awarded the Chief of Staff Prize for technological excellence.
“Seventy-year-old architecture on which current day processors are based is incapable of meeting the processing needs of deep learning applications,” notes Orr Danon, joint founder and Hailo CEO. “Hailo is revolutionizing the architecture which serves as the foundation of the microprocessor in order to boost deep learning processing capabilities by several orders of magnitude. Our multidisciplinary team has developed an innovative processing architecture structured around basic features of deep learning. For this purpose, we fundamentally redesigned the memory architecture, the control functions as well as computer processing, and the connection between them.”
In addition to its suitability for the automotive industry, the Hailo processor, which is expected to be rolled out in the first half of 2019, is designated for incorporation in a wide range of edge devices, from drones to smart home appliances. Hailo predicts that the chip will dramatically cut costs, reduce area budget and decrease power consumption in comparison to existing chips. The chip will be capable of processing high-volume video feed and data streaming in real time by using state-of-the-art deep learning algorithms. Hailo is currently cooperating with several of the largest companies in the automotive industry and other fields, which are examining the possibility of incorporating its technology.