One of the hottest AI topics these days is self-driving cars. According to the recent Boston Consulting Group study, self-driving cars will account for 25% of new car sales by the year 2035, with the market for partially and fully autonomous vehicles hitting a staggering US$77 billion.
RoboSense is positioning itself to capitalize on this explosion. Based in Shenzhen — the hardware capital of China — the company is in the business of LIDAR (Light Detection and Ranging) for autonomous vehicles. The global market for this versatile, high-precision, long-distance measuring “eye” will reach US$36 billion in 2030. Aside from established suppliers like Silicon Valley-based Velodyne; challengers such as Quanergy, Innoviz, and LeddarTech are also joining the LIDAR race. With its high cost-effective R&D and production, many industry insiders regard RoboSense as the dark horse — the DJI of LIDAR.
RoboSense launched its first generation 16-beam RS-LIDAR in October 2016. In June 2017 the it unveiled a multi-LIDAR coupling platform, and in September introduced its new 32-beam LIDAR models: RL32A and RL32B.
RoboSense CEO Chunxin Qiu finished his PhD on outdoor robotics perception at Harbin Institute of Technology, where he was first exposed to LIDAR technology. RoboSense was founded 2014 when Qiu asked his supervisor Xiaorui Zhu and PhD peer Letian Liu to join his team as Chief Scientist and CTO respectively.
RoboSense’s Quest Lower LIDAR’s Cost
One of the biggest LIDAR challenges is the technology’s high cost, and addressing this is an industry priority. Some companies are promoting ADAS (Advanced Driver-Assistance Systems) solutions to replace LIDAR with computer vision augmented cameras. However LIDAR remains necessary if we hope to put L4 and L5 level self-driving cars on the road.
LIDAR typically functions by spinning 360° atop a car to survey the surrounding environment. RoboSense is aiming to lower unit price to less than US$1,000 through the solid-state hybrid LIDAR, which eliminates this mechanical rotation, reducing cost and size while improving resolution, controllability, and system stability.
“The majority of our product is not too different from what’s on the market. Our technical breakthrough is solid-state LIDAR.” explains company COO Chunchao Qiu.
Solid-state LIDAR is not yet a fully mature solution, but companies are willing to give it a try. This year Silicon Valley-based startup Quangergy released its solid-state S3 system at the CES Expo. The new design uses an optical phased array as transmitter and has zero moving parts. Israeli self-driving car platform Innoviz is using a micro-electromechanical system (MEMS)-based design for the same purpose. Others like Innoclue and LedderTech are exploring their own paths.
RoboSense also has a multi-LIDAR coupling platform to address high cost issues. Compared with standalone 64-beam LiDAR, the standard coupling solution (comprises four 16-beam LiDARs) delivers the same point cloud density required at only a quarter of the cost of a 64-beam LiDAR. It’s worth mentioning that the platform can be customized according to customer needs. Now companies like drive.ai and Ford are also focusing on similar solutions.
Solving the Challenge of Production
As COO Chunchao Qiu tells us, RoboSense underestimated the difficulty of producing products at scale: four months delivery time was stretched into half a year. “Back then we needed to finish market development at the same time polish the product.”
The company now has over 100 dedicated LIDAR production lines, and Shenzhen’s high-end manufacturing industry is helping out big time in terms of money and talent.
In 2016 the company decided to focus on 16-beam LIDAR production. Current applications are low speed vehicles including street sweepers, logistic trucks, tool trucks, shuttle buses, rapid transit systems (BRT), and vehicles operating on designated road segments.
LIDAR used on autonomous vehicles can be divided into single-beam and multi-beam types. Single-beam LIDAR scans the target with a single laser beam, producing quick measurements with high resolution while consuming little energy. A leading single-beam supplier is Germany’s SICK. Multi-beam LIDAR include 4-beam, 8-beam, 32-beam and 64-beam products which can make multi-dimensional measurements. Well-known suppliers include Ibeo and Velodyne.
“For multi-beam LIDAR we are spending most of our energy on the 16-beam model,” explains Chunchao Qiu. He says that the 32-beam model benefits greatly from the 16-beam model’s R&D. “When the market matures for a 32-beam model, we will do a direct launch. But in the short-term 32-beam is not a priority for production, product lines should be tailored to market demand.”
The self-driving industry is huge, spanning from R&D teams to suppliers and governments. “LIDAR is a key component, but we are just a supporting role for autonomous driving environment perception. However, we want to perform at the highest level in that supporting role.” says Chaochun Qiu.
Industry Analyst: Jingyi Gao | Localization: Meghan Han | Editor: Michael Sarazen