RoboSense (Suteng Innovation Technology Co., Ltd.) is a world-leading provider of Smart LiDAR Sensor Systems. With a complete portfolio of LiDAR sensors, AI perception and IC chipsets, RoboSense transforms conventional 3D LiDAR sensors with comprehensive data analysis and interpretation systems. Its mission is to innovate outstanding hardware and AI capabilities to create smart solutions that enable robots, including autonomous vehicles, to have perception capabilities superior to humans.
RoboSense is headquartered in Shenzhen, China, with talent from top global companies and scientific research institutions, empowering the team’s innovation capabilities. As of 2021, RoboSense owns more than 700 patents globally. Its leading products are built on its multi-disciplinary strengths in technology across diverse level of the business. With a market-oriented strategy, RoboSense provides customers with various Smart LiDAR perception system solutions, including MEMS LiDAR, mechanical LiDAR, fusion hardware, and perception software.
RoboSense’s portfolio has been globally recognized. To date, RoboSense LiDAR systems have been widely applied to future mobility solutions, including autonomous driving passenger cars, commercial vehicles, automated logistics vehicles, robots, the company’s own Robotaxi, Robotruck, Robobus, as well as new smart transportation infrastructures. Among them, the pre-installed mass production projects cover supercars, coupes, and SUVs, heavy-duty trucks and other types of vehicles.
RoboSense’s latest RS-LiDAR-Perception software is built into RS-Cube 2.0 to analyze LiDAR point cloud data of autonomous driving in real time and provide semantic environment perception information to handle various complex road scenarios. The perception software platform is configured with a powerful NVIDIA Jetson Orin computing page, which provides peak computing power of up to 32 TOPS, supports the work of multiple LiDARs and fulfills the point cloud fusion task of multiple LiDARs.
LiDAR sensor provides high-precision geometric distribution data of the surrounding environment. Compared with camera vision sensor, LiDAR sensor is not interfered by ambient light, has a long detection range and distance with no distortion and high accuracy and reliability, which are irreplaceable advantages in the fields of robotics, logistics, autonomous driving, etc.
However, compared with visual processing perception software and solutions which have widespread application, LiDAR faces the challenges of scarce post-processing perception software and high development difficulty. In particular, the point cloud data is 3D data, the processing of which requires high computing power. At present, mainstream perception software, including point cloud processing software with high ranking in KITTI autonomous driving data set, almost all rely on the “gigantic” industrial personal computers (IPC) to support real-time operation.
Taking autonomous driving as an example, all the complex practical issues in automatic driving, including various mix-ups of weather and road conditions, means that the requirements for accuracy and reliability of sensors, the running speed of perception software and the efficiency of real-time data processing are very high. On top of that, point cloud AI technology is relatively complex and has a high R&D threshold, while traditional perception software teams are confronted with lack of point cloud processing experience, accessory development tools and effective data sets, high power consumption and large size of hardware platform, and other problems.
Based on above challenges, RoboSense chose NVIDIA Jetson Orin platform with low power consumption and high computing power, which meets the system requirement for high computing power to support the simultaneous operation of multiple LiDARs and data processing, and greatly improves the speed at which perception software process large point cloud data, supporting and growing businesses to the next level.
1. Empowered by NVIDIA Jetson Orin hardware platform, RoboSense is able to quickly complete prototype testing, effectively shorten the R&D cycle time and minimize trial costs, which enables the R&D team to focus on the construction of multi-scenario and large-scale dedicated point cloud data sets and the performance improvement of perception software, rather than over-emphasizing on the hardware underlying development and adaptation.
2. NVIDIA Jetson Orin's small size and low power consumption make actual road test installation, commissioning, software deployment and easier, and evidently improve the efficiency of perception software development and road tests; its rich interfaces facilitate its connection to other sensors and equipment, and makes it flexibly adapt to various hardware testing platforms, saving adaptation costs to a certain extent.
3. NVIDIA Jetson Orin development platform has a mature upper system environment, which can quickly and seamlessly migrate desktop development achievements to it. WYSIWYG development ensures the consistency between development and deployment and eliminates embedded transplantation, testing and other links, which in turn accelerates product development.
4. For autonomous driving customers, NVIDIA Jetson Orin supports the use of multiple LiDAR integrated applications flexibly, and its small size and plug-and-play design further enhances user experience. The product facilitates RoboSense’s safe implementation of multiple autonomous driving medium and low speed scenarios and applications including V2X cooperative vehicle infrastructure, medium and low speed logistics robots, and industrial autonomous transportation.
5. Based on the wide coverage of NVIDIA's product lines and standard tool chains, customers can smoothly and quickly migrate from non-automotive grade to automotive-grade applications. For instance, software models developed based on Xavier can be migrated to automotive-grade in-vehicle domain controller platforms based on Pegasus Orin or other automotive-grade SOC.
NVIDIA Jetson Orin is a low-cost, high-performance embedded computing device. It can run the perception and positioning software of RoboSense to help customers better perceive the surrounding environment using LiDAR. Orin has become the no.1 solution for customers in environments with high cost and space requirements such as unmanned logistics and V2X. Xavier and RoboSense's integrated solutions enable customers to directly obtain structured data of the surrounding environment on top of individual spatial point cloud data, helping customers skip the complex point cloud software development step and achieve plug and play, and substantially shortening the overall development cycle of user products.
With NVIDIA Jetson Orin, the platform provides peak computing power of up to 32 TOPS. RS-Cube 2.0 can support multiple RoboSense's LiDARs and their integrated applications including RS-LiDAR-16, RS-Bpearl, RS-Ruby Lite and RS-LiDAR-M1, so as to realize real-time omni-directional perception of obstacles in the surrounding environment, such as detection, categorizing, tracking, semantic segmentation, etc. In an easy, fast and flexible way, it enables autonomous driving LiDAR with perception and analysis capabilities to meet the needs for a high computing power system to support the simultaneous operation of multiple LiDARs and data processing.
In RS-Cube 2.0 application, customers can quickly and seamlessly embed the LiDAR perception module into their own solution, no cumbersome configuration required, which meaningfully realizes “one tap access LiDAR environment perception ability” and accelerates customers' autonomous or smart transportation projects.