Shulou(Shulou.com)11/24 Report--

Recently, Chaochao Information and Baidu Intelligent Cloud jointly released the first roadside edge computing unit RSCU, which is currently being deployed in the first autopilot L2~L4 test section to be built in Beijing. RSCU is a roadside edge server with open architecture design, supporting mainstream x86 processors and two leading AI acceleration cards, integrated network convergence, nanosecond clock synchronization and wireless interconnection. Through Baidu Zhilu OS operating system, it can use video, point cloud and other raw data collected by roadside sensor devices to realize all-weather roadside traffic scene perception and real-time intelligent guidance to self-driving vehicles.

Facing the vehicle-road collaborative scene of high-level self-driving at urban intersections, this product has strong roadside edge computing power, can support the data calculation of two-way 8-lane intersections, and realizes the real-time detection and analysis of all elements of vehicles, roads, environment and traffic events in the intersection range. The position accuracy and speed accuracy are higher than the general level of the industry. The average end-to-end delay perceived by roadside objects is ≤ 300ms. It is lower than the general level of the industry, and can better meet the needs of high-precision and low-delay calculation of massive roadside data in high-level autopilot, digital twin roads, intelligent transportation and other scenarios.

In order to adapt to the harsh working environment of the intersection, RSCU also adopts advanced isolation heat dissipation design, which can adapt to the ambient temperature of-25 to 55 ℃ under the condition of high performance and high power consumption, and has the extreme environmental adaptability of dustproof, waterproof, lightning protection and corrosion protection. In addition, the first generation of RSCU can be equipped with Baidu's open and compatible Zhilu OS operating system, and through the cooperation of software and hardware, it can better carry upper-layer applications, realize the coordination of vehicles, roads and clouds, effectively reduce uncivilized driving behaviors such as running red lights, illegal parking and retrograde, and greatly improve traffic safety.

The end of vehicle-road coordination is intelligent transportation and smart city.

In recent years, under the dual promotion of policy and market, the landing of autopilot technology is accelerated, and the basic supporting of industrial chain and market development are becoming more and more mature. According to data from the Ministry of Industry and Information Technology, in the first half of 2023, the sales of new passenger cars with combined driving assistance functions reached 42.4 percent, and the construction of intelligent road reconstruction and cloud control basic platform was accelerated, leading to the improvement of traffic and social operation efficiency.

Bicycle intelligence and vehicle-road coordination are two mainstream autopilot technology routes at present. In recent years, with the comprehensive breakthrough of computing power, algorithm and data, the target detection accuracy of bicycle intelligence has been improved rapidly, but due to the limitation of cost and on-board computing platform, it is still difficult to achieve global perception without dead angles. the security hidden danger caused by blind area still exists. On the other hand, vehicle-road coordination can gain insight into the blind areas of intelligent bicycles in advance with the help of the "God's perspective" of roadside perception, so as to make autopilot more safe and efficient.

More importantly, vehicle-road coordination has gone beyond the "original intention" of autopilot and become an important technical solution for intelligent transportation. In a sense, the starting point of vehicle-road coordination is autopilot, and the end may be intelligent transportation, smart city and intelligent society.

Vehicle-road coordination puts forward new requirements for edge computing facilities.

Vehicle-road collaboration not only requires that the vehicle itself has strong on-board computing power, high-precision sensors, operating systems, etc., but also makes the roadside have the ability of perception, calculation and communication, and can achieve multi-level cloud-edge coordination with edge cloud and data center cloud. Higher requirements are put forward for the design of roadside edge computing infrastructure.

Sun Bo, general manager of Tide Information Edge Computing products Department, believes that the design of roadside edge computing infrastructure should fully consider three elements: environmental adaptability, computing performance and edge security. "first of all, edge computing nodes are widely distributed, and the environment is quite different. there are more stringent requirements on the system, such as heat dissipation, dustproof, wide temperature, anti-electromagnetic interference and so on. Second, edge computing will not deal with a single task in the future, but will be combined with urban intelligence and will participate in systematic decision analysis, requiring more computing power. Finally, edge computing nodes generally lack various hardware protection mechanisms of data center servers, lack of firewall isolation protection, and are easy to become targets for intruders, so we must consider the security of specific hardware devices, network environment and applications to ensure device security, data security and operation and maintenance security. "

Roadside edge computing unit: to provide computational support for high-level autopilot and intelligent transportation

Combining the advantages of both sides in infrastructure, operating system, algorithm and application, Chaochao Information and Baidu jointly released the first roadside edge computing unit, which has six characteristics of high performance, intelligence, openness, compatibility, cooperation and security. build a soft and hard highly collaborative traffic computing base to comprehensively improve the efficiency of vehicle-road-cloud coordination. The product is oriented to L2 to L4 high-grade self-driving car road coordination scene, real-time computing support two-way 8-lane intersection signal lights, lidar, road signs, weather stations and other data comprehensive perception, at the same time support Baidu open and compatible Zhilu OS operating system, with "smart road" to enable "smart car", to achieve car-road-cloud collaborative intelligent transportation services.

Wang Miao, chief architect of Baidu car-Road Synergy, said: "the roadside computing unit integrates all the sensor data of the road section, while providing a computing platform to realize the digitization of the roadside. The digitization and intelligence of the roadside will be extended from vehicles to traffic management, and even with urban governance and urban big data operation to form an overall data aggregation to achieve 'software-defined traffic'."

According to reports, the "perception-computing-communication" roadside edge intelligent system based on the vehicle-road collaborative core computing unit is being deployed and tested in the Beijing High-level self-driving demonstration Zone, and will provide services for the establishment of Beijing city-level "vehicle-road-cloud integration" demonstration application zone in the future. The test data show that the roadside edge intelligent system can realize the real-time detection and analysis of all elements of vehicles, roads, environment and traffic events in the intersection range, with location accuracy ≤ 1.0m (man-machine non-average), speed accuracy ≤ 1.5m/s (mean), traffic object perceived location type identification accuracy ≥ 90%, roadside object perceived end-to-end delay (including communication delay) ≤ 300ms (mean).

Sun Bo believes that the project results and outputs of joint innovation with Baidu can be applied to more edge computing scenarios, such as water conservancy, manufacturing, mining, power grid, and so on, quickly replicating from a single point to other industries to accelerate the industrialization of edge computing.

With the further integration of edge computing technology and terminal devices, and cloud computing, big data, artificial intelligence and other technologies, edge computing infrastructure will play a role in all aspects of people's production and life, such as industrial Internet, car networking, smart city and wisdom, so that computing can reach the boundary of digital perception. E2X (Edge to everything) will become a "super neural network" in the future digital fusion world. It is necessary to strengthen industrial collaborative innovation and unleash the real potential of edge computing technology.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.