Tencent Vehicle-Road Co-operation System relies on road-side unit (RSU) , road sensing equipment (such as camera) and 5G base station to realize the information closed-loop among the connected vehicles. In addition, the edge cloud computing of the system is connected to central csloud (traditional data center) through external network lines (such as 4G/5G, CPE or carriers’ special lines), which achieves large-scale network coverage with 5G base station, makes up the lack of network coverage in roadside infrastructure, and realize the integration of vehicle-road-cloud-network by using the powerful computing power of the central cloud (such as traffic brain).
Edge cloud computing infrastructure can improve data processing capacity of existing roadside infrastructure. With the deployment of AI processing workstation, the edge cloud infrastructure can analyze each target in road-side camera (include vehicles, pedestrians, scatters, traffic accidents, etc.). The main functions are as follows:
Target recognition and detection: recognize and detect road targets (including vehicles, pedestrians, scatters, traffic accidents, etc.).
Target tracking: track the road targets accurately. Unique recognize and exclude repetitive tracking when multiple cameras capture the same target.
Location & velocity & direction: locate the target in real time and compute the speed and direction of it.
GNSS high precision positioning: based on the original GNSS position measured by the vehicle, the edge cloud can realize high precision positioning without special RTK terminal. Besides, the information can be delivered to connected vehicles and other modules of the edge cloud to judge the main car and abnormal scene. Compared with the method of using GPS as positioning reference, the accuracy is greatly improved without increasing the cost.
Abnormal scene recognition: the recognition scenes include emergency braking, parking, abnormal lane change, left turn, reverse driving, large trucks, vulnerable traffic participants, scatters, Lane level congestion, etc.
Picture 1: Visual recognition of emergency braking scene
Picture 2: Visual recognition of vehicle left turn scene
Picture 3: Visual recognition of lane changing scene
Based on our test in semi closed software Park, municipal traffic road and other scenes, the main technical indicators of the vision module are as follows:
Accuracy of road target detection: the average miss detection rate of the latest visual recognition module is less than 7%;
Accuracy of traffic scene judging: the success rate of judging abnormal traffic scenes and traffic accident scenes specified by CSAE is higher than 90%.
System end to end delay: based on the existing low-cost roadside sensors, the end-to-end delay is less than 300ms.
Open platform ecology
In Tencent Vehicle-Road Co-operation open platform, for edge cloud and central cloud system, we built a vehicle-road co-operation edge computing platform with good ecological openness based on Tencent cloud-edge collaborative system TMEC:
Flexible aggregation of third party system: the platform is completely decoupled from the third-party system, and the original road data collected by roadside infrastructure is directly provided to the third party system by means of senseless forwarding through the programmable data retransmission system. In addition, edge cloud and center cloud connect with third-party systems (such as traffic monitoring center) through specific open data interfaces. Edge cloud can share structured v2x information with third-party systems, while center cloud can provide big data interface to third-party systems, providing big data processing and analysis functions for third-party applications.
Seamless connection to 5G / V2X network: the platform supports the connected vehicles service deployed by edge cloud to seamlessly connect to 5G / V2x network, and can simultaneously adapt v2x messages to 5G base station (Uu Interface) and C-V2X RSU (PC5 interface) transmission, and support for 5G Internet of vehicles and cloud game business under high-speed vehicle scene, using innovative 5G end-to-end slice activation for application and location, AI weak network detection, MEC resource scheduling for Internet business and other technologies, to achieve seamless docking for 5G/V2X. The platform adopts Tencent self-developed micro service framework to realize service-oriented governance of edge cloud business including third-party business. The framework has been adopted by 3GPP SA2 R16 as one of the four main solutions of 5G core network service architecture evolution (eSBA) (the other three are traditional telecom equipment provider solutions), and has been partially adopted by 3GPP standards TS23.501 and TS23.502.
Low-cost platform: due to the openness of the platform, it can make full use of the existing infrastructure and flexibly aggregate the third-party system, so it can greatly reduce the cost for the users. The current system integrates Intel low-cost computing hardware in the edge cloud, which greatly reduces the computing power cost by 50% and reduces the energy consumption of the edge cloud by 60%. In addition, by using cloud GNSS high-precision positioning method, compared with the traditional RTK solution, the cost of connected vehicle terminal can be reduced by about 90% under the condition of slightly reducing the positioning accuracy (but meeting the requirements of lane level positioning).
In the Linux foundation, Tencent has led the establishment of “connected vehicle blueprint”, an open source project of vehicle-road collaboration based on 5G MEC. At present, with the cooperation of Tencent, Intel, Nokia, ARM and other companies, the project is promoting the platform to become an open-source solution for traditional information electromechanical providers in the process of vehicle road collaboration industry landing.
Advancement and innovation
Tencent vehicle road co-operation open-source platform is built based on TMEC, the edge cloud system of Tencent self-developed, and its main innovation points and competitiveness are as follows:
Openness and scalability. The platform has two capacities: (1) seamless docking 5G/V2X network, (2) flexible aggregation of third-party vehicle road collaborative system. For (1), we deployed the 5G / V2X network adaptation component developed by Tencent in the edge cloud, which can connect with the underlying business architecture and flexibly dock with 5G / V2X network, solving the problem of strong binding relationship between business and underlying network in the traditional vehicle road collaboration framework. For (2), the platform adopts the latest container and microservice technology, which can service the built-in V2X function and the third-party business deployed in the edge cloud. Besides, it supports the cooperation between the edge cloud and the center cloud, which realize the flexible cooperation between edge cloud and center cloud. For example, after the training of sensing model in center cloud, the model can be distributed to the edge cloud to realize the automatic deployment and update of the roadside sensing function. The micro service architecture based on container model is convenient for the third-party system deployment. In addition, the service governance function can reduce the operation and maintenance difficulty of the third party business, which ensures that the platform can quickly aggregate the third-party system and realize the function expansion.
Commercialization capability: the platform has built-in ability to touch Tencent to-C system. With the help of Tencent's strong user penetration ability in to-C field, it realizes the V2X information closed-loop between road and travel users, solve the problems of low user penetration and high user access cost commonly faced by the current vehicle road collaboration business, and lay a good foundation for further mining the value of roadside perception information. This feature greatly promotes the landing of commercialization of vehicle road collaboration business, and greatly enriches the commercialization path of vehicle road collaboration. For example:
For customers in the field of road infrastructure, the platform can be used as a part of the intelligent transportation system to support the deployment of roadside intelligent facilities and provide road operators with information services such as road safety and vehicle monitoring. The platform can fully coordinate with the existing system of road operators, effectively protect the existing investment and the implantation of new informatization capability.
For vehicle manufacturers, on the one hand, the platform can integrate with Tencent TAI system to provide an integrated vehicle road collaborative car machine solution; on the other hand, it can dock with third-party car machine system to provide cloud V2X data access service based on our interface.
For clients, the platform can dock with mobile map, and present rich road information on the application to improve the user’s experience. What’s more, V2X information can be directly provided to users in the form of WeChat applet.
Platform fully uses the capacity of roadside infrastructure: in terms of platform investment and operating costs, by making full use of the existing infrastructure and flexibly aggregating the third-party system, the cost can be greatly reduced for platform users. Technically, because of the high position of sensors deployed on roadside infrastructure (such as cameras), compared with the automatic driving system, the target shelter is less, which is conducive to multi-target detection, tracking and scene recognition. In addition, these roadside infrastructure deployment positions are fixed, so the deployment is easy, the location calibration is simple, and the operation management is more convenient.