Project background and business scenarios
At present, the Internet of Vehicles industry in China is in a period of deep integration with 5G technology. With the evolution of automobiles from traditional means of transportation to intelligence, networking and electrification, the Internet of Vehicles has become the most potential application for cross-border integration in the field of 5G transportation and automobiles, and has become an important development direction for strategic emerging industries in mainland China. By building a new Internet of Vehicles technology and industrial collaboration architecture, our company has carried out intelligent research and development with an innovation center and Ali algorithm team, and has now practiced and developed 17+ in-depth collaborative scenario innovation of "perception and early warning" and "pilot service", which will continue to evolve and upgrade in the future, and expand more innovative scenarios, which is of great significance for improving the efficiency and safety of the transportation system and realizing the sustainable development of the transportation system.
A digital twin of a campus (hereinafter referred to as "digital twin") applies roadside sensing infrastructure such as intelligent sensing devices and edge computing equipment, and uses a campus as the base site to complete 100% recognition and coverage of roads in the whole park, synchronize the twin display of people and vehicles, and support backtracking of the real scene when the event occurs, and dynamically present cloud-vehicle-road collaboration. Realistic restoration of test scenarios, efficient use of road mining data to generate simulation scenarios, and large-scale parallel acceleration in the cloud make simulation tests meet the closed-loop of the full stack of vehicle driving perception, decision-making planning and control.
Business pain points
1. An integrated display platform for multiple complex systems and perception devices
In the standard system of vehicle-road collaboration, the perception data collected by on-board sensors and roadside sensors is integrated, processed and released in the cloud service center using wired or wireless mobile communication technology, and then distributed to the twin platform through the data interface for business display. The multi-sensor fusion technology integrates low-latency, high-precision, and low-fault tolerance perception data, integrates various devices with different advantages and disadvantages (cameras, millimeter-wave radars, lidars, etc.), reasonably coordinates multi-source data to the greatest extent, makes full use of relevant information, improves comprehensive utilization, and integrates data indicators in multiple dimensions and levels to form the daily operation and global management system of networked vehicles.
2. It can be used as a vehicle-road synergy verification platform
In the twin platform, the real-time collected data can be synchronously completed by the deduction of the visual model, data-driven decision-making can be realized, and various collaborative scenarios under different traffic conditions can be reproduced in a visual form to verify the communication flow mechanism between the cloud-vehicle-road, so that the simulation platform can meet the closed-loop of Internet of Vehicles perception and decision-making planning control.
3. It can promote the implementation of vehicle-road collaborative innovation business and assist traffic decision-making
At present, the concept of vehicle-road collaboration is still in the early stage of development, and the specific implementation of technology, business model and construction mode is also being continuously explored, and the construction of this platform can start from the vehicle, truly restore the interaction and sharing of various traffic information in the park, and use the vehicle-mounted and roadside sensing equipment to provide real-time feedback on various collaborative disposal events, abnormal events, road condition monitoring and other information, so as to realize the landing of vehicle intelligent collaboration, so that vehicle-road collaboration gradually moves from concept to reality. In the future, it is believed that vehicle-road collaboration technology can also be applied to traffic management, realize efficient scheduling and management of traffic flow, and play a role in traffic supervision and command.
solution
Overall technical architecture
The data terminal of a campus digital twin platform uses the DataFlux Function (f(x)), a data development platform of the observation cloud (www.guance.com), to analyze, store, and process the native interfaces output by the operation control data platform, and generate interfaces that conform to RESTful specifications for DataVision (DataV, CityV) at the display layer to call and display.
The display end of a campus digital twin platform consists of two visualization tools, DataV and CityV, and CityV displays scene data, such as the distribution and online status of devices in the park, traffic flow tracks, and traffic lights at intersections. DataV displays data from a Kanban class, usually in the form of a pop-up box and panels on either side. Both tools accept the interface data output from f(x) and present it. Communication between the two tools is achieved via websockets.
The display end of VOC is composed of two visualization tools, DataV and CityV, CityV displays scene-based data, such as the passenger flow and vehicle flow of various venues in the Asian Games Village. DataV displays data from a Kanban class, usually in the form of a pop-up box and panels on either side. Both tools accept the interface data output from f(x) and present it. Communication between the two tools is achieved via websockets.
VOC platform
In the VOC platform, a real-time, high-fidelity, and infinitely close digital twin of the physical space is built in the digital space, and the perception, analysis and calculation, and information feedback of the Internet of Things in the real world are accurately mapped in the virtual space, so as to realize a digital twin that can be observed, analyzed, warned, and interacted.
Platform process
The data in the digital twin platform of a campus is collected in the back-end service through APIs, and after data processing and calculation, it is reported to the front-end display of the twin. The development tools used in the front-end of the twin, Datav and CityV, present the data in a visual way.
Annotation:
- The back-end service uses the Python programming language and the lightweight web framework Flask to start the service, and the deployment is monitored by the supervisor process management tool.
- DataV is a data visualization application building tool from Alibaba Cloud that helps users who need to display data quickly complete the construction of visualization applications.
- CityV is a UE5-based twin that quickly generates large-scale 3D models of cities, combined with handcrafting, to create more detailed and realistic models for specific buildings and scenes.
data processing
The back-end data processing adopts the DataFluxf(x2) data processing platform. The service completes operations such as data access, processing, and forwarding. It improves the efficiency of data docking and the ability to process massive data.
The backend has access to a total of 3 systems, 63 HTTP service interfaces, and opens up 68 authorization links for front-end data display. The Ali algorithm team integrates the device perception data and aggregates it into the trajectory data, and the backend collects and summarizes the trajectory data into the data format required for the cockpit. Data access and processing through the use of API interfaces, automation and real-time, suitable for data interaction with external systems. It can realize automatic synchronization and real-time update of data, which is convenient for real-time interaction with other systems.
Featured Highlights:
1. Full-stack capacity building for deep integration of vehicles and roads
The digital twin technology is used to present the real-time presentation of vehicle-road collaboration, and the voice broadcast in the real scene is transformed into a visual 3D scene. From the perception of the road test equipment, the data is uploaded to the cloud control decision-making, and the decision-making results are issued to the vehicle end for presentation, and at the same time, the front vehicle warns and the rear vehicle is jointly controlled to complete the cloud vehicle road-end link. In the overall link of the cloud vehicle-road, the road test equipment, connected bicycles, and vehicle-road scenarios are completely monitored. The vehicle no longer has to rely on the driver's perception and action to obtain information to take corresponding measures, managers can comprehensively and timely perceive the traffic flow status and predict potential dangers through the twin platform, collect scene test data, debug and analyze based on test data, and finally the vehicle road can be highly automated driving, and optimize the operation status of the vehicle based on the perception information to ensure driving safety and improve operation efficiency.
2. Multi-device perception monitoring
High-precision restoration is carried out for the campus, and the data collection of the campus is carried out through multi-mode device perception, and the real data drives the change of the running state of the scene. Based on the multi-sensor fusion technology, the intelligent combination of multiple sensor data is carried out by the intelligent light pole in the park, which solves the problems of all-day, all-weather and accurate perception, and provides the system with low-latency, high-precision and fault-tolerant perception results. The integration of roadside sensors with many different sensing devices can maximize the reasonable coordination of multi-source data, make full use of useful information, and improve the comprehensive utilization rate of information. And the refined management of all sensing devices is carried out, with the light pole as the main body, the perception status of each device is displayed, and the device perception coverage is comprehensively presented, which provides a basis for the vehicle-road collaborative scenario test.
3. Multi-vehicle road collaborative scene library
In view of the problems of strong accident accidentality, lack of traffic accident scene library, insufficient data coverage, high cost of scene construction, and low efficiency, based on the multi-modal data of vehicle-road-cloud intersection, the holographic restoration of the scene and the overall expansion of the scene library were carried out. Through the display and playback of 17+ vehicle-road collaborative scenes, historical scenes are explored and analyzed, and the application of scene libraries under vehicle-road collaboration is realized to improve road traffic safety. Car companies can improve the safety and overall competitiveness of intelligent driving vehicles, and traffic management departments can use the information from the traffic accident scenario library to assist in rapid decision-making and improve the development of traffic safety management technology, thereby accelerating the development of intelligent networked vehicles.
4. Real-time trajectory simulation
Based on the science and technology innovation park, the trajectory of the park and the roadside road is restored, and the multi-device IoT perception data is extracted and restored, converted into structured data for real-time display on the large screen, simulating the real-time traffic flow, and combining roadside information such as light status and stop line information, as well as driving data information such as vehicle speed and early warning, to remotely monitor and test the road status. At the same time, it can switch between real-time and historical trajectories, select the corresponding time period for the real presentation of the trajectory according to the previous vehicle road test data, and display the vehicle road conditions in the current state at the same time, which is convenient for auxiliary decision-making of vehicle road simulation tests.
5. The full status of the networked vehicle is presented
For test connected vehicles, complete status restoration and monitoring are carried out. Macro display of the overall triggering situation of the connected vehicle displays the status of the connected vehicle during the operation of the vehicle, and analyzes the recent triggering scenarios of the vehicle road. Real-time display of the current real-time operation of connected vehicles, real-time analysis and display of data of triggered scenes. Multi-view intelligent car following, synchronously display the current status of the connected vehicle and the details of the trigger scene, and link the nearest camera of the vehicle in real time for video surveillance, and present the multi-dimensional status of the connected vehicle in the digital twin to ensure the safety of the connected vehicle test scene. In the historical scenario, the online time period and trigger scene of the vehicle and the road are counted, and a certain time or scene is played back separately according to business needs, and the safety and stability of intelligent driving are improved based on the information interaction of vehicles, vehicles and vehicles, so as to accelerate the safer landing and promotion of intelligent driving.
6. Holographic parking guidance
Integrate the multi-source perception data of the park, associate the smart parking scene, analyze the scene combination with intelligent algorithms, assist the operation to realize the real-time informatization, digitization and visual upgrade of parking spaces, automatically plan parking guidance routes, and restore real-time parking path animation. Replicate the location of the parking spaces in the parking lot of the park, combine the induction of the parking lot and other Internet of Things devices to restore the parking status of vehicles in the parking lot, provide information support for the management of the parking lot, reduce management pressure and operation and maintenance costs, and bring more convenient, efficient and high-level management capabilities and image improvement.