Introduction  

With the rapid development of autonomous driving technology, functions based on high-precision maps have been initially applied and continue to evolve. During the R&D testing phase, various factors may prevent testers from obtaining the corresponding *OpenDRIVE file of the high-precision map within the controller, leading to a mismatch between the simulation scenario and the controller's internal high-precision map. This mismatch restricts autonomous driving functionality and hinders effective testing and validation. Additionally, real-world road testing is risky, costly, time-consuming, and lacks comprehensive coverage, making it difficult to ensure the reliability and safety of autonomous driving systems.

To address this core technical challenge, this article proposes a high-precision map simulation solution based on external data injection. This approach enhances testing effectiveness during the R&D phase of autonomous driving functions, significantly reduces the risks and costs associated with real-world road testing, accelerates the development and deployment of autonomous driving functions, and ensures the reliability and safety of autonomous driving systems.

What is a High-Precision Map?

A high-precision map is often misunderstood as a map with higher accuracy than a standard navigation map, but the two are fundamentally different.

Figure 1: Left - Navigation Electronic Map, Right - High-Precision Map

Navigation maps (SD MAP) are electronic maps designed for drivers, mainly storing road-level element information, such as road shapes, with an accuracy of approximately 5m-10m.

High-precision maps (HD MAP) are specifically designed for autonomous vehicles, providing precise (centimeter-level accuracy) vehicle location information and detailed road element data, such as traffic signs, lane markings, traffic lights, road slopes, and curvatures. These maps are critical for implementing autonomous driving functions and are mainly reflected in the following aspects:

- Precise Positioning: High-precision maps offer accurate road information and landmark data, aiding vehicles in achieving precise positioning, including current vehicle location.

- Environmental Perception: Based on high-precision map data, vehicles can more accurately perceive the surrounding environment, including road signs, traffic signals, and obstacles, thus improving traffic scene perception.

- Path Planning: High-precision maps provide autonomous driving systems with detailed road network data, helping the system plan optimal routes, avoid congested areas, and optimize driving paths.

- Decision Making: Based on high-precision maps and environmental perception data, autonomous driving systems can make intelligent decisions, such as avoiding obstacles, adjusting speed, and safely changing lanes, ensuring driving safety.

High-Precision Map Simulation Solution

When setting up a virtual simulation environment for testing and validating autonomous driving functions, it is crucial to ensure that the simulation scenario map and the internal high-precision map within the autonomous driving controller are perfectly aligned (lane-level accuracy: error less than 1cm). This alignment improves vehicle positioning accuracy, ensures the sensor model's road element information matches the high-precision map, and guarantees accurate results from the controller's fusion positioning and environmental perception. These factors provide precise input for path planning and decision-making, thus enhancing the effectiveness of autonomous driving function testing during the R&D phase.

Different testing backgrounds and requirements correspond to different simulation solutions:

- Solution 1: When testers have access to the OpenDRIVE file corresponding to the controller's internal high-precision map, a simulation based on the OpenDRIVE file can be used. This method is simple and effective.

- Solution 2: For situations where the corresponding OpenDRIVE file is not available (Solution 1 fails), Polelink has innovatively proposed a high-precision map simulation solution based on external data injection.

Solution 1: Simulation Based on OpenDRIVE File Positioning

When testers have access to the OpenDRIVE file corresponding to the controller's internal high-precision map, the OpenDRIVE file can be directly imported into the scene simulation software to automatically generate information such as traffic signs, lane markings, traffic lights, road slopes, and curvatures from the high-precision map. The simulation process is illustrated in the following figure:

Figure 2: Principle of Positioning Simulation Based on OpenDRIVE Files

1. VTD (scene simulation software) can output precise vehicle location information (UTM coordinate system) and transmit it to CANoe (test management software) via Ethernet;

2. CANoe (test management software) processes the vehicle's original coordinates for offset correction (coordinate conversion), i.e., UTM coordinate system -> WGS84 coordinate system -> GCJ02 coordinate system, and sends it to the controller via Ethernet/CANFD.

Solution 2: High-Precision Map Simulation Based on External Data Injection

When testers cannot obtain the OpenDRIVE file corresponding to the controller's internal high-precision map, it is impossible to ensure that the simulation scenario map matches the controller's internal high-precision map, limiting autonomous driving functionality. Therefore, for autonomous driving simulation testing, a high-precision map simulation solution based on external data injection is proposed to align the map data within the controller with the road element information in the simulation scenario, as shown in the following figure:

Figure 3: Principle of High-Precision Map Simulation Based on External Data Injection

Block the internal map module components of the controller, intercept data transmission between the internal map module and the fusion control module components, and replace it with an external self-made map module.

Analyze the map module output interface protocol (communication protocol, data encapsulation protocol), and develop middleware (consistent with the internal middleware of the controller, such as DDS) and the application MapComponent (map component) based on the Linux environment to establish communication with the internal components of the controller.

Develop a MapPlugin (high-precision map plugin) based on VTD to parse and extract road element information from OpenDRIVE and write it into shared memory.

MapComponent reads the road element information written into shared memory by MapPlugin, encapsulates the data according to the interface protocol, and sends it to the internal controller via middleware (e.g., DDS).

Principle of MapPlugin Development

Figure 4: Code Framework of MapPlugin Development

As shown in the figure above:

- The MapPlugin primarily extracts the ego vehicle's motion state, location, and other information through the data preprocessing module.

- On this basis, a map parsing module is developed to extract lane information, lane marking information, road boundary information, and traffic light information around the ego vehicle.

- Compile the MapPlugin code into a Plugin (.so file) and load and run it in VTD.

The road element information in OpenDrive is extensive, and the specific data to be extracted depends on the output interface protocol of the internal high-precision map component of the controller, generally including two types of data: static map data (such as lane markings, lane information, etc.) and dynamic map data (such as the distance from the ego vehicle to ramps, toll booths, etc.).

Principle of MapComponent Development

Figure 5: Principle Framework of MapComponent

As shown in the figure above:

- The MapComponent includes two parts: Shared Memory Data Reading (SHMRead) and Middleware Data Encapsulation and Transmission (MapAgent, such as DDS).

- SHMRead: Reads road element information written into shared memory by MapPlugin, processes it a second time, and assigns values to the interface signals in Proto.

- MapAgent: Encapsulates Proto interface data, serializes it, and publishes it (the development of MapAgent needs to be adapted to the internal environment framework of the controller).

- Map.yaml: Application program configuration, including communication ports, task enablement, interface data management, etc.

High-Precision Map Simulation Case  

The following is a simulation test case of high-precision map simulation based on external data injection, combined with other sensor simulations (e.g., LiDAR, Camera, Radar, USS), to complete the high-speed pilot function (automatically exiting the highway, entering ramps).

Figure 6: Left - VTD Simulation Scenario, Right - High-Precision Map Injected into the Controller

As shown in the figure above:

- The left image is the VTD simulation scenario (OpenDRIVE format map), and the right image is the high-precision map generated after MapPlugin parses and extracts the road element information from the VTD simulation scenario and injects it into the controller via MapComponent. In the right image, the green solid line represents the road boundary, the blue solid line represents the lane marking, the black solid line represents the recommended navigation path, and the pink solid line represents the non-recommended route.

- The ego vehicle drives on the highway, with the navigation endpoint set at the highway exit. After the pilot function is activated, the vehicle will automatically overtake, avoid obstacles, change lanes, and enter the ramp. Upon reaching the vicinity of the destination, the vehicle requests the driver to take over, completing the pilot task.

Summary  

As one of the key technologies in autonomous driving, high-precision maps play a crucial role during the product development phase. The inability to obtain the corresponding OpenDRIVE file for the internal high-precision map within the controller has hindered the testing and validation of autonomous driving functions to a certain extent. Polelink's high-precision map simulation solution based on external data injection effectively addresses this technical challenge, accelerating the R&D and testing of autonomous driving functions.

High-precision map technology has accelerated the deployment of autonomous driving functions. However, the real-time performance, data accuracy, coverage, privacy, and security of high-precision maps also limit the promotion of autonomous driving to some extent. Pure visual autonomous driving solutions, such as those represented by Tesla, have demonstrated a certain level of competitiveness. With the development and breakthroughs in new technologies, autonomous driving solutions are continually iterating, bringing autonomous driving functions closer to reality.

As a technical partner of Vector, Polelink covers autonomous driving MiL/SiL/HiL/ViL testing, V2X testing, and sensor perception testing, providing customers with high-quality autonomous driving testing solutions, system integration testing, and testing services, supporting the R&D, testing, and rapid iteration of autonomous driving products.

*OpenDRIVE is an open standard file format used to describe road networks and traffic environments. It provides a unified way to describe elements such as road geometry, lane information, traffic signs, and traffic lights, and can be used to create high-precision maps for autonomous driving systems.