Application and simulation of ultrasonic radar in APA automatic parking function

The automatic parking assist system (APA) is an important part of modern automobile intelligent driving technology. The system enables the vehicle to automatically complete the parking process without any human intervention. Specifically, during the parking process, the APA system uses sensors (surround view cameras, ultrasonic radars, etc.) arranged around the vehicle to sense and identify valid parking spaces and obstacles, and then plans the parking path and controls the vehicle's movement to complete the parking task. This article will focus on the application of ultrasonic radar in the APA automatic parking function and the simulation method in the HiL test environment.

Application of ultrasonic radar in APA automatic parking
In the application scenario of automatic parking, it is generally necessary to arrange 12 ultrasonic radars around the vehicle to complete the function of fully automatic parking. Compared with millimeter wave radar or other forms of radar, ultrasonic radar has many advantages such as low manufacturing cost, easy installation, and easy maintenance in the later stage.

Ultrasonic radars are mainly divided into two types. One is a short-range radar installed on the front and rear bumpers to detect obstacles. The detection distance is generally 15~250cm. This type of sensor is called a PDC sensor. The other type is a sensor installed on the side of the vehicle to detect the length of the parking space. The detection distance is generally 30~500cm. This type of sensor is called a PLA sensor. The main distribution of PDC and PLA is shown in the figure below.

In actual automatic parking application scenarios, ultrasound has multiple working modes, such as the Direct Echoes mode and the complex Cross Echoes mode. In the Direct Echoes mode, the ultrasonic radar can calculate the distance between the vehicle and the obstacle through the time of flight (TOF) of the sound wave. The principle of this mode is relatively simple, but it cannot obtain the two-dimensional coordinates of the obstacle, that is, it cannot obtain the spatial position of the obstacle relative to the main vehicle. The Cross Echoes mode uses multiple ultrasonic radars as receivers of sound waves to better obtain the spatial position of the obstacle relative to the main vehicle, but the calculation will also be more complicated.

Ultrasonic radar ranging principle and APA parking process

When the vehicle is parking, the ultrasonic radar can calculate and output the distance between the vehicle and the surrounding obstacles in real time. The controller software fits the outline, shape, relative position, etc. of the surrounding obstacles by processing the ultrasonic data. Taking the parking scene of a horizontal parking space as an example, the driver activates the vehicle's APA automatic parking function, and the vehicle keeps moving at a low speed. Combined with the ultrasonic perception fusion data, the APA system will identify the parking space that can be parked. After the driver enables it, the vehicle will enter the next necessary steps such as path planning and vehicle control until the vehicle successfully parks in the parking space and exits the APA automatic parking function. The following figure shows the entire APA automatic parking process.

Basic simulation method of ultrasonic radar After understanding the basic working principle of ultrasonic radar and vehicle layout, some important parameters that need to be paid attention to are:

Measuring range: the maximum detection distance of ultrasonic radar;

FOV: horizontal viewing angle range and vertical viewing angle range of ultrasonic radar.

Radar operating frequency: The operating frequency affects the diffusion of ultrasonic waves, background noise, and reflection loss. Generally, the operating frequency of ultrasonic radar is around 40kHz.

In the HiL simulation environment, the simulation of ultrasonic radar requires the help of professional scene simulation software (VTD) and bus simulation experiment management software (CANoe). In VTD, the various parameters of the simulated ultrasonic radar need to be correctly configured, including the radar's vehicle installation position, the ultrasonic radar's maximum detection distance, FOV, etc. In the bus simulation experiment management software, the communication protocol, transmission frequency, baud rate, etc. used by the ultrasonic wave need to be clearly defined.

In the entire simulation link, VTD uses the built-in perfect sensor or other custom development models with higher accuracy to package the closest distance to obstacles detected by each ultrasonic radar into UDP and send it to CANoe for data analysis and bus simulation. As shown in the figure below, VTD sends 12 groups of UDP data output by radar in real time. In this way, the transmission link can be effectively simplified and the performance of the entire simulation can be improved. Of course, according to different simulation requirements, the structure of the packaged data can be modified to output the three-dimensional coordinates, ID, attributes, etc. of the obstacle. By obtaining different data and combining CANoe, more types of APA automatic parking tests and verifications can be performed.

CANoe supports multiple bus simulation capabilities. Combined with Vector hardware such as VN1640, VN1670, VN5650 and other sensor communication devices, it can provide the controller with multiple protocol support such as CAN/CAN FD, TCP/UDP, SOME/IP, DSI3, etc. At the same time, the software provides a VN hardware configuration window so that users can easily manage the VN interface box in the simulation system.

Beihui Information's APA simulation function technical solution

In the delivered projects, Northlink Information has implemented a variety of ultrasonic simulation test solutions. For the ultrasonic radar simulation of SOME/IP, CAN/CAN FD bus, CANoe combined with VN1670, VN5650 or other communication interface boxes can provide the controller with a bus simulation platform with higher accuracy, more stable message sending cycle and rich fault injection types.

The DSI3 bus protocol has the advantages of master-slave one-to-many asynchronous single-line current and voltage communication, automatic ID allocation, asynchronous communication, low cost, strong anti-interference, and support for multiple data format transmission. It is very suitable for application scenarios such as automotive functional safety.

After CANoe and VTD are combined, the simulated obstacle distance detected by ultrasonic detection is output to CANoe. CANoe processes the distance data into the required flight time data format and sends it to the DSI3 communication device. The DSI3 communication device sends the simulated ultrasonic data to the controller under test through signal conversion, completing the simulation of 12-channel ultrasonic data of the controller.

When using perfect sensors to detect obstacles around the vehicle in VTD, there will usually be a certain distance error due to the perfect sensor's obstacle detection characteristics. The simulation solution of the perfect sensor cannot respond to the distance at the millimeter level. In order to compensate for this part of the error, more additional algorithms are needed to correct the error. This not only takes up more computing resources, but the results after processing may not be able to perfectly correct the error. See the figure below.

Beihui Information provides a set of ultrasonic models that use the principle of ray tracing for distance measurement. Compared with the use of 12 perfect sensors, the ray tracing sensor model can provide millimeter or even micron level distance detection, and integrate the 12 sensor models into a ray tracing ultrasonic model, which greatly simplifies the sensor layout of the project and improves the efficiency of simulation. The operating environment of the model requires the support of the NVIDIA CUDA environment, and supports the configuration of the ultrasonic radar's maximum detection distance, FOV, installation location, simulation frequency, etc. The following is a ray tracing principle diagram.

In summary,
the APA automatic parking function is an important part of intelligent driving. When dealing with increasingly complex parking environments, HiL testing can accelerate APA function development and test verification through virtual simulation, reduce the cost of APA real vehicle testing and verification, and verify the test coverage of the APA function by building more complex and rich parking scenarios.