CN118627201A - A sensor simulation modeling method and device for intelligent driving controller - Google Patents

Abstract

The present invention discloses a sensor simulation modeling method and device for an intelligent driving controller, and the sensor simulation modeling method includes the following steps: S1, firstly, reconstructing the autonomous driving scene in the digital world from the collected data, including cars, pedestrians, roads, buildings and traffic signs, controlling the reconstructed scene to simulate, and generating some rare key scenes; S2, simulation modeling of ideal environment sensors in a virtual environment, including static background, dynamic objects and objects outside the area; S3, simulation modeling of physical perception sensors in a virtual environment, including camera simulation, laser radar simulation, and millimeter wave radar simulation; S4, simulation modeling of hybrid sensors in a virtual environment, taking target list/environmental parameters as input and target list/sensor raw data as output; S5, simulation verification of sensor function model. The sensor simulation modeling of the present application can accurately simulate the real world and realize the free interaction effect between the unmanned vehicle and the environment.

Description

A sensor simulation modeling method and device for intelligent driving controller

Technical Field

The present invention relates to the field of intelligent driving technology, and in particular to a sensor simulation modeling method and device for an intelligent driving controller.

Background Art

In recent years, with the rapid development of autonomous driving technology, unmanned vehicles can perform well in most common scenarios. However, current technology still cannot guarantee safe deployment. The reason is that there are many safety-critical scenarios in the real world and these boundaries are crucial. Simulation testing has become an effective means to help researchers generate a large number of boundary scenarios in a low-cost manner, thereby comprehensively testing and training existing autonomous driving models. Since unmanned vehicles perceive the real world by equipping various sensors, real and scalable sensor simulation has become an important part of the entire simulation system.

Self-driving cars perceive the surrounding environment through sensors and transmit the information to the path planning stage to determine and predict future behavior and control the vehicle's lateral and longitudinal drive systems accordingly. In order to allow self-driving cars to drive smoothly on real roads and improve the accuracy of perception, multi-sensor perception and fusion technology has been widely used, laying the foundation for accurately understanding the surrounding environment and transmitting data accordingly.

One of the main reasons that hinder the popularization of autonomous driving is that safety is still insufficient and the real world is too complex, especially with the long tail effect. Boundary scenarios are crucial for safe driving, they are diverse, but difficult to encounter. It is very difficult to test the performance of autonomous driving systems in these scenarios because these scenarios are difficult to encounter and it is very expensive and dangerous to test them in the real world.

In order to solve this challenge, both industry and academia have begun to pay attention to the development of simulation systems. Many types of sensors are used in self-driving cars, such as cameras, radars, lidars, ultrasonic waves, etc., which can be used for a variety of different purposes and detection ranges. The actual self-driving system needs to achieve high-reliability environmental perception through sensor fusion technology in order to effectively complete path planning and behavior prediction. However, the sensors of the intelligent driving controller in the existing technology are difficult to accurately simulate the real world, and the free interaction between the unmanned vehicle and the environment is poor.

Summary of the invention

In order to solve the above problems, the present invention provides a sensor simulation modeling method and device for an intelligent driving controller to solve the problems raised in the above background technology.

The above technical objectives of the present invention are achieved through the following technical solutions: A sensor simulation modeling of an intelligent driving controller comprises the following steps:

S1. First, reconstruct the autonomous driving scene in the digital world from the collected data, including cars, pedestrians, roads, buildings and traffic signs, control the reconstructed scene for simulation, and generate some rare key scenes;

S2, simulation modeling of ideal environmental sensors in virtual environments, including static background, dynamic objects and objects outside the area;

S3, simulation modeling of physical perception sensors in virtual environments, including camera simulation, lidar simulation, and millimeter-wave radar simulation;

S4, hybrid sensor simulation modeling in virtual environment, taking target list/environmental parameters as input and target list/sensor raw data as output;

S5. Simulation verification of sensor functional model.

Further, in step S1, the functional modules of the sensor functional model framework are composed of object extraction, occlusion simulation, output simulation and error simulation, and the object extraction quickly extracts the sensed object from the simulation scene according to the position, sensing range and sensing object type of the sensor;

Occlusion simulation, based on the geometric outlines of objects and their relative position to the sensor, determines the final visible objects according to the graphic geometry;

Output simulation, based on the characteristics of the sensor itself and the output format of the perceived object, calculates the ideal output data for the visible object that needs to be output externally, and obtains the ideal output object;

Error simulation characterizes the sensor characteristics and adds Gaussian white noise to the output data of the ideal output object to form the final perception output.

Furthermore, the static background includes buildings, roads and traffic signs; the dynamic objects include pedestrians and cars; and the objects outside the area include the sky and roads.

Furthermore, in step S2, when modeling the dynamic object, the network is used to generate corresponding object features and share shape information, thereby generating a complete car shape.

Furthermore, in step S3, the camera simulation generates realistic images based on the geometric space of environmental objects, and then adds color and optical properties to the three-dimensional model through computer graphics according to the real material and texture of the object to simulate image synthesis; the lidar simulation refers to the scanning method of the real lidar, simulates each ray emitted and received by the lidar, and also intersects the emitted rays with all objects in the scene; the millimeter-wave radar simulation emits a series of virtual continuous frequency-modulated millimeter waves in different directions according to the field of view and resolution information of the radar configured for the test vehicle, and receives the reflected signal of the target.

Furthermore, in step S4, the hybrid sensor considers factors such as noise and target object properties on the basis of the ideal sensor, and adds an error signal to the data information of the ideal output object; performs partial physical level modeling on the sensor working link, and outputs the sensor raw data.

Furthermore, in step S5, the detection performance of the millimeter-wave radar sensor is simulated according to the physical mechanism, and the accuracy of the millimeter-wave radar sensor model is verified by fitting the millimeter-wave radar sensor model and the actual millimeter-wave radar measurement data and using statistical methods to compare and analyze the distance error, speed error and angle error.

The present application also discloses a sensor simulation modeling device for an intelligent driving controller, which adopts the above-mentioned simulation modeling method.

Compared with the prior art, the present invention has the following beneficial effects:

1. This application can accurately simulate the real world and achieve the effect of free interaction between the unmanned vehicle and the environment;

2. This application establishes a high-efficiency sensor function model that can be applied to concurrent real-time simulation of multiple intelligent vehicles in a virtual environment, can simulate certain physical phenomena, and can support real-time simulation of complex test scenarios in which multiple intelligent vehicles participate simultaneously;

3. This application simulates the detection performance of the millimeter-wave radar sensor based on the physical mechanism. By fitting the millimeter-wave radar sensor model and the actual millimeter-wave radar measurement data, a statistical method is used to compare and analyze the distance error, speed error and angle error, thereby verifying the accuracy of the millimeter-wave radar sensor model.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a flow chart of a sensor simulation modeling method of an intelligent driving controller in the present invention.

DETAILED DESCRIPTION

The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

Please refer to FIG1 . The present invention provides a technical solution: a sensor simulation modeling method for an intelligent driving controller, comprising the following steps:

S1. First, reconstruct the autonomous driving scene in the digital world from the collected data, including cars, pedestrians, roads, buildings and traffic signs, control the reconstructed scene for simulation, and generate some rare key scenes;

Among them, the functional modules of the sensor function model framework are composed of object extraction, occlusion simulation, output simulation and error simulation. Object extraction quickly extracts the perceived object from the simulation scene according to the sensor's position, perception range and perceived object type;

Occlusion simulation, based on the geometric outlines of objects and their relative position to the sensor, determines the final visible objects according to the graphic geometry;

Output simulation, based on the characteristics of the sensor itself and the output format of the perceived object, calculates the ideal output data for the visible object that needs to be output externally, and obtains the ideal output object;

Error simulation, characterizing the sensor characteristics, adding Gaussian white noise to the output data of the ideal output object to form the final perception output;

S2, simulation modeling of ideal environmental sensors in virtual environments, including static background, dynamic objects and objects outside the area;

The static background includes buildings, roads and traffic signs; the dynamic objects include pedestrians and cars; the objects outside the area include the sky and roads;

When modeling dynamic objects, the network is used to generate corresponding object features and share shape information to generate the complete car shape;

S3, simulation modeling of physical perception sensors in virtual environments, including camera simulation, lidar simulation, and millimeter-wave radar simulation;

The camera simulation generates realistic images based on the geometric space of environmental objects, and then adds color and optical properties to the three-dimensional model through computer graphics according to the real material and texture of the object to simulate image synthesis; the laser radar simulation refers to the scanning method of the real laser radar, simulates each ray emitted and received by the laser radar, and also intersects the emitted rays with all objects in the scene; the millimeter wave radar simulation emits a series of virtual continuous frequency modulated millimeter waves in different directions according to the field of view and resolution information of the radar configured on the test vehicle, and receives the reflected signal of the target;

S4, hybrid sensor simulation modeling in virtual environment, taking target list/environmental parameters as input and target list/sensor raw data as output;

Among them, the hybrid sensor considers factors such as noise and target properties on the basis of the ideal sensor, adds error signals to the data information of the ideal output object; performs partial physical level modeling on the sensor working link, and outputs the sensor raw data;

S5. Simulation verification of the sensor function model. The detection performance of the millimeter-wave radar sensor is simulated according to the physical mechanism. By fitting the millimeter-wave radar sensor model and the actual millimeter-wave radar measurement data, a statistical method is used to compare and analyze the distance error, speed error and angle error, thus verifying the accuracy of the millimeter-wave radar sensor model.

In the present embodiment, a sensor simulation modeling method of an intelligent driving controller can accurately simulate the real world and achieve the effect of free interaction between the unmanned vehicle and the environment; a high-efficiency sensor function model suitable for concurrent real-time simulation of multiple intelligent vehicles in a virtual environment is established, which can simulate certain physical phenomena and support real-time simulation of complex test scenarios in which multiple intelligent vehicles participate simultaneously; the detection performance of the millimeter-wave radar sensor can be simulated according to the physical mechanism, and the accuracy of the millimeter-wave radar sensor model is verified by fitting the millimeter-wave radar sensor model and the actual millimeter-wave radar measurement data and using statistical methods to compare and analyze the distance error, speed error and angle error.

This embodiment also discloses a sensor simulation modeling device for an intelligent driving controller, which adopts the above-mentioned simulation modeling method.

The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims (8)

1. A sensor simulation modeling method of an intelligent driving controller is characterized by comprising the following steps:

s1, reconstructing an automatic driving scene including automobiles, pedestrians, roads, buildings and traffic signs in the digital world from acquired data, and controlling the reconstructed scene to simulate to generate some rare key scenes;

S2, simulation modeling of an ideal environment sensor in a virtual environment comprises a static background, a dynamic object and an object outside an area;

s3, simulating and modeling a physical perception sensor in a virtual environment, wherein the simulation and modeling comprise camera simulation, laser radar simulation and millimeter wave radar simulation;

S4, simulation modeling of the hybrid sensor in the virtual environment, wherein the target list/environment parameters are used as input, and the original data of the target list/sensor are used as output;

S5, simulation verification of the sensor function model.

2. The sensor simulation modeling method of the intelligent driving controller according to claim 1, wherein the sensor simulation modeling method is characterized in that: in step S1, a functional module of a sensor functional model framework consists of object extraction, shielding simulation, output simulation and error simulation, and the object extraction rapidly extracts a perceived object from a simulation scene according to the position, the perception range and the perceived object type of a sensor;

shielding simulation, namely determining a final visible object according to the geometric outline of the object and the position relation of the object relative to the sensor and the graph geometry;

Output simulation, namely calculating ideal output data for a visible object to be externally output according to the characteristics of the sensor and the output format of a perceived object to obtain an ideal object to be output;

and (3) error simulation, representing the characteristics of the sensor, and adding Gaussian white noise to the output data of the ideal object to be output to form final perception output.

3. The sensor simulation modeling method of the intelligent driving controller according to claim 2, wherein the sensor simulation modeling method is characterized in that: in step S2, the static background includes buildings, roads and traffic signs; the dynamic objects include pedestrians and automobiles; the out-of-area objects include sky and roads.

4. The sensor simulation modeling method of the intelligent driving controller according to claim 2, wherein the sensor simulation modeling method is characterized in that: in step S2, during modeling of the dynamic object, the network is used to generate corresponding object features, and shape information is shared, so as to generate a complete vehicle shape.

5. The sensor simulation modeling method of the intelligent driving controller according to claim 4, wherein the sensor simulation modeling method is characterized in that: in step S3, the camera simulation generates a vivid image based on the geometric space of the environmental object, and then adds colors, optical attributes and the like to the three-dimensional model through computer graphics according to the real material and texture of the object to simulate the synthesis of the simulated image; simulating each ray emitted and received by the laser radar according to the scanning mode of the real laser radar, and intersecting the emitted ray with all objects in the scene; the millimeter wave radar simulation transmits a series of virtual continuous frequency modulation millimeter waves to different directions according to the field angle and resolution information of the radar configured by the test vehicle, and receives the reflected signals of the target.

6. The sensor simulation modeling method of the intelligent driving controller according to claim 5, wherein the sensor simulation modeling method is characterized in that: in step S4, the mixed sensor considers factors such as noise, object properties and the like on the basis of an ideal sensor, and an error signal is added to data information of an ideal object to be output; and modeling the working link of the sensor at a partial physical level, and outputting the sensor raw data.

7. The sensor simulation modeling method of the intelligent driving controller according to claim 5, wherein the sensor simulation modeling method is characterized in that: in step S5, the detection performance of the millimeter wave radar sensor is simulated according to a physical mechanism, and the accuracy of the millimeter wave radar sensor model is verified by fitting the millimeter wave radar sensor model and real millimeter wave radar measurement data and comparing and analyzing the distance error, the speed error and the angle error by adopting a statistical method.

8. A sensor simulation modeling apparatus of an intelligent driving controller according to any one of claims 1 to 7, characterized in that the above simulation modeling method is adopted.