A valet parking test method and system based on indoor and outdoor fusion positioning and generalized control of interference targets

Abstract

The present application relates to the field of parking test technology, and more specifically, to a valet parking test method and system based on indoor and outdoor fusion positioning and generalized control of interference targets. The system includes: a tested vehicle, equipped with a variety of sensors, an on-board terminal and a gateway; during a parking test in an underground parking lot, the on-board terminal obtains the position information of the tested vehicle based on data collected by a variety of sensors, and the gateway sends the position information to a central controller; the central controller generates trajectory information based on the behavior description, and controls the running trajectory of the interference target in the underground parking lot based on the trajectory information; the valet parking function of the tested vehicle is tested based on the position information of the tested vehicle and the running trajectory of the interference target. This embodiment can improve the accuracy and comprehensiveness of the test.

Description

Technical Field

[0001] The present application relates to the field of parking test technology, and in particular to a valet parking test method and system based on indoor and outdoor fusion positioning and generalized control of interference targets. Background Art

[0002] As an important part of the intelligent transportation field, valet parking technology has made significant technological progress in recent years. The valet parking system uses hardware devices such as sensors, cameras, radars, and LiDAR, as well as complex algorithms and artificial intelligence (AI) technology, enabling vehicles to autonomously complete the entire process from finding a parking space to parking the vehicle. This system not only reduces the parking pressure of drivers, but also improves parking efficiency and safety.

[0003] The current parking test system includes simulation test, bench test and field vehicle test. For field test, it is mostly to use multiple types of sensors to obtain video and position information of the vehicle during driving to determine the parking performance of the vehicle. It does not fully consider the complex environment of underground parking lots, the loss of underground positioning signals of the test system, and the precise control of interference targets when the positioning signal is lost. Faced with the complex and changing parking environment and the continuous upgrading of intelligent parking technology, the existing test system still needs to be continuously optimized and improved to improve the accuracy and comprehensiveness of the test and ensure the safety and practicality of valet parking technology.

[0004] In view of this, this application is filed. Summary of the invention

[0005] The purpose of this application is to provide a valet parking test method and system based on indoor and outdoor fusion positioning and generalized control of interference targets to improve the accuracy and comprehensiveness of the test.

[0006] In order to achieve the above objectives, this application adopts the following technical solutions:

[0007] In a first aspect, the present application provides a valet parking test system based on indoor and outdoor fusion positioning and generalized control of interference targets, including:

[0008] The tested vehicle is equipped with a plurality of sensors, an on-board terminal and a gateway; during a parking test in an underground parking lot, the on-board terminal obtains the position information of the tested vehicle according to the data collected by the plurality of sensors, and the gateway sends the position information to a central controller;

[0009] The central controller generates trajectory information according to the behavior description, and controls the running trajectory of the interference target in the underground parking lot according to the trajectory information;

[0010] The central controller tests the valet parking function of the tested vehicle according to the posture information of the tested vehicle and the running track of the interference target object.

[0011] In a second aspect, the present application provides a valet parking test method based on indoor and outdoor fusion positioning and generalized control of interference targets, including:

[0012] During a parking test of a tested vehicle in an underground parking lot, position information of the tested vehicle is obtained based on data collected by a plurality of sensors;

[0013] generating trajectory information according to the behavior description, and controlling the running trajectory of the interference target object in the underground parking lot according to the trajectory information;

[0014] The valet parking function of the tested vehicle is tested according to the position information of the tested vehicle and the running track of the interference target object.

[0015] Compared with the prior art, the beneficial effects of this application are:

[0016] This application is applicable to the scenario of testing the valet parking function of a vehicle in an underground parking lot. A variety of sensors, vehicle-mounted terminals and gateways are configured on the vehicle to be tested to accurately and in real time obtain the position and posture of the underground vehicle. The central controller can automatically generate trajectory information based on the behavior description. This method of obtaining the actual trajectory based on the description can improve the generalization of the trajectory, so as to comprehensively test the parking performance of the vehicle; and remotely control the running trajectory of the interference target object. The central controller can be configured on the ground, obtain the posture information of the tested vehicle and the running trajectory of the interference target object through communication, and test the valet parking function of the tested vehicle. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the specific implementation methods of the present application or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 is a structural schematic diagram of a valet parking test system based on indoor and outdoor fusion positioning and generalized control of interference targets provided in an embodiment of the present application;

[0019] Figure 2 It is a schematic diagram of arranging fixed beacons along the outline of a road line provided in an embodiment of the present application;

[0020] Figure 3is a schematic diagram of a beacon fixed on a parking space provided in an embodiment of the present application;

[0021] Figure 4 It is a schematic diagram of the structure of the gateway provided in the embodiment of the present application. DETAILED DESCRIPTION

[0022] The following is a description of exemplary embodiments of the present application in conjunction with the accompanying drawings, including various details of the embodiments of the present application to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present application. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0023] This embodiment provides a valet parking test system based on indoor and outdoor fusion positioning and interference target generalization control, which is suitable for testing the valet parking function of a vehicle in an underground parking lot. Figure 1 The system provided in this embodiment includes: a tested vehicle, a central controller and an interference target object. Detailed description is given below.

[0024] The tested vehicle is equipped with a parking algorithm, and the parking algorithm in the tested vehicle is tested through the parking performance of the tested vehicle in an underground parking lot and under the influence of interference targets.

[0025] The underground parking scene in this embodiment is specifically: an underground parking lot with one or more floors, where the parking lot is connected to the ground and between different parking lots, through a road that can interfere with the target road and the vehicle's travel, called a ramp. Since the GPS (Global Positioning System) signal in the underground parking lot is poor, in order to accurately locate the vehicle's position, the following arrangements need to be made in the underground parking lot before testing:

[0026] See also Figure 2 and Figure 3, fixed beacons are arranged along the outline of the road line in the ramps and parking areas (including parking spaces and aisles between parking spaces) of the underground parking lot. One fixed beacon is arranged at each of the three corner points of each parking space. Among them, the fixed beacon serves as a positioning base station for the underground parking lot and can emit UWB (Ultra Wideband) pulse signals. UWB is a wireless communication technology that does not require the use of a carrier, but instead transmits a series of very short energy pulses and uses an orthogonal method or direct sequencing to transmit pulses within a specified frequency range. Correspondingly, a signal receiver capable of receiving UWB pulse signals also needs to be installed on the tested vehicle. In addition, an inertial measurement unit is installed on the tested vehicle to locate the attitude angle and acceleration of the vehicle through inertial navigation signals. A positioning module is also installed on the tested vehicle to receive GPS positioning signals. It should be noted that the tested vehicle can also be equipped with sensors such as cameras, radars, and infrared sensing devices, which are not limited in this application.

[0027] When the vehicle enters the underground parking lot from the ground, the vehicle terminal determines the vehicle's position information based on the data of multiple fixed beacons, the data of the inertial measurement unit, and the data of the positioning module. Specifically, the vehicle terminal obtains the UWB pulse signal from the signal receiver, the vehicle's attitude angle and acceleration from the inertial measurement unit, and the GPS positioning signal from the positioning module, and fuses the above signals to obtain the real-time position and real-time attitude of the tested vehicle.

[0028] This embodiment divides different geographical areas according to the signal strength of fixed beacons, inertial measurement units and positioning models. When the tested vehicle is in different geographical areas, different fusion processing methods are used. Specifically, the path of the tested vehicle from the ground to the underground parking lot is divided into three geographical areas: the area with GPS signals on the ground, the area with poor GPS positioning signals such as ramps, and the area with no GPS positioning signals in the underground parking lot. (1) In the area with GPS signals on the ground, the vehicle terminal with GPS positioning function can receive GPS signals and inertial measurement unit signals to achieve accurate positioning of the vehicle and obtain attitude angles and accelerations. (2) In areas with poor GPS positioning signals such as ramps, the UWB pulse signal, the inertial measurement unit signal and the GPS positioning signal are fused to accurately position the vehicle. For example, when the GPS positioning signal can be received, the GPS positioning signal is used; when the GPS positioning signal cannot be received, the UWB pulse signal provides high-precision distance information for calculating the approximate position of the tested vehicle. The inertial measurement unit provides real-time acceleration and angular velocity information for correcting the measurement error and drift of the UWB. Through data fusion technology, such as Kalman filtering, the data of the two sensors are processed and estimated to obtain more accurate target location information. (3) In the stage where there is no GPS positioning signal at all, the UWB pulse signal received by the signal receiver and the signal of the inertial measurement unit are used to achieve underground indoor positioning.

[0029] Specifically, when analyzing the location of the tested vehicle through the UWB pulse signal, you can choose from a variety of algorithms such as the time difference of arrival (TDOA) algorithm and the time of arrival (TOA) algorithm. The UWB pulse signal is in the base station coordinate system, the inertial measurement unit signal is in the vehicle coordinate system, and the GPS signal is in the geodetic coordinate system, so the coordinate systems need to be unified. Assume that the geographical longitude, latitude, and altitude are used as the three mutually perpendicular axes of the geodetic coordinate system. The base station coordinate system established based on the parking lot is (X, Y, Z), and the vehicle coordinate system is (x, y, z), with the vehicle's driving direction as the x-axis, the side perpendicular to the vehicle body as the y-axis, and the vertical upward direction of the vehicle as the z-axis. The three coordinate systems can be transformed, and finally the inertial measurement unit signal and the GPS signal are converted to the base station coordinate system. When the tested vehicle reaches the entrance of the underground garage, it needs to pass the starting test position, which is the origin of the base station coordinate system of the parking lot (X0, Y0, Z0). The original attitude angle (α0, β0, γ0) of the vehicle at the starting test position is obtained through the inertial measurement unit. It is known that the geodetic coordinate system of the starting test position is (L0, B0, H0). The starting test position and the original attitude angle are converted to the base station coordinate system. The real-time acceleration is integrated for the starting test position to obtain the real-time position of the tested vehicle, which is used to correct the position obtained by the UWB pulse signal. The real-time attitude angle is integrated for the original attitude angle to obtain the real-time attitude angle of the tested vehicle.

[0030] This embodiment improves the accuracy of posture information by fusing multiple sensor data. The vehicle terminal sends real-time posture information to the gateway of the tested vehicle through the in-vehicle network. The gateway sends the real-time posture information to the central controller on the ground through Ethernet.

[0031] See also Figure 2 When testing the parking function, it is necessary to set movable obstacles around the tested vehicle or on the parking route, which are called interference targets. Interference targets are used to simulate people or simulated vehicles to test whether the vehicle will collide with the interference targets. Optionally, the interference targets include a load-bearing platform (such as a sliding base plate), an object (such as a dummy) located on the load-bearing platform, and wheels located under the load-bearing platform. The central controller can remotely control the motor of the interference target, control the rotation of the wheel under the drive of the motor, and thus change the running trajectory of the interference target. In other embodiments, a track matching the wheel can be set on the parking lot, so that the interference target can move along the track by only controlling the wheel to move forward / backward.

[0032] The gateway in the central controller receives the position information of the tested vehicle, generates trajectory information according to the behavior description through the processing unit, and controls the running trajectory of the interference target in the underground parking lot according to the trajectory information.

[0033] Among them, the behavior description refers to the natural language description or text description of the running trajectory of the interference target, including dynamic targets, actions, duration and environment. The dynamic target S is the category of the key object, including adult pedestrians, children, cars, sprinkler trucks and other categories. The environment E includes weather, time, road type, and current lane. Assuming that the dynamic target S is a vehicle, the action A includes acceleration, deceleration, left turn, lane change, etc., which is an action set. It is mainly divided into three categories: 1) speed control actions, including acceleration, deceleration, rapid deceleration, slow straight driving, straight driving at a constant speed, parking, waiting and reversing; 2) turning actions, including turning left, turning right and U-turn; 3) lane control actions, including changing lanes to the left, changing lanes to the right, turning slightly to the left and turning slightly to the right. Assuming that the dynamic target S is a pedestrian, the action set is two categories: 1) speed control actions, including acceleration, deceleration, rapid deceleration, etc.; 2) lane control actions, including crossing the road, driving longitudinally, driving diagonally from near to far, and avoiding the vehicle in front to the left. The duration D indicates how long the action should last or when it should start. The trajectory information includes a plurality of three-dimensional coordinate points arranged in time sequence.

[0034] The central controller inputs the behavior description into the pre-trained model and obtains the trajectory information output by the model: the trajectory curve. Optionally, a large language model is used as the basic model for training, such as the GPT (Generative Pre-trained Transformer) model, the LLaMA (Large Language Model Meta AI) model, etc.

[0035] Before using the model, the model needs to be trained. Specifically, the central controller includes a training module for obtaining training samples and using the training samples to train the model; wherein obtaining training samples includes: collecting multiple continuous images through the vehicle's front-view camera, marking dynamic targets, actions, durations and environments in the multiple continuous images, and determining the trajectory information of the dynamic targets according to the positions of the dynamic targets in the multiple continuous images; converting the trajectory information of the dynamic targets to the base station coordinate system of the underground parking lot. Specifically, the trajectory points of the dynamic targets in the image coordinate system are converted to three-dimensional trajectory points in the front-view camera coordinate system through perspective projection transformation, and the three-dimensional trajectory points are converted to the base station coordinate system of the underground parking lot.

[0036] In a specific implementation, a front-view camera with a central viewing angle of the front windshield is selected to collect multiple continuous images, and key frames in the multiple continuous images are found, such as the moment when the dynamic target object undergoes a significant change, and the continuous images where the key frames are located are annotated. The key frame is used as the starting point T0 for annotation. Assuming that the front-view camera collects 20 frames per second, every 20 frames are annotated once, and a total of 10 times are annotated, that is, 10 seconds of data are annotated. The specific annotation content of each frame includes: 1) Behavior description: dynamic target S, action A, duration D, environment E; 2) The pixel coordinates of the corner points of the minimum circumscribed rectangular frame of the dynamic target. Then, the annotated samples are further processed to calculate the central pixel coordinates of the dynamic target of the 10 annotated frames [(uC1, vC1), (uC2, vC3), ..., (uC10, vC10)], where (uC1, vC1) is the horizontal and vertical coordinates of the dynamic target in the image in the first frame, and so on, the horizontal and vertical coordinates of the dynamic target in the image in 10 frames are calculated. Through perspective projection transformation, the horizontal and vertical coordinates in the image are converted into three-dimensional trajectory points in the front camera coordinate system [(xC1, yC1, zC1), (xC2, yC2, zC2), ..., (xC10, yC10, zC10)]. Among them, (xC1, yC1, zC1) is the coordinate of the dynamic target in the first frame in the front camera coordinate system, and the same is true for the subsequent ones, and a total of 10 three-dimensional trajectory points are obtained. Among them, the front camera is installed at a fixed position of the vehicle, and there is a translation and rotation relationship between the front camera coordinate system and the vehicle body coordinate system. This relationship can be used to convert the three-dimensional trajectory points in the front camera coordinate system to the vehicle body coordinate system. Further, according to the vehicle body coordinate system and the position of the vehicle in the parking lot, the three-dimensional trajectory points are converted to the base station coordinate system of the underground parking lot. The three-dimensional trajectory points are curve fitted in the base station coordinate system to obtain a trajectory curve. Optionally, the trajectory curve can be represented by multiple three-dimensional trajectory point coordinates arranged in time sequence, or by the parameters of the curve equation.

[0037] For example, after completing the model training, the behavior description is directly input into the model, for example, on a sunny day with two lanes in both directions, in the first lane (E), a child (S) accelerates to cross the road (A) for 10 seconds (D). The trained model generates a trajectory curve W variation according to the behavior description.

[0038] In some cases, the trajectory curve directly generated by the model cannot be directly applied to the current scene, and the trajectory curve needs to be corrected before it can be used to control the interference target. Specifically, the trajectory curve is corrected according to the parking test area and kinematic constraints.

[0039] Assuming that the parking test area is 100m×100m, the range of the trajectory curve should also be within the range of 100m×100m. If there is a part that exceeds the range, the 3D trajectory points that exceed the range will be reset to the parking test area. Kinematic constraints refer to the constraints on the travel speed of common targets (such as people, vehicles, etc.). If there is a part in the trajectory curve that exceeds the speed constraint, the position of the 3D trajectory points of this part will be reset so that the 3D trajectory points meet the speed constraint.

[0040] In some embodiments, model control and rule control can be used to determine trajectory information. For example, the central controller obtains the test task and test style (including generalization control and rule control) of the tested vehicle. If the test style is generalization control, the behavior description is determined according to the test task of the vehicle; if the test style is rule control, the travel speed, path and collision point of the interference target are determined according to the test task of the vehicle; and the trajectory information is generated according to the travel speed, path and collision point of the interference target.

[0041] If the test style is generalized control, that is, the interference target is controlled according to the trajectory curve of the interference target extracted by the model. Specifically, the central controller extracts multiple three-dimensional trajectory points from the trajectory curve, and converts them into control instructions for the motor according to the three-dimensional trajectory points, including steering control instructions and speed control instructions. The motor of the interference target responds to the control instruction and drives the soft target carrying platform to move, so that the soft target carrying platform (with six degrees of freedom) drives the interference target to travel according to the trajectory curve. Since the model has a certain generalization ability, even if the same behavior description is input, different trajectory information may be obtained. Therefore, trajectory information can be generated multiple times according to the same behavior description to comprehensively test the parking function.

[0042] If the test style is rule control, the trajectory curve of the interference target is calculated based on the fixed driving path, speed, collision requirements and other parameters of the interference target. There are the following implementation methods:

[0043] 1. Taking the interference target as a dummy as an example, the test task is pedestrian crossing, that is, the vehicle is driving in the aisle area of ​​the parking lot, a certain distance away from the dummy. The central controller arranges the track position of the dummy in advance according to the positioning information of the tested vehicle and the requirements of the collision point in the test task, calculates the driving speed of the dummy, and controls the speed of the dummy's guide conveyor belt through the motor controller of the interference target, so as to keep the collision position of the dummy and the tested vehicle consistent each time. The collision point can be the middle of the road or the center point of the contour line of the front of the vehicle.

[0044] 2. Taking the interference target as a dummy car as an example, the test task is meeting, that is, the tested vehicle meets the dummy car of the target vehicle in the aisle area. The central controller arranges the initial position of the sliding bottom plate carried by the dummy car in advance according to the positioning information of the tested vehicle and the requirements of the meeting trajectory in the test task, and controls the driving speed and direction of the sliding bottom plate through the motor controller of the interference target. When the body of the tested vehicle completely passes the dummy car, the test ends.

[0045] 3. Taking the parking in and out of the target parking space as an example, the test task is parking in and out of the parking space. Fixed beacons need to be installed at the three corner points of the target parking space, and the positions of the three corner points of the target parking space are known. The fixed beacon communicates with the tested vehicle, so that the central controller obtains the real-time position of the tested vehicle according to the UWB signal; the vehicle attitude angle is obtained from the inertial measurement unit, and the relative parallel relationship between the vehicle and the parking space is known.

[0046] 4. Taking cross-layer cruise parking as an example, the ramp connected across layers can be a one-way or two-way lane with a certain curvature and slope. Fixed beacons are arranged along both ends and left and right sides of the ramp, so that each position of the tested vehicle is covered by pulse signals of 3-4 fixed beacons during driving to ensure the positioning effect.

[0047] The central controller tests the valet parking function of the tested vehicle based on the position information of the tested vehicle and the running trajectory of the interference target. The central controller matches the corresponding evaluation and outputs the test results based on the movement trajectory of the tested vehicle, the interaction between the tested vehicle and the interference target, and the parking in and out situation. There are three specific evaluation indicators: 1) Evaluate the safety of parking based on the relative distance between the tested vehicle and the interference target; 2) Evaluate the accuracy of parking based on the position and posture of the tested vehicle; 3) Evaluate the parking efficiency based on the number of times the tested vehicle rubs the parking lot.

[0048] Specifically, the real-time position and real-time posture of the tested vehicle are known through the signal receiver, inertial measurement unit and positioning module. The position of the interfering target is obtained through the trajectory curve. The relative distance between the two at the same time is calculated. If the relative distance is less than the set value, there is a risk of collision, and the parking function is not safe. The vehicle posture after the vehicle is parked is determined through the trajectory information, posture angle and position information of the target parking space of the tested vehicle, and the difference between the posture (such as yaw angle) and the parking reference posture (such as yaw angle) is calculated. If the difference is not within the set range, it is not accurate. With the help of sensor information arranged in the parking lot or on the vehicle body, such as external cameras, internal cameras, laser radars, etc., the number of times the vehicle rubs the garage is determined and the parking efficiency is evaluated. The above three test indicators can be visualized on the interface of the central controller.

[0049] In combination with the above-mentioned embodiments, the valet parking test system based on the indoor and outdoor fusion positioning system provided in the present application can realize the seamless connection of the positioning of the ground and underground parking tests of the valet parking system; link the interference target objects to realize the precise control of the driving trajectory of the interference target objects; realize the two testing methods of rule-based and model-based generalization, so as to better fit various working conditions in the real world.

[0050] The embodiments of the present application achieve accurate positioning of the tested vehicle throughout the entire process, both above and below ground, by integrating ground GPS positioning signals, UWB pulse signals, and inertial navigation signals, combined with the conversion of coordinate systems. By controlling the tested vehicle and the interference target object in linkage, accurate non-standard trajectory control in the basement scenario is achieved, ensuring the accuracy and reproducibility of the test items. Rule-based control is too fixed, and it is difficult to exhaustively enumerate the test conditions of various types of targets. Using a model to obtain the driving style of the interference target object and form executable control instructions can be closer to the real world and generalize the test conditions, which is conducive to improving the comprehensiveness and accuracy of parking tests.

[0051] The embodiment of the present application also provides a valet parking test method based on indoor and outdoor fusion positioning and generalized control of interference targets, which is suitable for testing the valet parking function of a vehicle in an underground parking lot scenario.

[0052] See also Figure 4 The method provided in this embodiment is executed by a valet parking test system based on indoor and outdoor fusion positioning and interference target generalization control, and specifically includes the following operations:

[0053] S110. During a parking test of a tested vehicle in an underground parking lot, position information of the tested vehicle is obtained based on data collected by a plurality of sensors.

[0054] S120: Generate trajectory information according to the behavior description, and control the running trajectory of the interference target object in the underground parking lot according to the trajectory information.

[0055] S130: Testing the valet parking function of the tested vehicle according to the position information of the tested vehicle and the running track of the interference target object.

[0056] Optionally, generating trajectory information based on the behavior description includes: inputting the behavior description into a pre-trained model to obtain a trajectory curve output by the model; and modifying the trajectory curve according to a parking test area and kinematic constraints; wherein the behavior description includes a dynamic target, action, duration and environment.

[0057] Optionally, before S120, a training sample is obtained; and the model is trained using the training sample. The trajectory information of the dynamic target object is converted to the base station coordinate system of the underground parking lot, including: converting the trajectory points of the dynamic target object in the image coordinate system to three-dimensional trajectory points in the front camera coordinate system through perspective projection transformation, converting the three-dimensional trajectory points to the base station coordinate system of the underground parking lot, and performing curve fitting on the three-dimensional trajectory points in the base station coordinate system to obtain a trajectory curve.

[0058] Optionally, generating trajectory information according to the behavior description includes: obtaining the test task and test style of the tested vehicle; if the test style is generalized control, determining the behavior description according to the test task of the vehicle. Further, if the test style is rule control, determining the travel speed, path and collision point of the interference target according to the test task of the vehicle; generating trajectory information according to the travel speed, path and collision point of the interference target.

[0059] Optionally, in the process of the vehicle entering the underground parking lot from the ground, the posture information of the tested vehicle is obtained based on the data collected by multiple sensors, including: determining the vehicle's posture information based on data from multiple fixed beacons, data from the inertial measurement unit and data from the positioning module.

[0060] Optionally, the valet parking function of the tested vehicle is tested based on the position information of the tested vehicle and the running trajectory of the interference target, including: evaluating the safety of parking based on the relative distance between the tested vehicle and the interference target; evaluating the accuracy of parking based on the position and posture of the tested vehicle; and evaluating the parking efficiency based on the number of times the tested vehicle enters the parking lot.

[0061] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this application can be executed in parallel, sequentially or in different orders, as long as the expected results of the technical solution disclosed in this application can be achieved, and this document is not limited here.

[0062] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A valet parking test system based on indoor and outdoor fusion positioning and interference target generalization control, characterized in that: include: The tested vehicle is equipped with a variety of sensors, on-board terminals and gateways; During the parking test in the underground parking lot, the vehicle-mounted terminal obtains the position information of the tested vehicle based on the data collected by various sensors, and the gateway sends the position information to the central controller; The central controller generates trajectory information according to the behavior description, and controls the running trajectory of the interference target in the underground parking lot according to the trajectory information; the behavior description refers to a natural language description or text description of the running trajectory of the interference target, including a dynamic target, an action, a duration and an environment, wherein the environment includes weather, time, road type, and current lane; the action includes three categories: speed control action, steering action and lane control action; the duration indicates how long the action should last or when it should start; the trajectory information includes a plurality of three-dimensional coordinate points arranged in time sequence; When the central controller generates trajectory information according to the behavior description, the behavior description is input into the pre-trained model to obtain the trajectory curve output by the model; the model has generalization ability, and different trajectory information is obtained by inputting the same behavior description. Trajectory information is generated multiple times according to the same behavior description to comprehensively test the parking function; The central controller tests the valet parking function of the tested vehicle according to the posture information of the tested vehicle and the running track of the interference target object.

2. The system according to claim 1, characterized in that After obtaining the trajectory curve output by the model, the method further includes: The trajectory curve is modified according to the parking test area and kinematic constraints.

3. The system according to claim 2, characterized in that Also includes: A training module, used to obtain training samples; Using the training samples to train the model; Among them, obtaining training samples includes: collecting multiple continuous images through the vehicle's front-view camera, marking dynamic targets, actions, duration and environment in the multiple continuous images, and determining the trajectory information of the dynamic targets according to the positions of the dynamic targets in the multiple continuous images; converting the trajectory information of the dynamic targets to the base station coordinate system of the underground parking lot.

4. The system according to claim 3, characterized in that Converting the trajectory information of the dynamic target object to the base station coordinate system of the underground parking lot includes: The trajectory points of the dynamic target object in the image coordinate system are converted into three-dimensional trajectory points in the forward-looking camera coordinate system through perspective projection transformation, and the three-dimensional trajectory points are converted to the base station coordinate system of the underground parking lot. The three-dimensional trajectory points are curve fitted in the base station coordinate system to obtain a trajectory curve.

5. The system according to claim 4, characterized in that The central controller is further used to obtain the test task and test style of the tested vehicle; if the test style is generalized control, determine the behavior description according to the test task of the vehicle; If the test style is rule control, the travel speed, path and collision point of the interference target object are determined according to the test task of the vehicle; and trajectory information is generated according to the travel speed, path and collision point of the interference target object.

6. The system according to claim 1, characterized in that The underground parking lot is provided with a plurality of fixed beacons; The tested vehicle is provided with a signal receiver, an inertial measurement unit and a positioning module; When the vehicle enters the underground parking lot from the ground, the vehicle terminal determines the vehicle's position information based on data from multiple fixed beacons, data from an inertial measurement unit, and data from a positioning module.

7. The system according to claim 1, characterized in that The interference target object includes a carrying platform, an object located on the carrying platform, and a wheel located under the carrying platform.

8. The system according to claim 1, characterized in that When the central controller tests the valet parking function of the tested vehicle based on the position information of the tested vehicle and the running trajectory of the interference target, it evaluates the safety of parking based on the relative distance between the tested vehicle and the interference target; evaluates the accuracy of parking based on the position and posture of the tested vehicle; and evaluates the parking efficiency based on the number of times the tested vehicle rubs the parking space.

9. A valet parking test method based on indoor and outdoor fusion positioning and interference target generalization control, characterized in that: During a parking test of a tested vehicle in an underground parking lot, position information of the tested vehicle is obtained based on data collected by a plurality of sensors; Generate trajectory information according to the behavior description, and control the running trajectory of the interference target in the underground parking lot according to the trajectory information; the behavior description refers to a natural language description or text description of the running trajectory of the interference target, including a dynamic target, an action, a duration and an environment, wherein the environment includes weather, time, road type, and current lane; the action includes three categories: speed control action, steering action and lane control action; the duration indicates how long the action should last or when it should start; the trajectory information includes a plurality of three-dimensional coordinate points arranged in time sequence; When generating trajectory information according to the behavior description, the behavior description is input into the pre-trained model to obtain the trajectory curve output by the model; the model has generalization ability, and different trajectory information is obtained by inputting the same behavior description, and trajectory information is generated multiple times according to the same behavior description to comprehensively test the parking function; The valet parking function of the tested vehicle is tested according to the position information of the tested vehicle and the running track of the interference target object.

Images