Training method and device for sensor error prediction model for autonomous driving

Abstract

The present application discloses a training method and device for a sensor error prediction model for autonomous driving, the method comprising: obtaining high-precision positioning information of an autonomous driving vehicle; obtaining target sensor information of the autonomous driving vehicle when the high-precision positioning information meets preset training conditions, the target sensor information including the positioning information and original positioning error information of the target sensor; determining the true correction error of the target sensor based on the high-precision positioning information and the positioning information of the target sensor; and training a sensor error prediction model using the original positioning error information and the true correction error of the target sensor to obtain a trained sensor error prediction model. The present application determines the true correction error of the sensor based on high-precision positioning information that meets preset training conditions, and uses this to train the sensor error prediction model, thereby improving the prediction accuracy of positioning errors between sensors and providing strong support for the fusion positioning of autonomous driving vehicles.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving technology, and in particular to a training method and device for a sensor error prediction model for autonomous driving. Background Art

[0002] In autonomous driving scenarios, it is often necessary to fuse data from multiple sensors to obtain fused positioning information in order to ensure the positioning accuracy of autonomous driving vehicles in complex road conditions in various cities.

[0003] However, due to factors such as calibration and time delay between sensors, there are large errors between the positioning information of each sensor obtained during fusion positioning. Direct fusion positioning will cause even greater errors. Summary of the Invention

[0004] The embodiments of the present application provide a method and apparatus for training a sensor error prediction model for autonomous driving to improve the prediction accuracy of positioning errors between sensors.

[0005] The embodiments of this application adopt the following technical solutions:

[0006] In a first aspect, an embodiment of the present application provides a method for training a sensor error prediction model for autonomous driving, wherein the method comprises:

[0007] Obtain high-precision positioning information for autonomous vehicles;

[0008] When the high-precision positioning information satisfies a preset training condition, obtaining target sensor information of the autonomous driving vehicle, the target sensor information including positioning information and original positioning error information of the target sensor;

[0009] determining a true correction error of the target sensor based on the high-precision positioning information and the positioning information of the target sensor;

[0010] The original positioning error information and the true correction error of the target sensor are used to train a sensor error prediction model to obtain a trained sensor error prediction model.

[0011] Optionally, when the high-precision positioning information meets a preset training condition, obtaining target sensor information of the autonomous driving vehicle includes:

[0012] Predicting the confidence of the high-precision positioning information using a positioning confidence prediction model;

[0013] If the confidence level of the high-precision positioning information is greater than a preset confidence threshold, determining that the high-precision positioning information meets the preset training condition;

[0014] Otherwise, it is determined that the high-precision positioning information does not meet the preset training condition.

[0015] Optionally, the original positioning error information includes a SLAM covariance and an original correction value, and the sensor error prediction model is trained using the original positioning error information and a true correction error of the target sensor to obtain the trained sensor error prediction model, comprising:

[0016] The SLAM covariance and the original correction amount are used as inputs of the sensor error prediction model, and the true correction error of the target sensor is used as a supervision signal to train the sensor error prediction model to obtain a trained sensor error prediction model.

[0017] Optionally, after determining a true correction error of the target sensor based on the high-precision positioning information and the positioning information of the target sensor, the method further includes:

[0018] Based on the original positioning error information and the true correction error of the target sensor, a nonlinear optimization algorithm is used to perform fitting optimization to obtain a corresponding relationship between the original positioning error information and the true correction error of the target sensor.

[0019] Optionally, the sensor error prediction model adopts an LSTM long short-term memory network.

[0020] In a second aspect, an embodiment of the present application further provides a fusion positioning method for an autonomous driving vehicle, wherein the method includes:

[0021] Acquire current target sensor information, where the current target sensor information includes current positioning information and current original positioning error information;

[0022] Predicting a true correction error of the target sensor using a sensor error prediction model based on the current target sensor information;

[0023] Correcting the current positioning information using the true correction error of the target sensor to obtain corrected positioning information;

[0024] Performing fusion positioning based on the corrected positioning information to obtain a fusion positioning result of the autonomous driving vehicle;

[0025] The sensor error prediction model is trained based on any of the aforementioned sensor error prediction model training methods.

[0026] In a third aspect, an embodiment of the present application further provides a training device for a sensor error prediction model for autonomous driving, wherein the device comprises:

[0027] A first acquisition unit is used to obtain high-precision positioning information of the autonomous driving vehicle;

[0028] a second acquiring unit, configured to acquire target sensor information of the autonomous driving vehicle when the high-precision positioning information satisfies a preset training condition, the target sensor information including positioning information and original positioning error information of the target sensor;

[0029] a determining unit, configured to determine a true correction error of the target sensor based on the high-precision positioning information and the positioning information of the target sensor;

[0030] The training unit is used to train a sensor error prediction model using the original positioning error information and the true correction error of the target sensor to obtain a trained sensor error prediction model.

[0031] In a fourth aspect, an embodiment of the present application further provides a fusion positioning device for an autonomous driving vehicle, wherein the device includes:

[0032] a third acquiring unit, configured to acquire current target sensor information, wherein the current target sensor information includes current positioning information and current original positioning error information;

[0033] a prediction unit, configured to predict a true correction error of the target sensor using a sensor error prediction model based on the current target sensor information;

[0034] A correction unit, configured to correct the current positioning information using a true correction error of the target sensor to obtain corrected positioning information;

[0035] a fusion positioning unit, configured to perform fusion positioning based on the corrected positioning information to obtain a fusion positioning result of the autonomous driving vehicle;

[0036] The sensor error prediction model is obtained by training based on the aforementioned sensor error prediction model training device.

[0037] In a fifth aspect, an embodiment of the present application further provides an electronic device, including:

[0038] processor; and

[0039] A memory arranged to store computer-executable instructions, which, when executed, cause the processor to perform any of the aforementioned training methods for sensor error prediction models for autonomous driving, or to perform the aforementioned fusion positioning method for autonomous driving vehicles.

[0040] In a sixth aspect, an embodiment of the present application further provides a computer-readable storage medium, which stores one or more programs. When the one or more programs are executed by an electronic device including multiple applications, the electronic device executes any of the aforementioned training methods for sensor error prediction models for autonomous driving, or executes the aforementioned fusion positioning method for autonomous driving vehicles.

[0041] At least one of the above technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects: the training method of the sensor error prediction model for autonomous driving in the embodiments of the present application first obtains high-precision positioning information of the autonomous driving vehicle; then, when the high-precision positioning information meets the preset training conditions, obtains the target sensor information of the autonomous driving vehicle, the target sensor information including the positioning information and original positioning error information of the target sensor; then, based on the high-precision positioning information and the positioning information of the target sensor, determines the true correction error of the target sensor; finally, uses the original positioning error information and the true correction error of the target sensor to train the sensor error prediction model to obtain the trained sensor error prediction model. The training method of the sensor error prediction model for autonomous driving in the embodiments of the present application determines the true correction error of the sensor based on the high-precision positioning information that meets the preset training conditions, and uses this to train the sensor error prediction model, thereby improving the prediction accuracy of the positioning error between sensors and providing strong support for the subsequent fusion positioning of autonomous driving vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0043] Figure 1 Schematic diagram of a flow chart of a method for training a sensor error prediction model for autonomous driving according to an embodiment of the present application;

[0044] Figure 2 This is a flow chart of a fusion positioning method for an autonomous driving vehicle in an embodiment of the present application;

[0045] Figure 3 Schematic diagram of the structure of a training device for a sensor error prediction model for autonomous driving in an embodiment of the present application;

[0046] Figure 4 This is a schematic structural diagram of a fusion positioning device for an autonomous driving vehicle in an embodiment of the present application;

[0047] Figure 5 This is a schematic structural diagram of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION

[0048] To make the purpose, technical solutions, and advantages of this application more clear, the technical solutions of this application will be clearly and completely described below in conjunction with the specific embodiments of this application and the corresponding drawings. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0049] The following describes in detail the technical solutions provided by various embodiments of the present application in conjunction with the accompanying drawings.

[0050] The present application embodiment provides a method for training a sensor error prediction model for autonomous driving, such as Figure 1 As shown, a flow chart of a method for training a sensor error prediction model for autonomous driving according to an embodiment of the present application is provided. The method includes at least the following steps S110 to S140:

[0051] Step S110: Obtain high-precision positioning information of the autonomous driving vehicle.

[0052] When training the sensor error prediction model in the embodiment of the present application, it is necessary to first obtain high-precision positioning information of the autonomous driving vehicle. The high-precision positioning information here can refer to the positioning information output by the inertial navigation RTK (Real-time kinematic). Inertial navigation RTK refers to the positioning information output after the IMU (Inertial Measurement Unit) positioning data collected by the inertial navigation device is integrated with the RTK real-time differential algorithm. Compared with pure inertial navigation positioning data and RTK positioning data, it has higher positioning accuracy and can therefore serve as the basis for subsequent measurement of sensor positioning errors.

[0053] Step S120, when the high-precision positioning information meets the preset training conditions, obtain the target sensor information of the autonomous driving vehicle, where the target sensor information includes the positioning information and original positioning error information of the target sensor.

[0054] The "high-precision positioning information" in the embodiments of the present application mainly refers to the positioning information obtained based on inertial navigation RTK, but it does not mean that the positioning information always has high accuracy. For example, the positioning information may be affected by the quality of the satellite positioning signal, resulting in reduced accuracy. The RTK positioning status can roughly reflect the accuracy of the high-precision positioning information to a certain extent, but simply relying on the RTK status to judge the accuracy of the high-precision positioning information cannot meet the needs of autonomous driving. For example, in complex urban road scenes, when the RTK positioning status is a fixed solution, the high-precision positioning information may still have a large positioning error.

[0055] Based on this, the embodiment of the present application needs to use certain strategies to determine whether the accuracy of the high-precision positioning information meets the preset training conditions. Only when the preset training conditions are met can the high-precision positioning information be used as the basis for measuring sensor errors, thereby improving the accuracy and effect of model training.

[0056] If the current high-precision positioning information meets the preset training conditions, then the target sensor information of the autonomous driving vehicle can be further obtained. The target sensor can be understood as the target object for which a sensor error prediction model needs to be constructed. For example, it can be a lidar, camera, etc. The target sensor information can specifically include the positioning information and original positioning error information output by the target sensor. The original positioning error information can be regarded as a preliminary positioning error information calculated by the target sensor based on its own positioning algorithm and strategy. The positioning error information is generally of low accuracy, so positioning correction cannot be performed directly based on the positioning error information. This is also the main reason for constructing a sensor error prediction model in the embodiment of the present application.

[0057] Step S130 : determining a true correction error of the target sensor according to the high-precision positioning information and the positioning information of the target sensor.

[0058] As previously mentioned, high-precision positioning information, when it meets preset training conditions, can serve as the basis for measuring the positioning error of the target sensor. Therefore, the positioning deviation between the target sensor's positioning information and the high-precision positioning information can be calculated as the true correction error. The ultimate goal of model training is to accurately predict the correction error, which serves as the basis for correcting positioning information.

[0059] Therefore, the embodiment of the present application is equivalent to starting the inertial navigation RTK equipment and the positioning algorithm based on the target sensor at the same time during the training phase, so as to determine the actual correction error.

[0060] Step S140 , using the original positioning error information and the actual correction error of the target sensor, training a sensor error prediction model to obtain a trained sensor error prediction model.

[0061] Although the original positioning error information cannot be used directly to correct the positioning error, it can be used as input information in the model training stage. The true correction error of the target sensor is the actual positioning error with high accuracy. Therefore, it can constrain the model training process. Based on the information of these two dimensions, the sensor error prediction model can be trained to obtain the trained sensor error prediction model.

[0062] The training method of the sensor error prediction model for autonomous driving in the embodiment of the present application determines the actual correction error of the sensor based on high-precision positioning information that meets preset training conditions, and uses this to train the sensor error prediction model, thereby improving the prediction accuracy of positioning errors between sensors and providing strong support for the subsequent fusion positioning of autonomous driving vehicles.

[0063] In some embodiments of the present application, the sensor error prediction model of the embodiments of the present application can be trained online based on the interaction between the vehicle and the cloud. For example, the vehicle of an autonomous vehicle can collect inertial navigation RTK positioning information in real time and determine its accuracy. When the accuracy meets the preset training conditions, the inertial navigation RTK positioning information and target sensor information are sent to the cloud for online training, thereby reducing the computing load on the vehicle and meeting the real-time requirements. Of course, offline training can also be used. The specific method to be adopted can be flexibly selected by those skilled in the art according to actual needs.

[0064] In some embodiments of the present application, the high-precision positioning information of the embodiments of the present application can also be collected by high-precision inertial navigation equipment installed on the vehicle. The positioning effect of the high-precision inertial navigation equipment is little affected by external factors and has higher positioning accuracy than general inertial navigation equipment.

[0065] In some embodiments of the present application, when the high-precision positioning information meets the preset training conditions, obtaining the target sensor information of the autonomous driving vehicle includes: using a positioning confidence prediction model to predict the confidence of the high-precision positioning information; if the confidence of the high-precision positioning information is greater than a preset confidence threshold, determining that the high-precision positioning information meets the preset training conditions; otherwise, determining that the high-precision positioning information does not meet the preset training conditions.

[0066] When determining whether the high-precision positioning information meets the preset training conditions, the embodiment of the present application can first determine the confidence of the currently acquired high-precision positioning information, that is, the degree of credibility of the positioning result. Here, a pre-trained positioning confidence prediction model can be used to predict the confidence of the high-precision positioning information, and then the confidence of the high-precision positioning information is compared with the preset confidence threshold. If it is greater than the preset confidence threshold, it is considered that the accuracy of the high-precision positioning information meets the subsequent training requirements. Otherwise, it cannot be used for subsequent training to ensure the accuracy of the model training.

[0067] The size of the above-mentioned preset confidence threshold can be flexibly set according to actual training requirements. For example, it can be set to 0.95. When the confidence of the high-precision positioning information exceeds 0.95, it can be considered that the accuracy is high and can be used for subsequent training. Of course, those skilled in the art can also flexibly adjust it according to actual needs, and no specific limitation is made here.

[0068] The positioning confidence prediction model of the embodiment of the present application can be trained in the following manner: obtaining inertial navigation RTK information and corresponding high-precision positioning information, the inertial navigation RTK information including inertial navigation RTK positioning information, the absolute time of the inertial navigation RTK information, the RTK positioning status, the horizontal position precision factor, and the number of satellites; determining the positioning error of the inertial navigation RTK positioning information based on the high-precision positioning information; determining the confidence of the inertial navigation RTK positioning information based on the positioning error of the inertial navigation RTK positioning information; and training the positioning confidence prediction model using the inertial navigation RTK information and the confidence of the inertial navigation RTK positioning information to obtain a trained positioning confidence prediction model.

[0069] In an embodiment of the present application, when determining the confidence level of the inertial navigation RTK positioning information based on the positioning error of the inertial navigation RTK positioning information, the following method can be specifically adopted: if the positioning error of the inertial navigation RTK positioning information is not greater than the first preset error threshold, the confidence level of the inertial navigation RTK positioning information is determined to be the first confidence level; if the positioning error of the inertial navigation RTK positioning information is not less than the second preset error threshold, the confidence level of the inertial navigation RTK positioning information is determined to be the second confidence level; if the positioning error of the inertial navigation RTK positioning information is greater than the first preset error threshold and less than the second preset error threshold, the confidence level of the inertial navigation RTK positioning information is determined to be the third confidence level; wherein the first preset error threshold is less than the second preset error threshold, the first confidence level is greater than the third confidence level, and the third confidence level is greater than the second confidence level.

[0070] In some embodiments of the present application, the original positioning error information includes SLAM covariance and original correction amount, and the use of the original positioning error information and the true correction error of the target sensor to train the sensor error prediction model to obtain the trained sensor error prediction model includes: using the SLAM covariance and the original correction amount as inputs of the sensor error prediction model, using the true correction error of the target sensor as a supervision signal, training the sensor error prediction model, and obtaining the trained sensor error prediction model.

[0071] As mentioned above, the target sensor of the embodiment of the present application may refer to a lidar and a camera, etc. The data collected by the lidar can realize a positioning solution based on laser SLAM (Simultaneous Localization And Mapping), and the data collected by the camera can realize a positioning solution based on visual SLAM.

[0072] Whether it is laser SLAM or visual SLAM, its main function is to provide auxiliary positioning when high-precision positioning signals such as RTK are interfered with or cannot provide high-precision positioning information. Therefore, the positioning deviation between laser SLAM or visual SLAM and other sensors also affects the final fusion positioning accuracy of the autonomous driving vehicle. Based on this, the purpose of training the sensor error prediction model in the embodiment of the present application is to accurately predict the positioning deviation between sensors such as laser SLAM or visual SLAM, so as to perform positioning correction and improve the fusion positioning accuracy.

[0073] The positioning algorithm based on laser SLAM / visual SLAM will output a SLAM covariance and a corresponding raw correction. The SLAM covariance represents the size of the positioning deviation between the laser SLAM / visual SLAM and other sensors, while the raw correction is an initial correction value calculated by the laser SLAM / visual SLAM positioning algorithm based on the size of the covariance. This correction value is generally of low accuracy and has high sensitivity only when the positioning error is large.

[0074] Therefore, the embodiment of the present application can use the SLAM covariance and the corresponding original correction amount as the input of the sensor error prediction model. Although there is a certain error in the original correction amount, it can improve the training efficiency of the model as a reference information input, and use the actual correction error of the target sensor as the output of the sensor error prediction model, thereby supervising and constraining the training process of the model. When the model prediction accuracy reaches the preset accuracy requirement, the training is terminated, and the trained sensor error prediction model is output.

[0075] In some embodiments of the present application, after determining the true correction error of the target sensor based on the high-precision positioning information and the positioning information of the target sensor, the method further includes: performing fitting optimization using a nonlinear optimization algorithm based on the original positioning error information and the true correction error of the target sensor to obtain the correspondence between the original positioning error information and the true correction error of the target sensor.

[0076] In addition to using adaptive training to train the sensor error prediction model to predict the sensor correction error, the embodiment of the present application can also use certain nonlinear optimization algorithms such as least squares, LM (Levenberg-Marquardt), etc. to fit and optimize the correspondence between the original positioning error information and the actual correction error of the target sensor, and then the fitting optimization result can be used as the basis for subsequent prediction of the actual correction error of the target sensor.

[0077] In some embodiments of the present application, the sensor error prediction model adopts an LSTM long short-term memory network.

[0078] The sensor error prediction model of the embodiment of the present application can be trained using the structure of the LSTM (Long Short-Term Memory) long short-term memory network. LSTM is a special recurrent neural network that has the ability to learn long-term dependencies. The ability of LSTM to remember information for a long time is actually its own attribute, not an ability acquired through learning or training. The SLAM covariance and the corresponding original correction amount in the embodiment of the present application are parameters related to time series, so the use of the LSTM network structure for training is more in line with the needs of actual scenarios.

[0079] Of course, it should be noted that in addition to using the LSTM network to train the sensor error prediction model, the traditional BP (Back Propagation) neural network can also be used for training. Those skilled in the art can flexibly choose according to actual needs, and they are not listed here one by one.

[0080] The present application also provides a fusion positioning method for an autonomous driving vehicle, such as Figure 2 As shown, a flow chart of a fusion positioning method for an autonomous driving vehicle in an embodiment of the present application is provided, wherein the method includes at least the following steps S210 to S240:

[0081] Step S210: obtaining current target sensor information, where the current target sensor information includes current positioning information and current original positioning error information;

[0082] Step S220, predicting a true correction error of the target sensor using a sensor error prediction model based on the current target sensor information;

[0083] Step S230, correcting the current positioning information using the true correction error of the target sensor to obtain corrected positioning information;

[0084] Step S240, performing fusion positioning based on the corrected positioning information to obtain a fusion positioning result of the autonomous driving vehicle;

[0085] The sensor error prediction model is trained based on any of the aforementioned sensor error prediction model training methods.

[0086] In actual fusion positioning scenarios, it is necessary to first obtain the current target sensor information, including the current positioning information such as laser SLAM / visual SLAM positioning information, and the current original positioning error information, such as SLAM covariance and original correction value.

[0087] The above information is input into the sensor error prediction model trained in the above embodiment to predict the true correction error (dx', dy') of the target sensor. The true correction error is used to correct the current positioning information (dx, dy) to obtain the corrected positioning information (dx1, dy1), which can be expressed as follows:

[0088] dx1=dx+dx'

[0089] dy1=dy+dy'

[0090] Finally, the corrected positioning information can be used as observation information and input into the Kalman filter or extended Kalman filter together with the positioning data of other sensors such as IMU positioning data and wheel speed data for fusion positioning, thereby obtaining the fusion positioning result of the autonomous driving vehicle.

[0091] The embodiment of the present application also provides a training device 300 for a sensor error prediction model for autonomous driving, such as Figure 3 As shown, a schematic diagram of the structure of a training device for a sensor error prediction model for autonomous driving according to an embodiment of the present application is provided. The device 300 includes: a first acquisition unit 310, a second acquisition unit 320, a determination unit 330, and a training unit 340, wherein:

[0092] A first acquisition unit 310 is configured to acquire high-precision positioning information of the autonomous driving vehicle;

[0093] A second acquisition unit 320 is configured to acquire target sensor information of the autonomous driving vehicle when the high-precision positioning information meets a preset training condition, the target sensor information including positioning information and original positioning error information of the target sensor;

[0094] a determining unit 330, configured to determine a true correction error of the target sensor based on the high-precision positioning information and the positioning information of the target sensor;

[0095] The training unit 340 is configured to train a sensor error prediction model using the original positioning error information and the actual correction error of the target sensor to obtain a trained sensor error prediction model.

[0096] In some embodiments of the present application, the second acquisition unit 320 is specifically used to: use a positioning confidence prediction model to predict the confidence of the high-precision positioning information; if the confidence of the high-precision positioning information is greater than a preset confidence threshold, it is determined that the high-precision positioning information meets the preset training condition; otherwise, it is determined that the high-precision positioning information does not meet the preset training condition.

[0097] In some embodiments of the present application, the original positioning error information includes SLAM covariance and original correction amount, and the training unit 340 is specifically used to: use the SLAM covariance and the original correction amount as input of the sensor error prediction model, use the actual correction error of the target sensor as the supervision signal, train the sensor error prediction model, and obtain a trained sensor error prediction model.

[0098] In some embodiments of the present application, the device also includes: a fitting unit, which is used to perform fitting optimization based on the original positioning error information and the actual correction error of the target sensor using a nonlinear optimization algorithm to obtain the corresponding relationship between the original positioning error information and the actual correction error of the target sensor.

[0099] In some embodiments of the present application, the sensor error prediction model adopts an LSTM long short-term memory network.

[0100] It can be understood that the above-mentioned training device for the sensor error prediction model for autonomous driving can implement the various steps of the training method for the sensor error prediction model for autonomous driving provided in the aforementioned embodiments. The relevant explanations on the training method for the sensor error prediction model for autonomous driving are applicable to the training device for the sensor error prediction model for autonomous driving and will not be repeated here.

[0101] The embodiment of the present application also provides a fusion positioning device 400 for an autonomous driving vehicle, such as Figure 4As shown, a schematic diagram of the structure of a fusion positioning device for an autonomous driving vehicle in an embodiment of the present application is provided. The device 400 includes: a third acquisition unit 410, a prediction unit 420, a correction unit 430, and a fusion positioning unit 440, wherein:

[0102] A third acquiring unit 410 is configured to acquire current target sensor information, where the current target sensor information includes current positioning information and current original positioning error information;

[0103] A prediction unit 420 is configured to predict a true correction error of the target sensor using a sensor error prediction model based on the current target sensor information;

[0104] A correction unit 430 is configured to correct the current positioning information using the actual correction error of the target sensor to obtain corrected positioning information;

[0105] a fusion positioning unit 440, configured to perform fusion positioning based on the corrected positioning information to obtain a fusion positioning result of the autonomous driving vehicle;

[0106] The sensor error prediction model is obtained by training based on the aforementioned sensor error prediction model training device.

[0107] It can be understood that the above-mentioned fusion positioning device of the autonomous driving vehicle can implement the various steps of the fusion positioning method of the autonomous driving vehicle provided in the aforementioned embodiments. The relevant explanations on the fusion positioning method of the autonomous driving vehicle are applicable to the fusion positioning device of the autonomous driving vehicle and will not be repeated here.

[0108] Figure 5 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present application. Figure 5 At the hardware level, the electronic device includes a processor and, optionally, an internal bus, a network interface, and memory. The memory may include internal memory, such as high-speed random-access memory (RAM), or non-volatile memory, such as at least one disk drive. Of course, the electronic device may also include other hardware required for its services.

[0109] The processor, network interface, and memory can be interconnected via an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 5 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0110] The memory is used to store programs. Specifically, the program may include program code, which includes computer operating instructions. The memory may include internal memory and non-volatile memory, and provides instructions and data to the processor.

[0111] The processor reads the corresponding computer program from the non-volatile memory into the internal memory and then runs it, forming a training device for the sensor error prediction model for autonomous driving at the logical level. The processor executes the program stored in the memory and is specifically used to perform the following operations:

[0112] Obtain high-precision positioning information for autonomous vehicles;

[0113] When the high-precision positioning information satisfies a preset training condition, obtaining target sensor information of the autonomous driving vehicle, the target sensor information including positioning information and original positioning error information of the target sensor;

[0114] determining a true correction error of the target sensor based on the high-precision positioning information and the positioning information of the target sensor;

[0115] The original positioning error information and the true correction error of the target sensor are used to train a sensor error prediction model to obtain a trained sensor error prediction model.

[0116] The above application Figure 1The method performed by the training device for a sensor error prediction model for autonomous driving disclosed in the illustrated embodiment can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by hardware integrated logic circuits in the processor or by software instructions. The above processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in conjunction with the embodiments of this application can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium well-known in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. The storage medium is located in the memory, and the processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above method.

[0117] The electronic device may also perform Figure 1 A method for executing a training device for a sensor error prediction model for autonomous driving, and implementing a training device for a sensor error prediction model for autonomous driving in Figure 1 The functions of the illustrated embodiment will not be described in detail in the embodiments of the present application.

[0118] The embodiment of the present application also provides a computer-readable storage medium, which stores one or more programs, wherein the one or more programs include instructions, which, when executed by an electronic device including multiple application programs, can enable the electronic device to execute Figure 1 The method performed by the training device for the sensor error prediction model for autonomous driving in the illustrated embodiment is specifically used to perform:

[0119] Obtain high-precision positioning information for autonomous vehicles;

[0120] When the high-precision positioning information satisfies a preset training condition, obtaining target sensor information of the autonomous driving vehicle, the target sensor information including positioning information and original positioning error information of the target sensor;

[0121] determining a true correction error of the target sensor based on the high-precision positioning information and the positioning information of the target sensor;

[0122] The original positioning error information and the true correction error of the target sensor are used to train a sensor error prediction model to obtain a trained sensor error prediction model.

[0123] It should be noted that the electronic device of the embodiment of the present application can also be used to perform Figure 2 The method performed by the fusion positioning device of the autonomous driving vehicle disclosed in the illustrated embodiment will not be described in detail.

[0124] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0125] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0126] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0127] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0128] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0129] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0130] Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0131] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0132] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0133] The foregoing is merely an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

Claims

1. A training method for a sensor error prediction model for autonomous driving, wherein: The method comprises: Obtain high-precision positioning information for autonomous vehicles; When the high-precision positioning information satisfies a preset training condition, obtaining target sensor information of the autonomous driving vehicle, the target sensor information including positioning information and original positioning error information of the target sensor; determining a true correction error of the target sensor based on the high-precision positioning information and the positioning information of the target sensor; Using the original positioning error information and the true correction error of the target sensor, a sensor error prediction model is trained to obtain a trained sensor error prediction model; The original positioning error information includes SLAM covariance and original correction value, and the sensor error prediction model is trained by using the original positioning error information and the real correction error of the target sensor to obtain the trained sensor error prediction model. The SLAM covariance and the original correction amount are used as inputs of the sensor error prediction model, and the true correction error of the target sensor is used as a supervision signal to train the sensor error prediction model to obtain a trained sensor error prediction model.

2. The method according to claim 1, wherein: When the high-precision positioning information satisfies a preset training condition, obtaining target sensor information of the autonomous driving vehicle includes: Predicting the confidence of the high-precision positioning information using a positioning confidence prediction model; If the confidence level of the high-precision positioning information is greater than a preset confidence threshold, determining that the high-precision positioning information meets the preset training condition; Otherwise, it is determined that the high-precision positioning information does not meet the preset training condition.

3. The method according to claim 1, wherein: After determining a true correction error of the target sensor based on the high-precision positioning information and the positioning information of the target sensor, the method further includes: Based on the original positioning error information and the true correction error of the target sensor, a nonlinear optimization algorithm is used to perform fitting optimization to obtain a corresponding relationship between the original positioning error information and the true correction error of the target sensor.

4. The method according to any one of claims 1 to 3, wherein: The sensor error prediction model adopts LSTM long short-term memory network.

5. A fusion positioning method for an autonomous driving vehicle, wherein: The method comprises: Acquire current target sensor information, where the current target sensor information includes current positioning information and current original positioning error information; Predicting a true correction error of the target sensor using a sensor error prediction model based on the current target sensor information; Correcting the current positioning information using the true correction error of the target sensor to obtain corrected positioning information; Performing fusion positioning based on the corrected positioning information to obtain a fusion positioning result of the autonomous driving vehicle; The sensor error prediction model is trained based on the sensor error prediction model training method according to any one of claims 1 to 4.

6. A training device for a sensor error prediction model for autonomous driving, wherein: The device comprises: A first acquisition unit is used to obtain high-precision positioning information of the autonomous driving vehicle; a second acquiring unit, configured to acquire target sensor information of the autonomous driving vehicle when the high-precision positioning information satisfies a preset training condition, the target sensor information including positioning information and original positioning error information of the target sensor; a determining unit, configured to determine a true correction error of the target sensor based on the high-precision positioning information and the positioning information of the target sensor; a training unit, configured to train a sensor error prediction model using the original positioning error information and a true correction error of the target sensor to obtain a trained sensor error prediction model; The original positioning error information includes SLAM covariance and original correction value, and the training unit is specifically used for: The SLAM covariance and the original correction amount are used as inputs of the sensor error prediction model, and the true correction error of the target sensor is used as a supervision signal to train the sensor error prediction model to obtain a trained sensor error prediction model.

7. A fusion positioning device for an autonomous driving vehicle, wherein: The device comprises: a third acquiring unit, configured to acquire current target sensor information, wherein the current target sensor information includes current positioning information and current original positioning error information; a prediction unit, configured to predict a true correction error of the target sensor using a sensor error prediction model based on the current target sensor information; A correction unit, configured to correct the current positioning information using a true correction error of the target sensor to obtain corrected positioning information; a fusion positioning unit, configured to perform fusion positioning based on the corrected positioning information to obtain a fusion positioning result of the autonomous driving vehicle; The sensor error prediction model is obtained by training based on the sensor error prediction model training device according to claim 6.

8. An electronic device comprising: processor; as well as A memory arranged to store computer-executable instructions, which, when executed, cause the processor to execute the training method of the sensor error prediction model for autonomous driving as described in any one of claims 1 to 4, or the fusion positioning method for the autonomous driving vehicle as described in claim 5.

9. A computer-readable storage medium storing one or more programs, which, when executed by an electronic device comprising a plurality of applications, enables the electronic device to execute the training method of a sensor error prediction model for autonomous driving as described in any one of claims 1 to 4, or to execute the fusion positioning method for an autonomous driving vehicle as described in claim 5.

Images