Automatic driving target positioning method and system based on multi-source data fusion

Abstract

The application discloses an automatic driving target positioning method and system based on multi-source data fusion, and relates to the technical field of data fusion.The method comprises the following steps: performing multi-source perception fusion on a to-be-charged object of a mobile storage and charging device to obtain a target perception sequence; performing multi-level positioning on the to-be-charged object to obtain a target positioning first feature and a target positioning second feature; performing perception interference correction to obtain a target positioning third feature and a target positioning fourth feature; performing driving level analysis on the mobile storage and charging device to construct a driving level factor; performing driving control decision on the mobile storage and charging device to obtain a first driving control space, and performing iterative optimization on the first driving control space to determine a driving control strategy; and performing driving management on the mobile storage and charging device and updating the driving control in combination with a multi-source perception device.The technical problem of inaccurate target positioning of an automatic driving mobile storage and charging device in the prior art is solved, and the technical effect of improving positioning accuracy is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data fusion, in particular to an automatic driving target positioning method and system based on multi-source data fusion. BACKGROUND

[0002] In the intelligent charging scene of electric vehicles, mobile storage and charging equipment can be flexibly moved to the place where charging is needed to provide convenient charging services for new energy vehicles, effectively alleviating the problems of insufficient layout of fixed charging piles, long charging queue time, etc. However, the existing mobile storage and charging equipment usually relies on a single type of sensor for environment perception and target positioning during automatic driving. Due to the limited perception source and the influence of environmental light, shielding, noise interference and other factors, the target positioning accuracy is insufficient, and there are problems such as charging object recognition deviation, unstable path planning and poor reliability of driving control strategy. This not only reduces the work efficiency of the automatic driving storage and charging equipment in complex scenes, but also affects the safety and stability of the overall operation. SUMMARY

[0003] The present application provides an automatic driving target positioning method and system based on multi-source data fusion, which solves the technical problem of inaccurate target positioning of the existing automatic driving mobile storage and charging equipment due to a single perception method.

[0004] In a first aspect, the present application provides an automatic driving target positioning method based on multi-source data fusion, which comprises:

[0005] The target perception sequence is obtained by multi-source perception fusion of the to-be-charged object of the mobile storage and charging equipment by the multi-source sensor. The target positioning first feature and the target positioning second feature are obtained by multi-level positioning of the to-be-charged object according to the target perception sequence. The target positioning third feature and the target positioning fourth feature are obtained by perception interference correction of the target positioning first feature and the target positioning second feature based on the multi-source sensor. The driving level factor is constructed by driving level analysis of the mobile storage and charging equipment according to the target positioning third feature and the target positioning fourth feature. The driving control strategy is determined by iterative optimization of the first driving control space according to the driving risk analysis model based on the driving control decision of the mobile storage and charging equipment based on the driving level factor. The driving management of the mobile storage and charging equipment is based on the driving control strategy, and the updated driving control is combined with the multi-source sensor.

[0006] In a second aspect, the present application provides an automatic driving target positioning system based on multi-source data fusion, which comprises:

[0007] The multi-source perception fusion module: through the multi-source perception of the mobile charging device by the multi-source perception device, a target perception sequence is obtained; the multi-level positioning module: according to the target perception sequence, the target charging object is positioned in multiple levels, and target positioning first features and target positioning second features are obtained; the perception interference correction module: based on the multi-source perception device, the target positioning first features and the target positioning second features are corrected, and target positioning third features and target positioning fourth features are obtained; the driving level analysis module: according to the target positioning third features and the target positioning fourth features, the mobile charging device is analyzed in driving level, and a driving level factor is constructed; the driving control decision module: based on the driving level factor, the mobile charging device is controlled and decided, a first driving control space is obtained, and the first driving control space is iteratively optimized according to a driving risk analysis model, and a driving control strategy is determined; the driving management module: based on the driving control strategy, the mobile charging device is managed and driven, and the driving control is updated in combination with the multi-source perception device.

[0008] One or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0009] Firstly, through the multi-source perception fusion of the mobile charging device by the multi-source perception device, a target perception sequence is obtained. Then, according to the target perception sequence, the target charging object is positioned in multiple levels, and target positioning first features and target positioning second features are obtained. Further, based on the multi-source perception device, the target positioning first features and the target positioning second features are corrected, and target positioning third features and target positioning fourth features are obtained. Next, according to the target positioning third features and the target positioning fourth features, the mobile charging device is analyzed in driving level, and a driving level factor is constructed. Then, based on the driving level factor, the mobile charging device is controlled and decided, a first driving control space is obtained, and the first driving control space is iteratively optimized according to a driving risk analysis model, and a driving control strategy is determined. Finally, based on the driving control strategy, the mobile charging device is managed and driven, and the driving control is updated in combination with the multi-source perception device. The technical problem of inaccurate target positioning caused by single perception means in the prior art is solved, and the technical effect of improving target positioning accuracy through multi-source data fusion and interference correction is achieved. BRIEF DESCRIPTION OF DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0011] Figure 1 A flowchart of an automatic driving target positioning method based on multi-source data fusion is provided for the embodiments of the present application.

[0012] Figure 2 A structural diagram of an automatic driving target positioning system based on multi-source data fusion is provided for the embodiments of the present application.

[0013] Legend: multi-source perception fusion module 11, multi-level positioning module 12, perception interference correction module 13, driving level analysis module 14, driving control decision module 15, and driving management module 16. DETAILED DESCRIPTION

[0014] The present application provides an automatic driving target positioning method and system based on multi-source data fusion, which solves the technical problem of inaccurate target positioning of automatic driving mobile charging equipment caused by single perception means in the prior art.

[0015] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0016] It should be noted that the terms "comprise" and "have" are intended to cover non-exclusive inclusion, for example, a process, method, system, product or server comprising a series of steps or units does not have to be limited to those clearly listed steps or units, but can include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices.

[0017] Embodiment one, as shown in the present application provides an automatic driving target positioning method based on multi-source data fusion, wherein the method comprises: Figure 1

[0018] The multi-source perception fusion module 11 performs multi-source perception fusion on the to-be-charged object of the mobile charging equipment through the multi-source perception device, and obtains a target perception sequence.

[0019] ​The multi-source sensor can be composed of a laser radar, a millimeter wave radar, an infrared imaging sensor, a depth camera, and an ultrasonic sensor, and is used for omnidirectional and heterogeneous environmental perception of the object to be charged during driving of the mobile charging device. The laser radar is used to collect three-dimensional point cloud data of the object to be charged, and obtain the spatial geometric shape and relative distance of the object. The millimeter wave radar is used to obtain the speed information and motion trajectory of the object in a low-visibility or insufficient-light scene. The infrared imaging sensor is used to capture the thermal features of the object in the night or complex light conditions, and realize thermal signal compensation. The depth camera is used to collect two-dimensional images and depth information of the object, and enhance the geometric contour recognition capability. The ultrasonic sensor is used for boundary detection and obstacle recognition in a short distance range.

[0020] By synchronizing the time stamps of the original data streams collected by the multi-source sensor and cleaning the data, redundant and noise information is removed to obtain a multi-source perception effective data set. The multi-source perception effective data set is subjected to feature extraction and space-time alignment, and the three-dimensional point cloud features, motion trajectory features, thermal features, image features, and boundary features are weighted and fused according to a preset fusion algorithm to generate a target perception sequence. The target perception sequence is used to represent the multi-dimensional perception information of the spatial position, attitude, motion trend, and charging interface region of the object to be charged.

[0021] Further, the target perception sequence is obtained by multi-source perception fusion of the object to be charged of the mobile charging device by the multi-source sensor, including:

[0022] According to the real-time perception of the object to be charged by the multi-source sensor, a target perception first set is obtained. The target perception first set is cleaned to obtain a target perception second set. The target perception second set is fused to generate the target perception sequence.

[0023] Based on the real-time perception of the object to be charged by the multi-source sensor, spatial three-dimensional point cloud information, motion speed information, appearance image information, and short-distance boundary information of the object to be charged are collected to form a target perception first set. The target perception first set is subjected to data cleaning processing, specifically including: time stamp alignment and spatial coordinate system unification of different perception source data, removal of abnormal values, noise points, and redundant information, and correction of missing data by interpolation compensation and filtering algorithm to generate a target perception second set with higher quality. The target perception second set is subjected to multi-source fusion processing. The fusion method can be feature-level fusion based on weight distribution. The fusion result can comprehensively reflect the spatial position, attitude feature, charging interface position, and dynamic change trend of the object to be charged, thereby generating the target perception sequence.

[0024] The feature-level fusion based on weight distribution includes: feature extraction on the perception data from the laser radar, millimeter wave radar, camera, infrared sensor and ultrasonic sensor to obtain point cloud features, speed features, image features, thermal imaging features and boundary features respectively; setting corresponding weight factors according to the reliability of different sensors in positioning accuracy, environmental adaptability and anti-interference ability, for example, assigning a higher spatial geometric weight to the laser radar point cloud features, assigning a dynamic tracking weight to the millimeter wave radar motion features, and assigning an appearance recognition and light compensation weight to the camera and infrared sensor; then, in a unified time coordinate and spatial coordinate system, the above features are weighted and superimposed or fused according to their respective weights to obtain a fused multi-dimensional comprehensive feature vector.

[0025] According to the target perception sequence, multi-level positioning is performed on the object to be charged to obtain target positioning first features and target positioning second features.

[0026] Further, the target positioning first features include the geographic position of the object to be charged, and the target positioning second features include the charging interface position of the object to be charged.

[0027] Based on the three-dimensional point cloud features and motion trajectory features acquired by the laser radar and the millimeter wave radar in the target perception sequence, combined with the global positioning system (such as GNSS / IMU) data of the mobile charging device, the global spatial position of the object to be charged is estimated to determine its absolute coordinates in the road or station, and the target positioning first features, i.e., the geographic position of the object to be charged, are obtained.

[0028] Based on the images and thermal features collected by the depth camera and infrared sensor in the target perception sequence, the geometric shape of the object to be charged is profiled and locally recognized, the charging interface area of the object to be charged is identified, and the charging interface area is finely positioned in a unified coordinate system. The spatial position of the charging interface is obtained through edge detection and region matching, and the target positioning second features, i.e., the charging interface position of the object to be charged, are obtained.

[0029] Based on the multi-source perception of the target positioning first features and the target positioning second features, the target positioning third features and the target positioning fourth features are obtained.

[0030] Further, based on the multi-source perception of the target positioning first features and the target positioning second features, the target positioning third features and the target positioning fourth features are obtained, including:

[0031] The multi-source perception device is subjected to perception scene collection based on the target perception sequence, to obtain a plurality of perception scene data; interference factors of the multi-source perception device are identified based on the plurality of perception scene data, to obtain a perception interference identification result; the target positioning first feature is subjected to positioning error correction based on the perception interference identification result, to obtain a target positioning third feature; and the target positioning second feature is subjected to positioning error correction based on the perception interference identification result, to obtain a target positioning fourth feature.

[0032] The multi-source perception device is subjected to perception scene collection based on the target perception sequence, specifically, during the operation of the mobile storage and charging device, sensors such as laser radars, millimeter wave radars, cameras, depth sensors, and infrared sensors are used to continuously collect environmental data of the object to be charged under different time periods, different light conditions, and different spatial positions, to form a plurality of perception scene data sets; the output results of the multi-source perception device are subjected to interference factor identification based on the plurality of perception scene data, wherein the interference factors include data abnormalities caused by insufficient light, existence of an obstacle, reflection interference, electromagnetic noise, and multipath effect, and the interference factor identification can be achieved through feature residual detection, statistical threshold determination, or an interference classification model based on deep learning, to obtain a perception interference identification result; the target positioning first feature is subjected to positioning error correction based on the perception interference identification result, for example, the positioning result is corrected through Kalman filtering, weighted average, or least square estimation, to obtain a target positioning third feature; and the target positioning second feature is subjected to positioning error correction based on the perception interference identification result, for example, the positioning result is corrected through edge detection correction, region matching optimization, or an interface recognition enhancement model based on a convolutional neural network, to obtain a target positioning fourth feature.

[0033] The mobile storage and charging device is subjected to driving level analysis based on the target positioning third feature and the target positioning fourth feature, to construct a driving level factor.

[0034] The third target positioning feature is used to determine the geographical position of the object to be charged in the global environment, and a macroscopic navigation path from the current position of the device to the target position is constructed in combination with the current geographical coordinates of the mobile storage and charging device. The fourth target positioning feature is used to determine the accurate position of the charging interface of the object to be charged in space, and a microscopic docking path is established in combination with the three-dimensional structural features of the target object. The macroscopic navigation path is used as the first driving level, and the microscopic docking path is used as the second driving level. The two are hierarchically analyzed to obtain a multi-level driving task from global driving to accurate docking. On this basis, the path elements from the current position of the device to the target position are used as the first driving factor, and the path elements from the external docking position of the target object to the position of the charging interface are used as the second driving factor. The two driving factors are modified and normalized according to the constraint conditions such as the distribution of environmental obstacles, the turning radius of the device, and the docking tolerance range, and finally the driving level factor is constructed.

[0035] Further, the driving level analysis of the mobile storage and charging device according to the third target positioning feature and the fourth target positioning feature, and the construction of the driving level factor, include:

[0036] According to the three-dimensional structural features of the object to be charged and the space features, a target three-dimensional space is constructed. Based on the target three-dimensional space, the end point tolerance fitting of the mobile storage and charging device is performed according to the third target positioning feature, and the end point tolerance region is obtained. The end point tolerance region is corrected according to the device structural features of the mobile storage and charging device, and the end point tolerance position is generated. The device geographical position of the mobile storage and charging device to the end point tolerance position is used as the first driving factor. The external charging position of the device corresponding to the end point tolerance position to the fourth target positioning feature is used as the second driving factor, and the first driving factor and the second driving factor are added to the driving level factor.

[0037] Based on the three-dimensional structural features of the object to be charged (such as the geometric shape and size parameters of the vehicle or device) and the space features (such as the parking area boundary, the distribution of surrounding obstacles, and the ground markings), a target three-dimensional space is constructed to describe the spatial topological relationship between the object to be charged and the surrounding environment.

[0038] The navigation end point of the mobile storage and charging device is tolerantly fitted in the target three-dimensional space based on the geographical position of the object to be charged represented by the target positioning third feature, that is, an allowed end point candidate area is established within the charging docking area range, and the end point tolerance area is obtained by setting a position deviation threshold and a docking tolerance parameter; combined with the device structure characteristics of the mobile storage and charging device (such as the minimum turning radius, the size of the device, and the telescopic range of the docking arm), the accessibility of the end point tolerance area is corrected, and the candidate area that cannot meet the device structure constraint is eliminated, thereby generating the end point tolerance position; the path elements from the device geographical position of the mobile storage and charging device to the end point tolerance position are taken as the first driving factor, which is used to describe the macroscopic driving path; the path elements from the device external charging position (the interface position on the mobile storage and charging device for energy exchange with the object to be charged) corresponding to the end point tolerance position to the target positioning fourth feature are taken as the second driving factor, which is used to describe the microscopic precise docking path; the first driving factor and the second driving factor are integrated and added to the driving level factor, so that the driving level factor can reflect the hierarchical control requirements of global navigation and local docking at the same time.

[0039] Based on the driving level factor, a driving control decision is made for the mobile storage and charging device to obtain a first driving control space, and an iterative optimization is performed on the first driving control space according to a driving risk analysis model to determine a driving control strategy.

[0040] Based on the multi-modal control constraint information of the mobile storage and charging device, including speed constraint, steering constraint, path smoothness constraint, and device docking accuracy constraint, a driving control constraint condition is constructed; based on the driving control constraint condition, a multi-objective decision analysis is performed on the driving level factor to generate a plurality of feasible driving control candidate schemes, and these schemes are combined to form a first driving control space. A driving risk analysis model is called to evaluate the risk of each candidate scheme in the first driving control space, wherein the driving risk analysis model includes multi-dimensional risk evaluation indexes such as driving safety risk, docking abnormal risk, and driving failure risk, and outputs a corresponding risk constraint matrix; according to the risk constraint matrix, an iterative optimization is performed on the first driving control space, and candidate schemes that do not meet the safety and reliability requirements are gradually eliminated, while the remaining schemes are risk sorted and optimized to obtain a driving control strategy that meets the driving safety, docking accuracy, and operation efficiency.

[0041] Further, based on the driving level factor, a driving control decision is made for the mobile storage and charging device to obtain a first driving control space, including:

[0042] The first driving factor and the second driving factor are used to collect driving scene features, and a first driving scene model is constructed. The first driving scene model is used to label obstacle features, and a second driving scene model is obtained. Multi-modal driving control constraint information of the mobile storage and charging device is collected, and driving control constraint conditions are generated. The first driving factor is subjected to driving control decision-making based on the second driving scene model and the driving control constraint conditions, and a first driving control decision set is obtained. The second driving factor is subjected to driving control decision-making based on the second driving scene model and the driving control constraint conditions, and a second driving control decision set is obtained. The first driving control decision set and the second driving control decision set are randomly combined to generate the first driving control space.

[0043] Based on the first driving factor and the second driving factor, the environmental perception data is collected for feature collection, and the road boundary, environmental obstacle distribution and target area geometric constraint are extracted to construct a preliminary first driving scene model. In the first driving scene model, the potential obstacles are labeled, for example, through laser radar point cloud clustering, camera target recognition or millimeter wave radar dynamic tracking, the position, size and motion trend of the obstacles are labeled and classified, and a second driving scene model enhanced by obstacles is obtained.

[0044] The multi-modal driving control constraint information of the mobile storage and charging device is collected, including the upper limit of the speed, the steering radius, the acceleration and deceleration smoothness, the path curvature threshold and the docking end posture constraint, and driving control constraint conditions are generated accordingly.

[0045] Based on the driving control constraint conditions, the first driving factor is subjected to control decision-making in combination with the second driving scene model, and a first driving control decision set is generated. The first driving control decision set is used to describe multiple feasible paths and control instructions of the device from the current position to the tolerant position of the endpoint. At the same time, based on the same constraint conditions and the second driving scene model, the second driving factor is subjected to control decision-making, and a second driving control decision set is generated. The second driving control decision set is used to describe multiple refined docking paths and control instructions of the device end during the docking process from the external charging position to the charging interface position. The first driving control decision set and the second driving control decision set are randomly combined to generate a first driving control space, so that it can cover the feasible control scheme set of different levels of driving factors under the constraint conditions.

[0046] Further, the first driving control space is subjected to iterative optimization according to the driving risk analysis model, and a driving control strategy is determined, including:

[0047] The driving risk analysis model is a multi-dimensional risk evaluation model based on machine learning, which is used for quantitative evaluation of multiple risks that may occur during driving. Optionally, the running data of the mobile storage and charging equipment in different driving scenarios is collected, including vehicle trajectory data, speed and acceleration data, obstacle approach distance data, charging interface docking deviation data, sensor state data, and control execution feedback data; a multi-dimensional risk label system is established based on the running data, and the risk label at least includes a driving safety risk label, a docking abnormal risk label, and a driving failure risk label, wherein the driving safety risk label is used to mark the event of collision or deviation of the vehicle during driving, the docking abnormal risk label is used to mark the event of deviation or docking failure in the charging interface recognition and docking process, and the driving failure risk label is used to mark the driving error event caused by sensor failure or controller abnormality. Feature engineering processing is performed on the running data, including data cleaning, time series segmentation, feature extraction and normalization, to obtain a high-dimensional feature vector representing the relationship between the vehicle operating state and the risk event; on this basis, a machine learning method is used to model the high-dimensional feature vector and the risk label, and a random forest model, a support vector machine model, a deep neural network model or an ensemble learning model can be optionally used for training to obtain a driving risk analysis model capable of predicting driving safety risk, docking abnormal risk and driving failure risk.

[0048] The driving risk analysis model is a multi-dimensional risk evaluation model based on machine learning, which is used for quantitative evaluation of multiple risks that may occur during driving. Optionally, the running data of the mobile storage and charging equipment in different driving scenarios is collected, including vehicle trajectory data, speed and acceleration data, obstacle approach distance data, charging interface docking deviation data, sensor state data, and control execution feedback data; a multi-dimensional risk label system is established based on the running data, and the risk label at least includes a driving safety risk label, a docking abnormal risk label, and a driving failure risk label, wherein the driving safety risk label is used to mark the event of collision or deviation of the vehicle during driving, the docking abnormal risk label is used to mark the event of deviation or docking failure in the charging interface recognition and docking process, and the driving failure risk label is used to mark the driving error event caused by sensor failure or controller abnormality. Feature engineering processing is performed on the running data, including data cleaning, time series segmentation, feature extraction and normalization, to obtain a high-dimensional feature vector representing the relationship between the vehicle operating state and the risk event; on this basis, a machine learning method is used to model the high-dimensional feature vector and the risk label, and a random forest model, a support vector machine model, a deep neural network model or an ensemble learning model can be optionally used for training to obtain a driving risk analysis model capable of predicting driving safety risk, docking abnormal risk and driving failure risk.

[0049] The multi-dimensional driving risk evaluation indexes in the driving risk analysis model are called, including driving safety risk, docking abnormal risk and driving failure risk, and constraint conditions are set for each type of risk evaluation index, such as setting a collision probability threshold constraint for driving safety risk, setting a docking deviation range constraint for docking abnormal risk, and setting an operation reliability threshold constraint for driving failure risk, and the above constraints are summarized to construct a driving risk constraint matrix. The driving risk constraint matrix is applied to the first driving control space, and the multiple driving control schemes in the first driving control space are screened one by one according to the risk constraints, and the schemes that do not meet the constraint conditions are removed, and a second driving control space that meets the constraint requirements is obtained.

[0050] Based on the multi-dimensional driving risk evaluation index, a weight configuration is performed to construct a driving comprehensive risk analysis model, that is, each risk index is weighted according to a preset importance ratio to construct a driving comprehensive risk function for calculating the comprehensive risk of any driving control scheme and outputting a driving comprehensive risk coefficient.

[0051] Driving comprehensive risk function: wherein R is a driving comprehensive risk coefficient, is a measurement value of the ith risk evaluation index, is a corresponding weight factor. Through the driving comprehensive risk function, all candidate schemes in the second driving control space can be uniformly quantitatively evaluated, and the driving comprehensive risk coefficient is taken as a sorting reference to sort from small to large, and the first K driving control candidate schemes are selected.

[0052] Based on the driving risk analysis model and the driving comprehensive risk analysis model, the K driving control candidate schemes are identified and optimized, new driving variation groups are generated through crossover and difference operators, and the third driving control space is established after risk constraint correction and comprehensive risk reordering; in the third driving control space, the multiple schemes in the third driving control space are iteratively optimized and path fitted to minimize the driving time, and the driving control scheme that meets the safety, reliability and time optimization is selected as the driving control strategy.

[0053] Further, based on the driving risk analysis model, the first driving control space is optimized according to the driving risk constraint matrix to establish a second driving control space, including:

[0054] extracting a driving control W-th scheme from the first driving control space, W being a positive integer; simulating driving of the mobile charging device based on the driving control W-th scheme to obtain W-th simulated driving data; inputting the W-th simulated driving data into the driving risk analysis model to obtain a W-th driving risk matrix; if the W-th driving risk matrix satisfies the driving risk constraint matrix, adding the driving control W-th scheme to the second driving control space; and continuing to perform driving risk constraint optimization on the first driving control space based on the driving risk analysis model and the driving risk constraint matrix to obtain the second driving control space.

[0055] extracting a driving control W-th scheme from the first driving control space; simulating driving of the mobile charging device based on the driving control W-th scheme to obtain W-th simulated driving data; inputting the W-th simulated driving data into the driving risk analysis model to obtain a W-th driving risk matrix; if the W-th driving risk matrix satisfies the driving risk constraint matrix, adding the driving control W-th scheme to the second driving control space; and continuing to perform driving risk constraint optimization on the first driving control space based on the driving risk analysis model and the driving risk constraint matrix to obtain the second driving control space.

[0056] Further, based on the driving risk analysis model and the driving comprehensive risk analysis model, performing mutation optimization on the K driving control candidate schemes to establish a third driving control space, including:

[0057] performing difference identification on the second driving control space based on the K driving control candidate schemes to obtain a plurality of driving control difference vectors; performing cross mutation on the second driving control space based on the driving hierarchical factors and the plurality of driving control difference vectors to obtain a first group of driving mutations; performing driving risk constraint optimization on the first group of driving mutations based on the driving risk analysis model and the driving risk constraint matrix to obtain a second group of driving mutations; expanding the second group of driving mutations based on the K driving control candidate schemes to obtain a third group of driving mutations; and performing driving comprehensive risk sorting optimization on the third group of driving mutations based on the driving comprehensive risk analysis model to obtain the third driving control space.

[0058] The second driving control space is identified according to K driving control candidate schemes, specifically, the candidate schemes are compared in terms of path length, steering angle distribution, speed change curve, and end-to-end docking attitude, and a plurality of driving control difference vectors are obtained to represent feature differences between the candidate schemes.

[0059] Based on the driving level factor, the plurality of driving control difference vectors are introduced into the crossover mutation operator to perform crossover combination and disturbance on the second driving control space, to generate a new driving control scheme group, to obtain a driving variation first group, thereby expanding the diversity and coverage of the original candidate scheme. Based on the driving risk analysis model, the driving variation first group is risk screened and optimized according to a preset driving risk constraint matrix, to eliminate control schemes that do not meet the safety threshold, to establish a driving variation second group; the driving variation second group is expanded according to the initial K driving control candidate schemes, so as to have mixed characteristics of the original candidate scheme and the variation scheme, thereby obtaining a driving variation third group; based on the driving comprehensive risk analysis model, all schemes in the driving variation third group are driving comprehensive risk sorted and optimized, and are sequentially sorted in order of comprehensive risk value from small to large, and the top K driving control schemes are selected and added to the third driving control space, to obtain the third driving control space after variation and optimization.

[0060] Based on the driving control strategy, the mobile storage and charging equipment is driven and managed, and the multi-source sensor is combined to update the driving control.

[0061] The driving control strategy generated by iterative optimization is issued to the execution unit of the mobile storage and charging equipment, and the driving path planning, speed control, steering adjustment, and charging interface docking operation of the vehicle are managed in real time, so as to ensure that the mobile storage and charging equipment automatically drives and docks according to the optimal control strategy.

[0062] During the driving process of the mobile storage and charging equipment, the multi-source sensor continuously dynamically perceives the to-be-charged object and the surrounding environment, collects updated perception data, including real-time position offset of the to-be-charged object, charging interface attitude change, and dynamic change of environmental obstacles; then, the updated perception data and the current driving control strategy are compared, if there is a significant deviation or environmental sudden change, the driving control update process is triggered, that is, the updated perception data is taken as input, the target perception sequence generation, multi-level positioning, interference correction, driving level analysis, control space construction, and risk optimization steps are repeated, a new driving control strategy is generated in real time, and the original control instruction is replaced by the updated driving control strategy and continues to be executed, so as to realize the dynamic closed loop of the driving management process.

[0063] In summary, the embodiments of the present application have at least the following technical effects:

[0064] Firstly, the multi-source perception fusion module 11 performs multi-source perception fusion on the object to be charged of the mobile charging device through the multi-source perception device to obtain a target perception sequence. Secondly, the multi-level positioning module 12 performs multi-level positioning on the object to be charged according to the target perception sequence to obtain a target positioning first feature and a target positioning second feature. Further, the perception interference correction module 13 performs perception interference correction on the target positioning first feature and the target positioning second feature based on the multi-source perception device to obtain a target positioning third feature and a target positioning fourth feature. Next, the driving level analysis module 14 performs driving level analysis on the mobile charging device according to the target positioning third feature and the target positioning fourth feature to construct a driving level factor. Then, the driving control decision module 15 performs driving control decision on the mobile charging device based on the driving level factor to obtain a first driving control space, and performs iterative optimization on the first driving control space according to a driving risk analysis model to determine a driving control strategy. Finally, the driving management module 16 performs driving management on the mobile charging device based on the driving control strategy, and updates the driving control in combination with the multi-source perception device. The technical problem of inaccurate target positioning caused by single perception means in the prior art is solved, and the technical effect of improving the target positioning accuracy through multi-source data fusion and interference correction is achieved.

[0065] In the embodiment two, based on the same inventive concept as the automatic driving target positioning method based on multi-source data fusion in the foregoing embodiments, as shown in the following table, the present application provides an automatic driving target positioning system based on multi-source data fusion, wherein the system comprises: Figure 2

[0066] The multi-source perception fusion module 11 performs multi-source perception fusion on the object to be charged of the mobile charging device through the multi-source perception device to obtain a target perception sequence. The multi-level positioning module 12 performs multi-level positioning on the object to be charged according to the target perception sequence to obtain a target positioning first feature and a target positioning second feature. The perception interference correction module 13 performs perception interference correction on the target positioning first feature and the target positioning second feature based on the multi-source perception device to obtain a target positioning third feature and a target positioning fourth feature. The driving level analysis module 14 performs driving level analysis on the mobile charging device according to the target positioning third feature and the target positioning fourth feature to construct a driving level factor. The driving control decision module 15 performs driving control decision on the mobile charging device based on the driving level factor to obtain a first driving control space, and performs iterative optimization on the first driving control space according to a driving risk analysis model to determine a driving control strategy. The driving management module 16 performs driving management on the mobile charging device based on the driving control strategy, and updates the driving control in combination with the multi-source perception device.

[0067] Further, the perception interference correction module 13 is configured to perform the following method:

[0068] ​The multi-source perception device is used for collecting perception scenes based on the target perception sequence, and a plurality of perception scene data are obtained; the multi-source perception device is used for identifying interference factors according to the plurality of perception scene data, and a perception interference identification result is obtained; the target positioning first feature is corrected for positioning error according to the perception interference identification result, and a target positioning third feature is obtained; the target positioning second feature is corrected for positioning error according to the perception interference identification result, and a target positioning fourth feature is obtained.

[0069] Further, the driving level analysis module 14 is used for executing the following method:

[0070] According to the three-dimensional structure characteristics and the space characteristics of the to-be-charged object, a target three-dimensional space is constructed; based on the target three-dimensional space, the mobile storage and charging device is navigated to an end-tolerance fitting point according to the target positioning third feature, and an end-tolerance region is obtained; the end-tolerance region is corrected for reachability according to the device structure characteristics of the mobile storage and charging device, and an end-tolerance position is generated; the device geographical position of the mobile storage and charging device to the end-tolerance position is taken as a first driving factor; the external charging position corresponding to the end-tolerance position of the device to the target positioning fourth feature is taken as a second driving factor, and the first driving factor and the second driving factor are added to the driving level factor.

[0071] Further, the driving control decision module 15 is used for executing the following method:

[0072] A first driving scene model is constructed based on the first driving factor and the second driving factor for driving scene feature collection; a second driving scene model is obtained based on obstacle feature labeling of the first driving scene model; a plurality of modal driving control constraint information of the mobile storage and charging device is collected, and driving control constraint conditions are generated; the first driving factor is subjected to driving control decision based on the second driving scene model and the driving control constraint conditions, and a first driving control decision set is obtained; the second driving factor is subjected to driving control decision based on the second driving scene model and the driving control constraint conditions, and a second driving control decision set is obtained; the first driving control decision set and the second driving control decision set are subjected to random combination, and a first driving control space is generated.

[0073] Further, the driving control decision module 15 is used for executing the following method:

[0074] The driving risk analysis model is used for driving risk constraint optimization of the first driving control space based on the driving risk constraint matrix, and a second driving control space is established; the driving risk analysis model is used for driving risk constraint optimization of the second driving control space based on the driving risk constraint matrix, and a third driving control space is established; and the driving control strategy is generated by driving time minimization optimization based on the third driving control space.

[0075] Further, the driving control decision module 15 is configured to perform the following method:

[0076] The driving control decision module 15 is configured to perform the following method:

[0077] Further, the driving control decision module 15 is configured to perform the following method:

[0078] The driving control decision module 15 is configured to perform the following method:

[0079] Further, the multi-source perception fusion module 11 is configured to perform the following method:

[0080] According to real-time sensing of the object to be charged by the multi-source sensor, a target sensing first set is obtained; the target sensing first set is cleaned to obtain a target sensing second set; and the target sensing second set is fused to generate the target sensing sequence.

[0081] Further, the multi-level positioning module 12 is configured to perform the following method:

[0082] The target positioning first feature includes a geographic position of the object to be charged, and the target positioning second feature includes a charging interface position of the object to be charged.

[0083] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. The above-described specific embodiments of the present application are described. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are possible or can be advantageous.

[0084] The above only describes the preferred embodiments of the present application, and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0085] The specification and drawings are merely exemplary of the present application, and any and all modifications, variations, combinations or equivalents that fall within the scope of the present application are considered to be covered by the present application. Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technology, the present application is intended to include these modifications and variations.

Claims

1. An automatic driving target positioning method based on multi-source data fusion, characterized in that, The method comprises: obtaining a target perception sequence by multi-source perception fusion of a to-be-charged object of a mobile storage and charging device through a multi-source perception device; performing multi-level positioning on the to-be-charged object according to the target perception sequence to obtain a target positioning first feature and a target positioning second feature; performing perception interference correction on the target positioning first feature and the target positioning second feature based on the multi-source perception device to obtain a target positioning third feature and a target positioning fourth feature; performing driving level analysis on the mobile storage and charging device according to the target positioning third feature and the target positioning fourth feature to construct a driving level factor; performing driving control decision on the mobile storage and charging device based on the driving level factor to obtain a first driving control space, and performing iterative optimization on the first driving control space according to a driving risk analysis model to determine a driving control strategy; performing driving management on the mobile storage and charging device based on the driving control strategy, and updating the driving control in combination with the multi-source perception device; wherein, based on the multi-source perception device, the target positioning first feature and the target positioning second feature are corrected to obtain the target positioning third feature and the target positioning fourth feature, comprising: based on the target perception sequence, collecting perception scene data of the multi-source perception device to obtain a plurality of perception scene data; based on the plurality of perception scene data, identifying interference factors of the multi-source perception device to obtain a perception interference identification result; based on the perception interference identification result, correcting the positioning error of the target positioning first feature to obtain the target positioning third feature; based on the perception interference identification result, correcting the positioning error of the target positioning second feature to obtain the target positioning fourth feature. 2.The automatic driving target positioning method based on multi-source data fusion of claim 1, wherein, According to the target positioning third feature and the target positioning fourth feature, the mobile storage and charging device is analyzed at the driving level to construct a driving level factor, comprising: constructing a target three-dimensional space according to the three-dimensional structural features and the space features of the to-be-charged object; based on the target three-dimensional space, fitting the navigation end point tolerance of the mobile storage and charging device according to the target positioning third feature to obtain an end point tolerance position domain; correcting the end point tolerance position domain according to the device structural features of the mobile storage and charging device to generate an end point tolerance position; taking the device geographic position of the mobile storage and charging device to the end point tolerance position as a first driving factor; taking the device external charging position corresponding to the end point tolerance position to the target positioning fourth feature as a second driving factor, and adding the first driving factor and the second driving factor to the driving level factor. 3.The automatic driving target positioning method based on multi-source data fusion of claim 1, wherein, Based on the driving level factor, the mobile storage and charging device is controlled to make a driving control decision to obtain a first driving control space, comprising: performing driving scene feature collection based on the first driving factor and the second driving factor to construct a first driving scene model; performing obstacle feature labeling based on the first driving scene model to obtain a second driving scene model; collecting multi-modal driving control constraint information of the mobile storage and charging device to generate driving control constraint conditions; obtaining a first driving control decision set by making driving control decisions on the first driving factor according to the second driving scene model based on the driving control constraint condition; obtaining a second driving control decision set by making driving control decisions on the second driving factor according to the second driving scene model based on the driving control constraint condition; generating the first driving control space by randomly combining the first driving control decision set and the second driving control decision set. 4.The automatic driving target positioning method based on multi-source data fusion of claim 1, wherein, determining a driving control strategy by iteratively optimizing the first driving control space according to a driving risk analysis model, including: generating a driving risk constraint matrix by configuring constraints according to multi-dimensional driving risk evaluation indexes of the driving risk analysis model, the multi-dimensional driving risk evaluation indexes including driving safety risk, interfacing abnormality risk and driving failure risk; establishing a second driving control space by performing driving risk constraint optimization on the first driving control space according to the driving risk constraint matrix based on the driving risk analysis model; configuring weights based on the multi-dimensional driving risk evaluation indexes to obtain a driving comprehensive risk analysis model; obtaining K driving control candidate schemes by performing driving comprehensive risk sorting optimization on the second driving control space according to the driving comprehensive risk analysis model, K being a positive integer; establishing a third driving control space by performing mutation optimization on the K driving control candidate schemes based on the driving risk analysis model and the driving comprehensive risk analysis model; generating the driving control strategy by performing driving time minimization optimization according to the third driving control space. 5.The automatic driving target positioning method based on multi-source data fusion of claim 4, wherein, determining a driving control strategy by iteratively optimizing the first driving control space according to a driving risk analysis model, including: extracting a driving control Wth scheme from the first driving control space, W being a positive integer; obtaining Wth simulation driving data by simulating driving of the mobile storage and charging equipment based on the driving control Wth scheme; inputting the Wth simulation driving data into the driving risk analysis model to obtain a Wth driving risk matrix; if the Wth driving risk matrix meets the driving risk constraint matrix, adding the driving control Wth scheme to the second driving control space; continuing to perform driving risk constraint optimization on the first driving control space according to the driving risk analysis model and the driving risk constraint matrix to obtain the second driving control space. 6.The automatic driving target positioning method based on multi-source data fusion of claim 4, wherein, determining a driving control strategy by iteratively optimizing the first driving control space according to a driving risk analysis model, including: performing difference identification on the second driving control space according to the K driving control candidate schemes to obtain a plurality of driving control difference vectors; performing cross mutation on the second driving control space according to the plurality of driving control difference vectors based on the driving hierarchical factors to obtain a driving variation first group; performing driving risk constraint optimization on the driving variation first group according to the driving risk constraint matrix based on the driving risk analysis model to establish a driving variation second group; According to the K driving control candidate schemes, the driving variation second group is expanded to obtain a driving variation third group; According to the driving comprehensive risk analysis model, the driving variation third group is ranked and optimized in driving comprehensive risk to obtain a third driving control space. 7.The automatic driving target positioning method based on multi-source data fusion of claim 1, wherein, The target perception sequence is obtained by multi-source perception fusion of the to-be-charged object of the mobile storage and charging equipment through a multi-source perception device, including: The target perception first set is obtained by real-time perception of the to-be-charged object through the multi-source perception device; The target perception second set is obtained by cleaning the target perception first set; The target perception sequence is generated by fusing the target perception second set. 8.The automatic driving target positioning method based on multi-source data fusion of claim 1, wherein, The target positioning first feature includes the geographic position of the to-be-charged object, and the target positioning second feature includes the charging interface position of the to-be-charged object.

9. An automatic driving target positioning system based on multi-source data fusion, characterized in that, The system for implementing the automatic driving target positioning method based on multi-source data fusion according to any one of claims 1-8, the system comprising: A multi-source perception fusion module: the target perception sequence is obtained by multi-source perception fusion of the to-be-charged object of the mobile storage and charging equipment through a multi-source perception device; A multi-level positioning module: the target perception sequence is used to perform multi-level positioning on the to-be-charged object to obtain target positioning first features and target positioning second features; A perception interference correction module: the target positioning first features and the target positioning second features are corrected based on the multi-source perception device to obtain target positioning third features and target positioning fourth features; A driving level analysis module: the target positioning third features and the target positioning fourth features are used to analyze the driving level of the mobile storage and charging equipment to construct driving level factors; A driving control decision module: the driving level factors are used to make driving control decisions for the mobile storage and charging equipment to obtain a first driving control space, and the first driving control space is iteratively optimized based on a driving risk analysis model to determine a driving control strategy; A driving management module: the driving control strategy is used to manage the driving of the mobile storage and charging equipment, and the multi-source perception device is used to update the driving control.

Images