Map building method and map positioning method and device

Abstract

The invention discloses a map building method and a map positioning method and device. A guiding vehicle is deployed in the driving direction of the autonomous vehicle, and structural features provided by the guiding vehicle are utilized to assist in construction of a point cloud map. And obtaining point cloud data, IMU data and UWB data. Then, a point cloud map is constructed based on the point cloud data including the features of the guided vehicle and the IMU data. And then, based on the UWB data, determining a base station position of at least one UWB base station in the point cloud map, and adding the at least one base station position in the point cloud map to obtain the point cloud map comprising the at least one base station position. And finally, inputting the point cloud data into a target detection model to enable the target detection model to determine a dynamic obstacle in the point cloud data based on the point cloud data, and removing the dynamic obstacle from the point cloud map containing the at least one base station position to obtain a target map. Therefore, the position of the UWB base station and rich semantic information are integrated in the point cloud map.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method for constructing a map, a method for positioning a map, and a device. Background Art

[0002] With the development of information technology and the rapid development of unmanned vehicles and intelligent robots, the demand for automatic mapping, positioning and obstacle avoidance control technologies for unmanned vehicles and intelligent robots has become increasingly urgent.

[0003] At present, autonomous vehicles are usually positioned through the Global Navigation Satellite System (GNSS). However, in the actual environment, there are many areas with weak or no GNSS signals, such as mining areas, tunnels, garages, etc. In these areas, the GNSS system cannot provide positioning information. In order to map and locate these areas, the inertial measurement unit IMU and LiDAR fusion method can be used.

[0004] However, in some degraded scenes such as tunnels and mines where the structural features are not rich enough, the point cloud data collected by the autonomous vehicle during driving is often highly similar, resulting in low accuracy in mapping and positioning. In addition, the positioning information of the IMU is obtained through integral calculations, which will inevitably be affected by error accumulation after long-term operation, thus affecting the accuracy of map construction and map positioning. Summary of the invention

[0005] The present specification provides a method, device, storage medium and electronic device for map construction to improve the accuracy of map construction and the accuracy of subsequent map positioning.

[0006] This manual adopts the following technical solutions:

[0007] This specification provides a method for map construction, which is applied to an autonomous driving vehicle, wherein a guide vehicle is deployed in the driving direction of the autonomous driving vehicle, and the method comprises:

[0008] Acquire point cloud data collected by the laser radar on the autonomous driving vehicle, IMU data collected by the inertial measurement unit IMU, and UWB data received by the ultra-wideband UWB tag;

[0009] constructing a point cloud map based on the point cloud data including the features of the guided vehicle and the IMU data;

[0010] Based on the UWB data, determine a base station position of at least one UWB base station in the point cloud map, and add the at least one base station position to the point cloud map to obtain a point cloud map including the at least one base station position;

[0011] The point cloud data is input into a pre-trained target detection model so that the target detection model determines dynamic obstacles in the point cloud data based on the point cloud data, and removes the dynamic obstacles from the point cloud map including at least one base station location to obtain a target map.

[0012] Optionally, four UWB tags are provided on the autonomous driving vehicle; and based on the UWB data, determining a base station position of at least one UWB base station in the point cloud map comprises:

[0013] Based on the position of the autonomous driving vehicle in the point cloud map, calculating the positions of four UWB tags in the point cloud map;

[0014] Based on the positions of the four UWB tags in the point cloud map and the distances between the four UWB tags and the at least one UWB base station, the base station position of the at least one UWB base station in the point cloud map is determined.

[0015] Optionally, the target detection model includes: a feature layer, a fusion layer, and a detection layer; based on the point cloud data, determining a dynamic obstacle in the point cloud data includes:

[0016] Inputting the point cloud data into the feature layer so that the feature layer divides the point cloud data into a plurality of grids based on an XY plane, and extracting point cloud features of the point cloud data in the plurality of grids to generate a pseudo image;

[0017] Inputting the pseudo image into the fusion layer, so that the fusion layer generates feature maps of different sizes based on the pseudo image, and fuses the feature maps of different sizes to obtain fusion features;

[0018] The fused features are input into the detection layer, so that the detection layer determines the dynamic obstacles in the point cloud data based on the fused features.

[0019] This specification provides a method for map positioning, which is applied to an autonomous driving vehicle, and includes:

[0020] Obtaining point cloud data collected by the laser radar on the autonomous driving vehicle;

[0021] Based on the point cloud data, calculating the curvature of each point in the point cloud data, and determining points whose curvature is greater than a set curvature threshold from the point cloud data, and constructing a plane point set;

[0022] Based on the plane point set, calculating a first residual between each point in the plane point set and a map plane corresponding thereto in a target map; the target map is constructed by the above-mentioned map construction method;

[0023] Based on the first residual and the position and posture of the autonomous driving vehicle, a first Jacobian matrix is ​​constructed, and the first position of the autonomous driving vehicle in the map coordinate system of the target map is calculated with minimizing the first residual in the Jacobian matrix as the optimization goal.

[0024] Optionally, the method further comprises:

[0025] Acquire UWB data received by an ultra-wideband (UWB) tag on the autonomous driving vehicle;

[0026] Determining eigenvalues ​​corresponding to six degrees of freedom in the posture of the autonomous driving vehicle based on the Jacobian matrix;

[0027] Analyze the eigenvalues ​​corresponding to the six degrees of freedom to determine whether there is a degenerate direction; if so, take the degenerate direction as the positive direction of the X-axis, construct a degenerate coordinate system, and calculate a rotation matrix between the degenerate coordinate system and the map coordinate system based on the degenerate coordinate system and the map coordinate system;

[0028] Determine a second residual between each UWB base station based on the position of each UWB base station and the autonomous driving vehicle in the map coordinate system and the position of each UWB base station and the autonomous driving vehicle in the degenerate coordinate system;

[0029] Based on the second residuals between the UWB base stations, the rotation matrix, and the degenerate direction, construct a second Jacobian matrix corresponding to the degenerate direction, and calculate a second posture of the autonomous driving vehicle corresponding to the degenerate direction in the map coordinate system with minimization of the second residual as an optimization goal;

[0030] The first pose of the autonomous driving vehicle in the map coordinate system corresponding to the degraded direction is replaced by the second pose of the autonomous driving vehicle in the map coordinate system corresponding to the degraded direction to obtain an optimized pose of the autonomous driving vehicle in the map coordinate system.

[0031] Optionally, based on the first residual and the position and posture of the autonomous driving vehicle, a first Jacobian matrix is ​​constructed, and minimizing the first residual in the Jacobian matrix is ​​used as an optimization goal to calculate the first position of the autonomous driving vehicle in the map coordinate system, including:

[0032] Based on the point cloud data, a normal vector of each point in the point cloud data is calculated, and each point where an angle between the normal vector and the degenerate direction is less than a set angle threshold is determined from the point cloud data to construct a structured feature point set;

[0033] Based on the structured feature point set, calculating a third residual between a normal vector of each point in the structured feature point set and the degenerate direction;

[0034] Based on the first residual, the third residual and the pose of the autonomous driving vehicle, a first Jacobian matrix is ​​constructed, and the first pose of the autonomous driving vehicle in the map coordinate system is calculated with the minimization of the first residual and the third residual in the Jacobian matrix as the optimization goal.

[0035] Optionally, the method further comprises:

[0036] Acquire IMU data collected by an inertial measurement unit (IMU) on the autonomous driving vehicle;

[0037] Based on IMU data, construct the IMU pre-integration factor between any two adjacent moments;

[0038] Constructing a mileage factor at each moment based on the point cloud data and the UWB data;

[0039] Constructing a factor graph based on the position and posture of the autonomous driving vehicle at each moment, the IMU pre-integration factor between any two adjacent moments, and the mileage factor at each moment; wherein the nodes of the factor graph are the mileage factors, and the edges of the factor graph are the IMU pre-integration factors;

[0040] Based on the factor graph, the position and posture of the autonomous driving vehicle at each moment is optimized to obtain the optimized position and posture of the autonomous driving vehicle at each moment.

[0041] This specification provides a map construction device, which is applied to an autonomous driving vehicle and deploys a guide vehicle in the driving direction of the autonomous driving vehicle; the device includes:

[0042] An acquisition module, used to acquire point cloud data collected by the laser radar on the autonomous driving vehicle, IMU data collected by the inertial measurement unit IMU, and UWB data received by the ultra-wideband UWB tag;

[0043] A construction module, configured to construct a point cloud map based on the point cloud data including the features of the guided vehicle and the IMU data;

[0044] An adding module, configured to determine a base station position of at least one UWB base station in the point cloud map based on the UWB data, and add the at least one base station position in the point cloud map to obtain a point cloud map including the at least one base station position;

[0045] A removal module is used to input the point cloud data into a pre-trained target detection model so that the target detection model determines the dynamic obstacles in the point cloud data based on the point cloud data, and removes the dynamic obstacles from the point cloud map including at least one base station location to obtain a target map.

[0046] This specification provides a map positioning device, which is applied to an autonomous driving vehicle; the device includes:

[0047] An acquisition module, used to acquire point cloud data collected by the laser radar on the autonomous driving vehicle;

[0048] A construction module, used to calculate the curvature of each point in the point cloud data based on the point cloud data, and determine the points whose curvature is greater than a set curvature threshold from the point cloud data to construct a plane point set;

[0049] A calculation module, used for calculating, based on the plane point set, a first residual between each point in the plane point set and a map plane corresponding thereto in a target map; the target map is constructed by the above-mentioned map construction method;

[0050] A positioning module is used to construct a first Jacobian matrix based on the first residual and the position and posture of the autonomous driving vehicle, and calculate the first position of the autonomous driving vehicle in the map coordinate system of the target map by minimizing the first residual in the Jacobian matrix as an optimization goal.

[0051] This specification provides an electronic device, including a communication interface, a processor, a memory and a bus, wherein the communication interface, the processor and the memory are interconnected via the bus;

[0052] The memory stores machine-readable instructions, and the processor executes the above-mentioned map construction method and map positioning method by calling the machine-readable instructions.

[0053] This specification provides a machine-readable storage medium, which stores machine-readable instructions. When the machine-readable instructions are called and executed by a processor, the above-mentioned map construction method and map positioning method are implemented.

[0054] At least one of the above technical solutions adopted in this specification can achieve the following beneficial effects:

[0055] In this specification, a guide vehicle is deployed in the driving direction of the autonomous driving vehicle, and the structural features provided by the guide vehicle are used to assist in the construction of the point cloud map. Then, based on the UWB data, the base station position of at least one UWB base station in the point cloud map is determined. Next, the point cloud data is input into a pre-trained target detection model so that the target detection model determines the dynamic obstacles in the point cloud data based on the point cloud data, and removes the dynamic obstacles from the point cloud map containing the location of at least one base station to obtain a target map. Thus, the location of the UWB base station and rich semantic information are integrated in the point cloud map. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The drawings described herein are used to provide a further understanding of this specification and constitute a part of this specification. The illustrative embodiments and descriptions of this specification are used to explain this specification and do not constitute an improper limitation on this specification. In the drawings:

[0057] Figure 1 is a flowchart of a method for map construction shown in an exemplary embodiment;

[0058] Figure 2 is a schematic diagram of a point cloud map including base station locations shown in an exemplary embodiment;

[0059] Figure 3 is a flow chart of a map positioning method shown in an exemplary embodiment;

[0060] Figure 4A is a schematic diagram of a recessed area for storing a fire hydrant, shown in an exemplary embodiment;

[0061] Figure 4B is a schematic diagram showing a method for storing point cloud data of a recessed area of ​​a fire hydrant according to an exemplary embodiment;

[0062] Figure 5 is a schematic diagram of factor graph optimization shown in an exemplary embodiment;

[0063] Figure 6 is a schematic diagram of map construction and positioning shown in an exemplary embodiment;

[0064] Figure 7 is a structural diagram of an electronic device in which a map building device is located, showing an exemplary embodiment;

[0065] Figure 8 is a structural diagram of a map construction device shown in an exemplary embodiment;

[0066] Fig. 9 It is a structural diagram of a map positioning device shown in an exemplary embodiment. DETAILED DESCRIPTION

[0067] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0068] It should be noted that: in other embodiments, the steps of the corresponding method are not necessarily performed in the order shown and described in this specification. In some other embodiments, the steps included in the method may be more or less than those described in this specification. In addition, a single step described in this specification may be decomposed into multiple steps for description in other embodiments; and multiple steps described in this specification may be combined into a single step for description in other embodiments.

[0069] At present, autonomous vehicles are usually positioned through the Global Navigation Satellite System (GNSS). However, in the actual environment, there are many areas with weak or no GNSS signals, such as mining areas, tunnels, garages, etc. In these areas, the GNSS system cannot provide positioning information. In order to map and locate these areas, the inertial measurement unit IMU and LiDAR fusion method can be used.

[0070] However, in some degraded scenes such as tunnels and mines where the structural features are not rich enough, the point cloud data collected by the autonomous vehicle during driving is often highly similar, resulting in low accuracy in mapping and positioning. In addition, the positioning information of the IMU is obtained through integral calculations, which will inevitably be affected by error accumulation after long-term operation, thus affecting the accuracy of map construction and map positioning.

[0071] Based on this, this specification proposes a technical solution that deploys a guide vehicle in the driving direction of the autonomous vehicle, uses the structural features provided by the guide vehicle to assist in the construction of a point cloud map, removes dynamic obstacles in the point cloud map through a target detection model, and integrates the location of the UWB base station and rich semantic information in the point cloud map. Thus, not only can a more accurate point cloud map be generated, but the positioning accuracy of the vehicle can also be significantly improved, thereby providing more reliable path planning for autonomous vehicles.

[0072] The technical solutions provided by the embodiments of this specification are described in detail below in conjunction with the accompanying drawings.

[0073] Figure 1 FIG. 1 is a flowchart of a method for constructing a map according to an exemplary embodiment, wherein the method specifically comprises the following steps:

[0074] S100: Acquire point cloud data collected by the laser radar on the autonomous driving vehicle, IMU data collected by the inertial measurement unit IMU, and UWB data received by the ultra-wideband UWB tag.

[0075] In the embodiments of this specification, the autonomous driving vehicle is equipped with a laser radar, an inertial measurement unit IMU, and an ultra-wideband UWB tag. The autonomous driving vehicle can obtain point cloud data collected by the laser radar, IMU data collected by the IMU, and UWB data received by the UWB tag.

[0076] In practical applications, some noise points will inevitably appear in the point cloud data obtained by LiDAR. If these noise points are not filtered out, it will have a bad impact on subsequent data processing. In addition, point cloud data has the characteristics of sparseness and unevenness, and cannot be directly used for feature extraction, point cloud registration, etc. It is necessary to perform preprocessing such as point cloud filtering, sampling, segmentation, and motion distortion compensation. The purpose of motion distortion compensation is to correct the position deviation of the point cloud.

[0077] In the embodiments of this specification, the preprocessing of point cloud data includes removing the point cloud that is too far away or invalid point cloud in the original point cloud data, etc., which requires the use of point cloud segmentation technology to effectively divide the target point cloud through geometric, spatial, texture and other features, and then effectively extract the target point cloud, remove the invalid point cloud, and apply the constant speed model to compensate for motion distortion. Among them, the constant speed motion model is a set of models built into the point cloud database, which can compensate for the distortion of laser point cloud data.

[0078] In actual applications, during the driving process of an autonomous vehicle, the high-speed rotation of the engine will cause high-frequency vibration of the body. The original IMU data collected by the autonomous vehicle often contains high-frequency noise, which will affect the subsequent real-time positioning.

[0079] In the embodiments of this specification, in order to filter out high-frequency noise in the original IMU data, a first-order low-pass filtering algorithm can be used. For a continuous system, the input-output relationship of the first-order low-pass filtering module is known, and the transfer function is obtained. The specific formula is as follows:

[0080]

[0081] In the above formula, Available f c Can be used to indicate the cutoff frequency of the filter. j is used to indicate the imaginary unit. w is used to indicate the signal frequency.

[0082] For discrete systems, the first-order backward difference method can be used Discretize the continuous system, the specific formula is as follows:

[0083]

[0084] In the above formula, T s is the sampling period of the discrete system, which can be used to represent the sampling period of IMU data.

[0085] Transform the above formula, the specific formula is as follows:

[0086]

[0087] Further converted into a difference equation, the specific formula is as follows:

[0088]

[0089] Then, let the filter coefficient And α∈[0,1], the difference equation can be rewritten as follows:

[0090] V out (n) = (1-α) V out (n-1)+α·V in

[0091] In the above formula, the formula can make the frequency f c Signal components within the frequency range can pass without attenuation, while those above the frequency f c The signal component will be attenuated after passing through the filter. It should be noted that the introduction of the first-order low-pass filter will cause a significant phase lag effect on the signal component near the cutoff frequency. Therefore, when selecting the cutoff frequency, it cannot be lower than the preset cutoff frequency to avoid affecting the low-frequency signal.

[0092] In actual applications, when an autonomous vehicle is driving, if there are obstacles (such as vehicles, pedestrians, etc.) between the UWB tag and the UWB base station, the UWB data will jump significantly.

[0093] Therefore, Kalman filtering can be used to smooth the UWB raw data to ensure the robustness of the positioning results. The state equation and measurement equation of Kalman filtering are as follows:

[0094]

[0095] In the above formula, x k It can be used to indicate the actual distance between the UWB base station and the UWB tag. k It can be used to indicate the observed distance between the UWB tag and the UWB base station. k It can be used to represent the state transition matrix from time k-1 to time k. k It can be used to represent the gain matrix of the input control vector at time k. k It can be used to represent the gain matrix from the state quantity to the observation quantity at time k. Among them, the state transfer matrix A k With the observation matrix C k are all unit matrices. k It can be used to represent the system input control vector at time k. Among them, the system input u k is a zero matrix. k Can be used to represent input noise. k It can be used to represent observation noise. Assume that both input noise and observation noise follow normal distribution, the covariance of input noise is Q, and the covariance of observation noise is R.

[0096] Ignore input noise and system input u k , we can get the optimal estimate of the state and covariance at time k:

[0097]

[0098] The Kalman gain K is calculated based on the above formula, and the above formula is transformed in combination with the Kalman gain K:

[0099]

[0100] In the above formula, I is the unit matrix. R can be used to represent the covariance of the observed noise. When the value of R is small, it means that the credibility of the observed data is higher, which will make the filter respond to changes in the observed data more quickly, thereby accelerating the convergence speed. Q can be used to represent the covariance of the input noise. When the value of Q is small, it means that the credibility of the predicted value at the previous moment is higher, which can better suppress noise, but it may also lead to slower convergence and response speed.

[0101] S102: Constructing a point cloud map based on the point cloud data including the features of the guided vehicle and the IMU data.

[0102] In practical applications, in degraded scenarios such as tunnels and mines, the point cloud data collected by autonomous vehicles during driving is often highly similar, which makes it difficult for autonomous vehicles to judge their own motion status based on point cloud data. In addition, the positioning information of IMU is obtained through integral calculations, which will inevitably be affected by error accumulation after long-term operation, which may make it difficult for the vehicle to perform effective mapping and positioning.

[0103] In the embodiment of this specification, a guide vehicle is deployed in the driving direction of the autonomous driving vehicle. The autonomous driving vehicle can judge its own motion state through the structural features of the guide vehicle in the collected point cloud data.

[0104] Then, the autonomous vehicle can build a point cloud map based on the point cloud data containing the features of the guide vehicle and the IMU data. In other words, the autonomous vehicle can use the structural features provided by the guide vehicle to assist in the construction of the point cloud map.

[0105] In the process of building the map, the guide vehicle is controlled to always be located in the driving direction of the autonomous driving vehicle. For example, when the autonomous driving vehicle enters the preset sensing range of the guide vehicle, the guide vehicle is controlled to stay away from the autonomous driving vehicle and maintain a preset distance.

[0106] Furthermore, when the autonomous vehicle and the guide vehicle are traveling at the same time, the position and speed of the guide vehicle in the point cloud data collected by the autonomous vehicle will also change continuously. This dynamic change makes it difficult for the structural features of the guide vehicle in the point cloud data to provide stable reference information. Therefore, the guide vehicle cannot travel at the same time as the autonomous vehicle.

[0107] It should be noted that there are multiple methods for constructing a point cloud map in this specification. For example, the point cloud data is preliminarily aligned using IMU data to estimate the approximate position and posture of the point cloud. Then, the point cloud data is further accurately aligned using a point cloud matching algorithm to determine the position of the point cloud in the global coordinate system. The aligned point cloud data is fused into the global map to construct a point cloud map. For another example, the IMU data is optimized using a low-pass filtering algorithm. Then, the point cloud data is aligned using the optimized IMU data. Next, the point cloud data is further accurately aligned using a point cloud matching algorithm to determine the position of the point cloud in the global coordinate system. The aligned point cloud data is fused into the global map to construct a point cloud map. This specification does not specifically limit the method for constructing a point cloud map.

[0108] S104: Based on the UWB data, determine the base station position of at least one UWB base station in the point cloud map, and add the at least one base station position to the point cloud map to obtain a point cloud map including the at least one base station position.

[0109] In the embodiments of this specification, the autonomous driving vehicle can determine the base station location of at least one UWB base station in the point cloud map based on UWB data, and add at least one base station location to the point cloud map to obtain a point cloud map containing at least one base station location. Figure 2 shown.

[0110] Figure 2 It is a schematic diagram of a point cloud map including base station locations shown in an exemplary embodiment.

[0111] exist Figure 2 In the figure, taking the tunnel scene as an example, the white dots can be used to represent UWB base stations. It can be seen that multiple UWB base stations are cross-deployed at a certain distance on both sides of the tunnel for subsequent positioning of the autonomous driving vehicle.

[0112] In practical applications, Time Difference of Arrival (TDOA) is a method of positioning using the time difference of arrival, also known as hyperbolic positioning. A UWB tag sends a UWB signal to the outside, and all UWB base stations within the signal transmission range of the UWB tag will receive the UWB signal. If two base stations with known coordinates receive the signal, and the distance between the tag and the two base stations is different, then the time points at which the two base stations receive the signal are different. The principle of TDOA positioning is to use the time difference of receiving signals from multiple base stations to determine the location of the tag.

[0113] In the embodiment of this specification, when the autonomous driving vehicle is located in a degraded scene such as a tunnel or a mine, several UWB base stations are deployed at intervals and cross-deployed according to a set distance on both sides of the tunnel direction or the mine direction. In addition, four UWB tags are set on the autonomous driving vehicle, and the four UWB tags are located at the four corners of the roof position of the autonomous driving vehicle.

[0114] The autonomous vehicle can calculate the positions of the four UWB tags in the point cloud map based on its own position in the point cloud map.

[0115] The autonomous driving vehicle can then determine the base station position of at least one UWB base station in the point cloud map based on the positions of the four UWB tags in the point cloud map and the distances between the four UWB tags and the at least one UWB base station.

[0116] Specifically, according to the principle of TDOA positioning, four UWB base stations can determine the location of a UWB tag, and correspondingly, four UWB tags can also determine the location of a UWB base station. The specific formula is as follows:

[0117]

[0118] In the above formula, It can be used to represent the locations of n UWB base stations in the map coordinate system. i，j It can be used to represent the distance observation value between the i-th UWB tag and the j-th UWB base station. ε i =[xu i , i ,zu i ], i=1, 2, 3, 4 can be used to represent the positions of four UWB tags in the map coordinate system.

[0119] It should be noted that the above formula is input into the Ceres optimizer for solution to obtain the position of each UWB base station in the map coordinate system.

[0120] Furthermore, the autonomous driving vehicle can calculate the covariance corresponding to the position of each UWB base station in the map coordinate system. The specific formula is as follows:

[0121]

[0122] In the above formula, ∑ j UWB can be used to represent the covariance corresponding to the position of the UWB base station in the map coordinate system. The covariance mentioned here can be used to represent the confidence of the position of the UWB base station. The larger the covariance, the less reliable the calculated position of the UWB base station, and the smaller the covariance, the more reliable the calculated position of the UWB base station. 2 It can be used to represent the distance noise of UWB observation and can be obtained from the product manual. Furthermore, It can be used to represent the distance between the UWB tag and the UWB base station, and the residual formed by the UWB observation data. The specific formula is as follows:

[0123]

[0124] S106: Input the point cloud data into a pre-trained target detection model so that the target detection model determines dynamic obstacles in the point cloud data based on the point cloud data, and removes the dynamic obstacles from the point cloud map including at least one base station location to obtain a target map.

[0125] In this illustrative embodiment, the autonomous driving vehicle can input point cloud data into a pre-trained target detection model so that the target detection model determines dynamic obstacles in the point cloud data based on the point cloud data, and removes the dynamic obstacles from a point cloud map including at least one base station location to obtain a target map.

[0126] Among them, the target detection model includes: feature layer, fusion layer, and detection layer.

[0127] The autonomous driving vehicle can input the point cloud data into the feature layer so that the feature layer divides the point cloud data into several grids based on the XY plane, and extracts the point cloud features of the point cloud data within the several grids to generate a pseudo image.

[0128] Then, the autonomous driving vehicle can input the pseudo image into the fusion layer so that the fusion layer generates feature maps of different sizes based on the pseudo image, and fuses the feature maps of different sizes to obtain fused features.

[0129] Next, the autonomous driving vehicle can input the fused features into the detection layer so that the detection layer can determine the dynamic obstacles in the point cloud data based on the fused features.

[0130] It should be noted that the target detection model can be PointPillars. The specific implementation of PointPillar is mainly divided into three steps: (1) Convert point cloud data into pseudo images. (2) Use the backbone network to learn high-dimensional features. (3) Determine dynamic obstacles through the SSD detection head.

[0131] First, the point cloud data is divided into several grids according to the x-axis and y-axis coordinates of each point in the point cloud. The point cloud data divided in the same grid constitutes a pillar. Each point can be represented by a D = 9-dimensional vector, that is, (x, y, z, r, x c ,y c , z c , x p ,y p ), where (x, y, z) can be used to represent the three-dimensional coordinates of the point in the laser system. r can be used to represent the reflection intensity of the point. c ,y c , z c ) can be used to represent the geometric center of all points on the pillar where the point cloud is located. (x p ,y p ) can be used to indicate the relative position of the point and the geometric center.

[0132] Assuming that each frame of point cloud has P non-empty pillars and each pillar has N point cloud data, the frame of point cloud can be represented as a tensor of (D, P, N). In order to ensure that each pillar has N point cloud data, the strategy of selecting more and filling less is adopted. When the number of point clouds in each pillar exceeds N, N points are randomly sampled. When the number of point clouds in each pillar is less than N, the less part is filled with 0 vectors. Thus, the irregular point cloud data is converted into stacked Pillars with fixed shape.

[0133] Next, a simplified PointNet is used to extract features from pillars to obtain a tensor of shape (C, P, N). Then, a MaxPooling operation is performed on the dimension where pillar is located to obtain a feature map of shape (C, P). In order to realize the pseudo-image feature, P is converted to (H, W) to obtain a pseudo-image of shape (C, H, W).

[0134] Then, the backbone network is divided into a downsampling network and an upsampling network. The downsampling network can generate feature maps of different sizes through convolution layers of different sizes. The upsampling network upsamples feature maps of different sizes through deconvolution operations so that feature maps of different sizes reach the same size. Then, the upsampled feature maps are spliced ​​to obtain the spliced ​​feature map.

[0135] Finally, the spliced ​​feature map is input into the SSD detection head to detect dynamic obstacles. Since the SSD detection head is mainly used to detect dynamic obstacles in the image, it can only determine the length and width of the dynamic obstacles. Therefore, after detecting the dynamic obstacles, it is also necessary to determine the maximum height of the point cloud data in the area where the dynamic obstacles are located as the height of the dynamic obstacles.

[0136] It should be noted that after detecting static obstacles, the object detection model can add the semantic information of the static obstacles to the point cloud map. Static obstacles can refer to obstacles that do not change position within a certain period of time, such as buildings, telephone poles, road signs, walls, tunnels, guardrails, etc.

[0137] In the embodiments of this specification, after the target map is constructed, the autonomous driving vehicle can be positioned through the target map.

[0138] Figure 3 The following is a flowchart of a method for map positioning shown in an exemplary embodiment, wherein the method specifically comprises the following steps:

[0139] S300: Obtaining point cloud data collected by the laser radar on the autonomous driving vehicle.

[0140] S302: Based on the point cloud data, calculate the curvature of each point in the point cloud data, and determine the points whose curvature is greater than a set curvature threshold from the point cloud data to construct a plane point set.

[0141] S304: Based on the plane point set, calculate a first residual between each point in the plane point set and a map plane corresponding thereto in the target map.

[0142] S306: Based on the first residual and the position and posture of the autonomous driving vehicle, a first Jacobian matrix is ​​constructed, and the first position of the autonomous driving vehicle in the map coordinate system of the target map is calculated with minimizing the first residual in the Jacobian matrix as the optimization goal.

[0143] In the embodiment of this specification, the autonomous driving vehicle can obtain point cloud data collected by the laser radar.

[0144] Secondly, the autonomous driving vehicle can calculate the curvature of each point in the point cloud data based on the point cloud data, and determine the points whose curvature is greater than the set curvature threshold from the point cloud data to construct a plane point set. The specific formula is as follows:

[0145]

[0146] In the above formula, c can be used to represent the curvature. S can be used to represent the point set constructed by each point in the vicinity of the i-th point. Can be used to represent the three-dimensional coordinates of the i-th point. Can be used to represent the three-dimensional coordinates of the j-th point.

[0147] If the curvature of the point is greater than the set curvature threshold, the point is classified as an edge feature and an edge point set is constructed. If the curvature of the point is not greater than the set curvature threshold, the point is classified as a plane feature and a plane point set is constructed.

[0148] Then, based on the plane point set, the first residual between each point in the plane point set and the corresponding map plane in the target map is calculated. The specific formula is as follows:

[0149]

[0150] In the above formula, p i It can be used to represent the i-th plane feature point at time k. a , p b , p c It can be used to represent the three points selected by the i-th plane feature point on the corresponding map plane.

[0151] Next, based on the first residual and the position and posture of the autonomous driving vehicle, the first Jacobian matrix is ​​constructed, and the first position of the autonomous driving vehicle in the map coordinate system of the target map is calculated with the first residual in the Jacobian matrix minimized as the optimization goal. The specific formula is as follows:

[0152]

[0153] In the above formula, Can be used to represent the Jacobian matrix at time k. k It can be used to represent the position and posture of the autonomous driving vehicle at time k. This specification can use the Gauss-Newton method to minimize the first residual and optimize the position and posture of the autonomous driving vehicle at time k.

[0154] In the embodiments of this specification, the autonomous driving vehicle can obtain UWB data received by an ultra-wideband UWB tag.

[0155] Secondly, the autonomous driving vehicle can determine the eigenvalues ​​corresponding to the six degrees of freedom in the posture of the autonomous driving vehicle based on the Jacobian matrix, wherein the transpose of the Jacobian matrix multiplied by the Jacobian matrix can determine the eigenvalues ​​corresponding to the six degrees of freedom in the posture of the autonomous driving vehicle.

[0156] Then, the autonomous driving vehicle can analyze the eigenvalues ​​corresponding to the six degrees of freedom to determine whether there is a degenerate direction. If so, the degenerate direction is taken as the positive direction of the X-axis to construct a degenerate coordinate system, and based on the degenerate coordinate system and the map coordinate system, the rotation matrix between the degenerate coordinate system and the map coordinate system is calculated. The specific formula is as follows:

[0157]

[0158] In the above formula, the r in the first column 11 、r 21 、r 31 Can be used to indicate the X-axis direction. 12 、r 22 、r 32 Can be used to indicate the Y-axis direction. 13 、r 23 、r 33 Can be used to indicate the Z-axis direction.

[0159] Next, the autonomous driving vehicle can determine the second residual between each UWB base station based on the position of each UWB base station and the autonomous driving vehicle in the map coordinate system, and the position of each UWB base station and the autonomous driving vehicle in the degraded coordinate system. The specific formula is as follows:

[0160]

[0161] In the above formula, It can be used to represent the position of the autonomous driving vehicle in the map coordinate system. It can be used to represent the position of the jth UWB base station in the map coordinate system. It can be used to represent the position of the autonomous driving vehicle in a degenerate coordinate system. It can be used to represent the position of the jth UWB base station in the degenerate coordinate system.

[0162] Then, the autonomous driving vehicle can construct the second Jacobian matrix corresponding to the degradation direction based on the second residual, rotation matrix and degradation direction between each UWB base station, and calculate the second posture of the autonomous driving vehicle corresponding to the degradation direction in the map coordinate system with minimizing the second residual as the optimization goal. The specific formula is as follows:

[0163]

[0164] Furthermore, considering that point cloud data is sufficient to constrain the posture of the autonomous driving device in the non-degenerate direction, the introduction of UWB data may affect the posture of the autonomous driving device in the non-degenerate direction. Therefore, UWB data is only used to optimize the posture in the degenerate direction, and the elements of the non-degenerate direction in the rotation matrix are set to zero to construct the second Jacobian matrix corresponding to the degenerate direction. The specific formula is as follows:

[0165]

[0166] In the above formula, assuming that the degenerate direction is the X-axis direction, the elements in the Y-axis direction and the Z-axis direction are set to zero.

[0167] Finally, the autonomous driving vehicle can replace the first posture of the autonomous driving vehicle in the map coordinate system corresponding to the degraded direction with the second posture of the autonomous driving vehicle in the map coordinate system corresponding to the degraded direction, and obtain the optimized posture of the autonomous driving vehicle in the map coordinate system.

[0168] In actual applications, in tunnel scenarios, there are some recessed areas on the inner wall of the tunnel for storing fire hydrants. The autonomous driving device can use the structural features provided by the recessed areas to assist in its own positioning. Figure 4A , Figure 4B shown.

[0169] Figure 4A is a schematic diagram of a recessed area for storing a fire hydrant, according to an exemplary embodiment.

[0170] Figure 4B It is a schematic diagram showing a method for storing point cloud data of a recessed area of ​​a fire hydrant according to an exemplary embodiment.

[0171] exist Figure 4A The area marked by the blue oval is the recessed area used to store fire hydrants. Figure 4B In the figure, blue points are used to represent the point cloud data of the concave area.

[0172] In an embodiment of the present specification, the autonomous driving device can calculate the normal vector of each point in the point cloud data based on the point cloud data, and determine from the point cloud data each point where the angle between the normal vector and the degradation direction is less than a set angle threshold, to construct a structured feature point set.

[0173] Then, the autonomous driving device can calculate the third residual between the normal vector and the degenerate direction of each point in the structured feature point set based on the structured feature point set.

[0174] Finally, the autonomous driving device can construct a first Jacobian matrix based on the first residual, the third residual and the position and posture of the autonomous driving vehicle, and calculate the first position and posture of the autonomous driving vehicle in the map coordinate system by minimizing the first residual and the third residual in the Jacobian matrix as the optimization goal.

[0175] In the embodiments of this specification, the autonomous driving device can obtain IMU data collected by the inertial measurement unit IMU on the autonomous driving vehicle.

[0176] Secondly, the autonomous driving device can construct the IMU pre-integration factor between any two adjacent moments based on the IMU data, and construct the mileage factor at each moment based on the point cloud data and UWB data.

[0177] Next, the autonomous driving device can construct a factor graph based on the posture of the autonomous driving vehicle at each time, the IMU pre-integration factor between any two adjacent times, and the mileage factor at each time. The nodes of the factor graph are mileage factors, and the edges of the factor graph are IMU pre-integration factors.

[0178] Finally, the autonomous driving device can optimize the posture of the autonomous driving vehicle at each moment based on the factor graph to obtain the optimized posture of the autonomous driving vehicle at each moment. Figure 5 shown.

[0179] Figure 5 It is a schematic diagram of a factor graph optimization shown in an exemplary embodiment.

[0180] exist Figure 5 In the algorithm, the mileage factor is used as a priori factor to generate a preliminary estimate of the position and posture of the autonomous vehicle. The IMU pre-integration factor is used as a constraint to optimize the position and posture of the autonomous vehicle at each moment. Thus, while ensuring global consistency, local accuracy is also taken into account, thereby providing accurate map positioning.

[0181] Figure 6 It is a schematic diagram of map construction and positioning shown in an exemplary embodiment.

[0182] exist Figure 6 In the process, the autonomous vehicle can build a point cloud map based on the point cloud data containing the characteristics of the guided vehicle and the IMU data.

[0183] Secondly, based on the UWB data, the base station position of at least one UWB base station in the point cloud map is determined, and the base station position of at least one UWB base station is added to the point cloud map.

[0184] Then, the target detection model is used to perform target detection on the point cloud map, and the dynamic obstacles in the point cloud map are removed to obtain the target map.

[0185] Then, the autonomous driving vehicle is positioned through the target map, and the IMU pre-integration factor between any two adjacent moments is constructed based on the IMU data, and the mileage factor at each moment is constructed based on the point cloud data and UWB data.

[0186] Next, a factor graph is constructed based on the position and posture of the autonomous driving vehicle at each moment, the IMU pre-integration factor between any two adjacent moments, and the mileage factor at each moment.

[0187] Finally, based on the factor graph, the position and posture of the autonomous driving vehicle at each moment is optimized to obtain the optimized position and posture of the autonomous driving vehicle at each moment.

[0188] It can be seen from the above method that a guide vehicle is deployed in the driving direction of the autonomous driving vehicle, and the structural features provided by the guide vehicle are used to assist in the construction of the point cloud map. Then, based on the UWB data, the base station position of at least one UWB base station in the point cloud map is determined. Next, the point cloud data is input into the pre-trained target detection model so that the target detection model determines the dynamic obstacles in the point cloud data based on the point cloud data, and removes the dynamic obstacles from the point cloud map containing at least one base station position to obtain the target map. Thus, the location of the UWB base station and rich semantic information are integrated in the point cloud map.

[0189] In the process of positioning the autonomous vehicle, the UWB data is used to optimize the position in the degraded direction, thereby improving the accuracy of the position of the autonomous vehicle in the map coordinate system. In addition, the autonomous driving device can use the structural features provided by the concave area to assist its own positioning.

[0190] Furthermore, the autonomous driving device can also determine the confidence of the determined position of the UWB base station through the covariance corresponding to the position of each UWB base station in the map coordinate system.

[0191] Corresponding to the above-mentioned embodiment of the method based on map construction, this specification also provides an embodiment of a device for map construction.

[0192] See also Figure 7 , Figure 7 It is a structural diagram of an electronic device in which a map construction device is located, which is shown as an exemplary embodiment. At the hardware level, the device includes a processor 702, an internal bus 704, a network interface 706, a memory 708, and a non-volatile memory 710, and of course may also include other required hardware. One or more embodiments of this specification can be implemented based on software, such as the processor 702 reading the corresponding computer program from the non-volatile memory 710 into the memory 708 and then running it. Of course, in addition to software implementation, one or more embodiments of this specification do not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc., that is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0193] See also Figure 8 , Figure 8 FIG. 1 is a structural diagram of a map construction device according to an exemplary embodiment. The map construction device can be applied to Figure 7 In the electronic device shown in the figure, the technical solution of this specification is implemented. The device may include:

[0194] An acquisition module 800 is used to acquire point cloud data collected by a laser radar on the autonomous driving vehicle, IMU data collected by an inertial measurement unit IMU, and UWB data received by an ultra-wideband UWB tag;

[0195] A construction module 802, configured to construct a point cloud map based on the point cloud data including the features of the guided vehicle and the IMU data;

[0196] An adding module 804 is used to determine a base station position of at least one UWB base station in the point cloud map based on the UWB data, and add the at least one base station position in the point cloud map to obtain a point cloud map including the at least one base station position;

[0197] The removal module 806 is used to input the point cloud data into a pre-trained target detection model so that the target detection model determines the dynamic obstacles in the point cloud data based on the point cloud data, and removes the dynamic obstacles from the point cloud map including at least one base station location to obtain a target map.

[0198] Optionally, four UWB tags are set on the autonomous driving vehicle; the adding module 804 is specifically used to calculate the positions of the four UWB tags in the point cloud map based on the position of the autonomous driving vehicle in the point cloud map; based on the positions of the four UWB tags in the point cloud map and the distances between the four UWB tags and the at least one UWB base station, determine the base station position of the at least one UWB base station in the point cloud map.

[0199] Optionally, the target detection model includes: a feature layer, a fusion layer, and a detection layer; the removal module 806 is specifically used to input the point cloud data into the feature layer, so that the feature layer divides the point cloud data into a number of grids based on the XY plane, and extracts the point cloud features of the point cloud data in the number of grids to generate a pseudo image; inputs the pseudo image into the fusion layer, so that the fusion layer generates feature maps of different sizes based on the pseudo image, and fuses the feature maps of different sizes to obtain fused features; inputs the fused features into the detection layer, so that the detection layer determines the dynamic obstacles in the point cloud data based on the fused features.

[0200] See also Fig. 9 , Fig. 9 FIG. 1 is a structural diagram of a map positioning device according to an exemplary embodiment. The map positioning device can be applied to Figure 7 In the electronic device shown in the figure, the technical solution of this specification is implemented. The device may include:

[0201] An acquisition module 900 is used to acquire point cloud data collected by a laser radar on the autonomous driving vehicle;

[0202] A construction module 902 is used to calculate the curvature of each point in the point cloud data based on the point cloud data, and determine the points whose curvature is greater than a set curvature threshold from the point cloud data to construct a plane point set;

[0203] A calculation module 904 is used to calculate, based on the plane point set, a first residual between each point in the plane point set and a map plane corresponding thereto in the target map;

[0204] The positioning module 906 is used to construct a first Jacobian matrix based on the first residual and the position and posture of the autonomous driving vehicle, and calculate the first position of the autonomous driving vehicle in the map coordinate system of the target map by minimizing the first residual in the Jacobian matrix as the optimization goal.

[0205] Optionally, the positioning module 906 is further specifically used to obtain UWB data received by the ultra-wideband UWB tag on the autonomous driving vehicle; based on the Jacobian matrix, determine the eigenvalues ​​corresponding to the six degrees of freedom in the posture of the autonomous driving vehicle; analyze the eigenvalues ​​corresponding to the six degrees of freedom to determine whether there is a degenerate direction; if so, use the degenerate direction as the positive direction of the X-axis to construct a degenerate coordinate system, and based on the degenerate coordinate system and the map coordinate system, calculate the rotation matrix between the degenerate coordinate system and the map coordinate system; based on the positions of each UWB base station and the autonomous driving vehicle in the map coordinate system, and the positions of each UWB base station and the autonomous driving vehicle in the map coordinate system The method comprises the steps of: determining the position of the autonomous driving vehicle in the degenerate coordinate system, determining the second residual between each UWB base station; constructing a second Jacobian matrix corresponding to the degenerate direction based on the second residual between each UWB base station, the rotation matrix and the degenerate direction, and calculating the second posture of the autonomous driving vehicle in the map coordinate system corresponding to the degenerate direction with minimizing the second residual as the optimization goal; replacing the first posture of the autonomous driving vehicle in the map coordinate system corresponding to the degenerate direction is replaced by the second posture of the autonomous driving vehicle in the map coordinate system corresponding to the degenerate direction, so as to obtain the optimized posture of the autonomous driving vehicle in the map coordinate system.

[0206] Optionally, the positioning module 906 is further specifically used to calculate the normal vector of each point in the point cloud data based on the point cloud data, and determine from the point cloud data each point at which the angle between the normal vector and the degenerate direction is less than a set angle threshold, to construct a structured feature point set; based on the structured feature point set, calculate the third residual between the normal vector of each point in the structured feature point set and the degenerate direction; based on the first residual, the third residual and the pose of the autonomous driving vehicle, construct a first Jacobian matrix, and calculate the first pose of the autonomous driving vehicle in the map coordinate system with the optimization goal of minimizing the first residual and the third residual in the Jacobian matrix.

[0207] Optionally, the positioning module 906 is further specifically used to obtain IMU data collected by the inertial measurement unit IMU on the autonomous driving vehicle; based on the IMU data, construct an IMU pre-integration factor between any two adjacent moments; based on the point cloud data and the UWB data, construct a mileage factor at each moment; based on the posture of the autonomous driving vehicle at each moment, the IMU pre-integration factor between any two adjacent moments, and the mileage factor at each moment, construct a factor graph; wherein the nodes of the factor graph are the mileage factors, and the edges of the factor graph are the IMU pre-integration factors; based on the factor graph, the posture of the autonomous driving vehicle at each moment is optimized to obtain the optimized posture of the autonomous driving vehicle at each moment.

[0208] The implementation process of the functions and effects of each unit in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.

[0209] Based on the same concept as the above method, this specification also provides an electronic device, including: a processor; a memory for storing processor executable instructions; wherein the processor implements the steps of the method described in any of the above embodiments by running the executable instructions.

[0210] Based on the same concept as the above method, this specification also provides a computer-readable storage medium on which computer instructions are stored. When the instructions are executed by a processor, the steps of the method described in any of the above embodiments are implemented.

[0211] Based on the same concept as the above method, this specification also provides a computer program product, including a computer program / instruction, which implements the steps of the method described in any of the above embodiments when executed by a processor.

[0212] It should be understood that although the terms first, second, third, etc. may be used to describe various information in one or more embodiments of this specification, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of this specification, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0213] The above description is merely a preferred embodiment of one or more embodiments of the present specification and is not intended to limit one or more embodiments of the present specification. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of one or more embodiments of the present specification shall be included in the scope of protection of one or more embodiments of the present specification.

Claims

1. A method for map construction, characterized in that: The method is applied to an autonomous driving vehicle, and a guide vehicle is deployed in the driving direction of the autonomous driving vehicle. The method includes: Acquire point cloud data collected by the laser radar on the autonomous driving vehicle, IMU data collected by the inertial measurement unit IMU, and UWB data received by the ultra-wideband UWB tag; constructing a point cloud map based on the point cloud data including the features of the guided vehicle and the IMU data; Based on the UWB data, determine a base station position of at least one UWB base station in the point cloud map, and add the at least one base station position to the point cloud map to obtain a point cloud map including the at least one base station position; The point cloud data is input into a pre-trained target detection model so that the target detection model determines dynamic obstacles in the point cloud data based on the point cloud data, and removes the dynamic obstacles from the point cloud map including at least one base station location to obtain a target map.

2. The method according to claim 1, characterized in that Four UWB tags are set on the autonomous driving vehicle; Determining a base station position of at least one UWB base station in the point cloud map based on the UWB data includes: Based on the position of the autonomous driving vehicle in the point cloud map, calculating the positions of four UWB tags in the point cloud map; Based on the positions of the four UWB tags in the point cloud map and the distances between the four UWB tags and the at least one UWB base station, the base station position of the at least one UWB base station in the point cloud map is determined.

3. The method according to claim 1, characterized in that The target detection model includes: a feature layer, a fusion layer, and a detection layer; Determining a dynamic obstacle in the point cloud data based on the point cloud data includes: Inputting the point cloud data into the feature layer so that the feature layer divides the point cloud data into a plurality of grids based on an XY plane, and extracting point cloud features of the point cloud data in the plurality of grids to generate a pseudo image; Inputting the pseudo image into the fusion layer, so that the fusion layer generates feature maps of different sizes based on the pseudo image, and fuses the feature maps of different sizes to obtain fusion features; The fused features are input into the detection layer, so that the detection layer determines the dynamic obstacles in the point cloud data based on the fused features.

4. A map positioning method, characterized in that: The method is applied to an autonomous driving vehicle, and the method comprises: Obtaining point cloud data collected by the laser radar on the autonomous driving vehicle; Based on the point cloud data, calculating the curvature of each point in the point cloud data, and determining points whose curvature is greater than a set curvature threshold from the point cloud data, and constructing a plane point set; Based on the plane point set, calculating a first residual between each point in the plane point set and a map plane corresponding thereto in a target map; the target map is constructed by the method according to any one of claims 1 to 3 above; Based on the first residual and the position and posture of the autonomous driving vehicle, a first Jacobian matrix is ​​constructed, and the first position of the autonomous driving vehicle in the map coordinate system of the target map is calculated with minimizing the first residual in the Jacobian matrix as the optimization goal.

5. The method according to claim 4, characterized in that The method further comprises: Acquire UWB data received by an ultra-wideband (UWB) tag on the autonomous driving vehicle; Determining eigenvalues ​​corresponding to six degrees of freedom in the posture of the autonomous driving vehicle based on the Jacobian matrix; Analyze the eigenvalues ​​corresponding to the six degrees of freedom to determine whether there is a degenerate direction; if so, take the degenerate direction as the positive direction of the X-axis, construct a degenerate coordinate system, and calculate a rotation matrix between the degenerate coordinate system and the map coordinate system based on the degenerate coordinate system and the map coordinate system; Determine a second residual between each UWB base station based on the position of each UWB base station and the autonomous driving vehicle in the map coordinate system and the position of each UWB base station and the autonomous driving vehicle in the degenerate coordinate system; Based on the second residuals between the UWB base stations, the rotation matrix, and the degenerate direction, construct a second Jacobian matrix corresponding to the degenerate direction, and calculate a second posture of the autonomous driving vehicle corresponding to the degenerate direction in the map coordinate system with minimization of the second residual as an optimization goal; The first pose of the autonomous driving vehicle in the map coordinate system corresponding to the degraded direction is replaced by the second pose of the autonomous driving vehicle in the map coordinate system corresponding to the degraded direction to obtain an optimized pose of the autonomous driving vehicle in the map coordinate system.

6. The method according to claim 5, characterized in that Based on the first residual and the position and posture of the autonomous driving vehicle, a first Jacobian matrix is ​​constructed, and the first position and posture of the autonomous driving vehicle in the map coordinate system are calculated by minimizing the first residual in the Jacobian matrix as an optimization goal, including: Based on the point cloud data, a normal vector of each point in the point cloud data is calculated, and each point where an angle between the normal vector and the degenerate direction is less than a set angle threshold is determined from the point cloud data to construct a structured feature point set; Based on the structured feature point set, calculating a third residual between a normal vector of each point in the structured feature point set and the degenerate direction; Based on the first residual, the third residual and the pose of the autonomous driving vehicle, a first Jacobian matrix is ​​constructed, and the first pose of the autonomous driving vehicle in the map coordinate system is calculated with the minimization of the first residual and the third residual in the Jacobian matrix as the optimization goal.

7. The method according to claim 5, characterized in that The method further comprises: Acquire IMU data collected by an inertial measurement unit (IMU) on the autonomous driving vehicle; Based on IMU data, construct the IMU pre-integration factor between any two adjacent moments; Constructing a mileage factor at each moment based on the point cloud data and the UWB data; Constructing a factor graph based on the position and posture of the autonomous driving vehicle at each moment, the IMU pre-integration factor between any two adjacent moments, and the mileage factor at each moment; wherein the nodes of the factor graph are the mileage factors, and the edges of the factor graph are the IMU pre-integration factors; Based on the factor graph, the position and posture of the autonomous driving vehicle at each moment is optimized to obtain the optimized position and posture of the autonomous driving vehicle at each moment.

8. A map construction device, characterized in that: The device is applied to an autonomous driving vehicle, and a guide vehicle is deployed in the driving direction of the autonomous driving vehicle; the device comprises: An acquisition module, used to acquire point cloud data collected by the laser radar on the autonomous driving vehicle, IMU data collected by the inertial measurement unit IMU, and UWB data received by the ultra-wideband UWB tag; A construction module, configured to construct a point cloud map based on the point cloud data including the features of the guided vehicle and the IMU data; An adding module, configured to determine a base station position of at least one UWB base station in the point cloud map based on the UWB data, and add the at least one base station position in the point cloud map to obtain a point cloud map including the at least one base station position; A removal module is used to input the point cloud data into a pre-trained target detection model so that the target detection model determines the dynamic obstacles in the point cloud data based on the point cloud data, and removes the dynamic obstacles from the point cloud map including at least one base station location to obtain a target map.

9. A map positioning device, characterized in that: The device is applied to an autonomous driving vehicle, and comprises: An acquisition module, used to acquire point cloud data collected by the laser radar on the autonomous driving vehicle; A construction module, used to calculate the curvature of each point in the point cloud data based on the point cloud data, and determine the points whose curvature is greater than a set curvature threshold from the point cloud data to construct a plane point set; A calculation module, configured to calculate, based on the plane point set, a first residual between each point in the plane point set and a map plane corresponding thereto in a target map; the target map is constructed by the method according to any one of claims 1 to 4 above; A positioning module is used to construct a first Jacobian matrix based on the first residual and the position and posture of the autonomous driving vehicle, and calculate the first position of the autonomous driving vehicle in the map coordinate system of the target map by minimizing the first residual in the Jacobian matrix as an optimization goal.

10. An electronic device, characterized in that: It includes a communication interface, a processor, a memory and a bus, wherein the communication interface, the processor and the memory are interconnected via the bus; The memory stores machine-readable instructions, and the processor executes the method according to any one of claims 1 to 7 by calling the machine-readable instructions.

11. A machine-readable storage medium, characterized in that: The machine-readable storage medium stores machine-readable instructions, and when the machine-readable instructions are called and executed by the processor, the method according to any one of claims 1 to 7 is implemented.

Images