Lane-level road binding method and system and medium

Abstract

The invention relates to the technical field of intelligent traffic, in particular to a lane-level road binding method and system and a medium, and the method comprises the steps: obtaining a current position point and a current candidate lane, and constructing a transmission probability based on the current position point and the current candidate lane; obtaining and constructing a transition probability based on the distance between the current position point and the previous position point and the path similarity between the current candidate lane position and the previous candidate lane position; a vehicle matching path is generated based on the transmission probability and the transition probability. According to the invention, the problems of poor effect, low efficiency and the like of network lane-level road binding for GNSS errors and complex roads in the prior art can be solved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent transportation, and particularly to a lane-level road binding method, system and medium. Background Art

[0002] The lane-level road binding method is a technology that precisely binds the driving trajectory of a vehicle to a specific lane through high-precision positioning and map data. The implementation process of the lane-level road binding method is a complex and crucial process, especially in the fields of vehicle navigation, traffic management, and autonomous driving. Road binding is specifically the process of associating Global Navigation Satellite System (GNSS) trajectory points with the road network, aiming to determine the actual road and driving path where the vehicle is located. Due to the limitations of GNSS errors and trajectory sampling rates, ensuring accurate matching of trajectory points is crucial, and this process includes key steps such as data collection, map data preparation, trajectory preprocessing, and road binding algorithms. Finally, road binding generates the sequence of road segments where the vehicle travels, providing necessary information for applications such as navigation, traffic management, and intelligent driving. In the face of GNSS errors and complex road networks, developing more precise matching algorithms is crucial to meet the requirements of different application scenarios, such as scenarios like autonomous driving and intelligent transportation systems.

[0003] Solving GNSS errors and the need for high-precision matching is an important driving force for the development of more precise matching algorithms. For scenarios such as autonomous driving and intelligent transportation systems, the matching algorithm not only requires high-precision matching but also real-time performance and robustness. Summary of the Invention

[0004] In view of the above technical problems, the present invention provides a lane-level road binding method, system and medium, which can provide a lane-level road binding based on a high-precision map with higher accuracy, higher efficiency and better applicability, and can operate in a real-time environment and can handle various complex traffic conditions and road scenarios to ensure that the vehicle can travel safely and efficiently.

[0005] To achieve the above object, the present invention adopts the following technical solutions.

[0006] The first aspect of the present invention provides a lane-level road binding method, including: Obtaining and constructing an emission probability based on the current position point and the current candidate lane; Obtaining and constructing a transition probability based on the distance between the current position point and the previous position point, and the path similarity between the current candidate lane position and the previous candidate lane position; Generating a vehicle matching path based on the emission probability and the transition probability.

[0007] As an optional implementation manner, the obtaining and constructing an emission probability based on the current position point and the current candidate lane includes: Obtain and construct the emission probability based on the original trajectory sequence and the original data points therein, as well as the road data and the road information therein.

[0008] As an alternative implementation, the obtaining and constructing the transition probability based on the distance between the current position point and the previous position point, and the path similarity between the current candidate lane position and the previous candidate lane position includes: Obtain and construct the transition probability based on the distance from the GPS original data point at the previous moment to the GPS original data point at the next moment, the shortest path on the earth's surface between the matching points on the candidate lane at the previous moment and the matching points on the candidate lane at the next moment, the weight of the similarity between the distance between the GPS original data points at adjacent moments and the shortest path on the earth's surface between the candidate lane matching points at adjacent moments, the distance from the GPS original data point at the previous moment to the corresponding candidate lane, and the distance from the GPS original data point at the next moment to the corresponding candidate lane.

[0009] As an alternative implementation, the obtaining the vehicle matching path based on the emission probability and the transition probability includes: Obtain the probability that each candidate lane was selected at the previous moment; Obtain the observation probability corresponding to the next moment based on the emission probability of the next moment; Obtain the probability that each candidate lane was selected at the next moment based on the observation probability corresponding to the next moment, the probability that each candidate lane was selected at the previous moment, and the transition probability corresponding to the previous moment; Obtain the matching road with the maximum probability at the target moment; Perform trajectory backtracking in the order from the target moment to the initial moment to obtain the vehicle matching path.

[0010] As an alternative implementation, the obtaining the probability that each candidate lane was selected at the previous moment includes: obtaining the observation probability corresponding to the previous moment based on the distance error and direction error at the previous moment, where the direction error includes the angle difference between the vehicle heading angle and the road direction at the previous moment; Obtain the probability that each candidate lane was selected at the previous moment based on the observation probability corresponding to the previous moment, the transition probability corresponding to the moment before the previous moment, and the probability that each candidate lane was selected in the moment before the previous moment.

[0011] As an alternative implementation, the obtaining the observation probability corresponding to the next moment based on the emission probability of the next moment includes: Construct the emission probability based on the distance error and direction error at the next moment, where the distance error is obtained based on the actual distance between the GPS observation point and the candidate lane, and the direction error is obtained based on the difference between the ideal distances and the similarity between the current vehicle direction and the candidate lane direction; Obtain the observation probability corresponding to the next moment based on the emission probability at the next moment.

[0012] As an optional implementation, the probability of each candidate lane being selected at the initial moment is the observation probability corresponding to the initial moment.

[0013] The second aspect of the present invention provides a lane-level road binding system, including: A first acquisition module, at least used to acquire and construct an emission probability based on the current position point and the current candidate lane; A second acquisition module, at least used to acquire and construct a transition probability based on the distance between the current position point and the previous position point, and the path similarity between the current candidate lane position and the previous candidate lane position; A generation module, at least used to generate a vehicle matching path based on the emission probability and the transition probability.

[0014] The third aspect of the present invention provides an electronic device, including: At least one processor; and at least one memory communicatively connected to the processor, where: the memory stores program instructions executable by the processor, and the processor can execute the method described in the first aspect of the embodiments of the present invention by invoking the program instructions.

[0015] The fourth aspect of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is run by a computer, it executes the method described in the first aspect of the embodiments of the present invention.

[0016] The present invention abstracts the lane matched by the vehicle into a hidden state, and abstracts the RTK position point (the real-time position of the vehicle) into an observation state, that is, the whole problem infers the lane sequence with the greatest possibility of vehicle matching through the sequence of RTK position points and high-precision map elements, which more realistically reflects the driving mode of the vehicle, thereby improving the road binding accuracy. Description of the Drawings

[0017] Figure 1 It is a schematic flowchart of a lane-level road binding method in an embodiment of the present invention.

[0018] Figure 2 It is a schematic flowchart of another lane-level road binding method in an embodiment of the present invention.

[0019] Figure 3It is a GPS raw data information diagram on a north-south elevated road overpass in an embodiment of the present invention.

[0020] Figure 4 It is a GPS abstract data information diagram on a north-south elevated road overpass in an embodiment of the present invention.

[0021] Figure 5 It is a time-lapse direction diagram of a lane-level road binding method in an embodiment of the present invention.

[0022] Figure 6 It is an abstract road backtracking diagram of a lane-level road binding method in an embodiment of the present invention.

[0023] Figure 7 It is a schematic diagram for excluding abnormal situations of a lane-level road binding method in an embodiment of the present invention.

[0024] Figure 8 It is a schematic diagram for loading a high-precision map of a lane-level road binding method in an embodiment of the present invention.

[0025] Figure 9 It is a schematic diagram of candidate lanes of a lane-level road binding method in an embodiment of the present invention.

[0026] Figure 10 It is a schematic diagram of candidate lanes at time t of a lane-level road binding method in an embodiment of the present invention.

[0027] Figure 11 It is a schematic diagram of candidate lanes at time t+1 of a lane-level road binding method in an embodiment of the present invention.

[0028] Figure 12 It is a schematic diagram of candidate lanes of a lane-level road binding method based on navigation data constraints in an embodiment of the present invention.

[0029] Figure 13 It is a schematic diagram of a solution framework of a lane-level road binding method based on an HMM model in an embodiment of the present invention.

[0030] Figure 14 It is a schematic diagram of the principle considering angle error in a lane-level road binding method in an embodiment of the present invention.

[0031] Figure 15 It is a schematic diagram of the pseudocode of a lane-level road binding method in an embodiment of the present invention.

[0032] Figure 16 It is a schematic diagram of the pseudocode of another lane-level road binding method in an embodiment of the present invention.

[0033] Figure 17 It is a schematic diagram of the visualization effect without using the lane-level road binding method shown in the present invention in an embodiment of the present invention.

[0034] Figure 18 This is a schematic diagram of the visualization effect of the lane-level path binding method shown in the present invention in the embodiments of the present invention.

[0035] Figure 19 This is a block diagram of a lane-level path binding system in the embodiments of the present invention.

[0036] Figure 20 This is a schematic structural diagram of an electronic device in the embodiments according to the present invention. Detailed implementation manners

[0037] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts shall fall within the protection scope of the present invention.

[0038] It should be understood that the terms "first", "second", "third", etc. in the claims, the description and the drawings of the present disclosure are used to distinguish different objects, rather than to describe a specific order. The terms "including" and "comprising" used in the description and claims of the present disclosure indicate the presence of the described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or their combinations. It should also be understood that the terms used in the description of the present disclosure herein are only for the purpose of describing specific embodiments, and are not intended to limit the present disclosure.

[0039] With the progress of technology and the increasing demand for precise matching, the development of path binding algorithms will continue to be an important issue in these application fields, providing more reliable support for vehicle navigation, traffic management, autonomous driving, etc.

[0040] As Figure 1 shown, the first aspect of the present invention provides a lane-level path binding method.

[0041] In this method, the lane-level road binding is evaluated as a process that advances over time to match the GPS raw data points to the corresponding roads. By comparing information such as the vehicle's position, direction, and attitude with the map data, information about the actual road where the vehicle is located and its position on the road can be obtained. Among them, for the road binding system, the comparison results of the vehicle's position, direction, and attitude information with the map data are observable information, that is, the observation state, while the vehicle's actual driving path and position are unobservable information for the road binding system, that is, the hidden state. Moreover, the comparison results of the observable vehicle information and the map data are generated by the unobservable actual driving path of the vehicle. And when the actual driving path is certain, as time advances, the sequence of observable states generated is random. Therefore, the road binding process has three essential elements: the initial state, the hidden state sequence, and the observable state sequence, and it has a certain degree of randomness over time. So the road binding process can be abstracted as a state transition process based on the Hidden Markov Model (HMM). Specifically, it mainly includes the following steps.

[0042] Step S100: Obtain and construct the emission probability based on the current position point and the current candidate lane.

[0043] Specifically, the emission probability, also known as the emission matrix or emission probability matrix, is used to represent the probability when the actual road where the vehicle is located is and the GPS raw data point p i is observed. Therefore, in the present invention, the emission probability is constructed by obtaining and based on the original trajectory sequence and the raw data point p i within it, as well as the road data and the road information e n within it.

[0044] Step S200: Obtain and construct the transition probability based on the distance between the current position point and the previous position point, and the path similarity from the current candidate lane position to the previous candidate lane position.

[0045] Specifically, the transition probability, also known as the transition matrix or transition probability matrix, is used to represent the probability when the vehicle is initially on the candidate road and the observed GPS raw data p i-1 point is, and then it transfers to the candidate road and the observed GPS raw data point at this time is p i at this time.

[0046] Step S300: Generate the vehicle matching path based on the emission probability and the transition probability.

[0047] Specifically, the emission probability and the transition probability are processed through the Viterbi algorithm. For a vehicle trajectory information P = p1 → … → p i → … → p T , in the HMM, the Viterbi algorithm can be used to calculate the hidden state sequence with the highest corresponding probability, which is the vehicle matching path and also represents the most likely driving trajectory of the vehicle.

[0048] Herein, in this way, the present invention abstracts the lane-level road binding problem into a prediction problem of an HMM model for processing. After inputting information such as a high-precision map and vehicle pose, it can obtain the most likely lane matching result, meeting the requirements of high precision and high efficiency for lane-level road binding, and is more applicable to the lane-level road binding method based on a high-precision map.

[0049] As Figure 2 shown, in an embodiment of the present invention, first, the abnormal input values of the system are filtered out through anomaly detection, and information such as RTK and IMU is fused to smooth the vehicle position and angle. Secondly, a high-precision map is loaded according to certain rules and cached in the map buffer area. Then, candidate lanes that meet the requirements near the current frame position are obtained. Next, for the vehicle position point, vehicle state, and corresponding candidate lanes of the current frame, an initial probability matrix, an emission probability matrix, and a transition probability matrix are calculated. Finally, the Viterbi algorithm is used to search for the candidate lane sequence corresponding to the most likely path, thereby obtaining the final result.

[0050] As Figure 3 shown, taking the map of the North-South Elevated Road and Middle Ring Road Overpass in Shanghai as the application scenario map, it can be seen from the figure that at this time the vehicle is driving on the road and generates 3 different position GPS raw data information at different times, and data abstraction is performed on it, key roads are extracted and drawn to obtain Figure 4 .

[0051] As Figure 4 shown, the vehicle is driving on the road, and p1, p2, and p3 are the GPS position information obtained by the vehicle at different times. Due to the positioning error of GPS, the positioning points do not match the road on which the vehicle is actually driving. Therefore, perpendicular projection lines are drawn for the sections e1 to e7 relatively close to p1, p2, and p3, and these sections are used as candidate roads, and the intersection points of the perpendicular lines and the roads are used as possible matching points. As time goes by, the number of GPS position information collected will gradually increase. As a result, the probability of the true driving route will continue to increase, while the probability of the route where the candidate road of the interference term is located will decrease accordingly. Finally, the road and position point on which the vehicle is actually driving are determined and used as the result.

[0052] As Figure 5 and Figure 6 shown, further, the problem of tying roads based on HMM is described as a time-varying process as follows: when t = 1, there are two candidate roads e1 and e2 for p1; when t = 2, there are two candidate roads e3 and e4 for p2; when t = 3, there are three candidate roads e5, e6 and e7 for p3.

[0053] In an embodiment of the present invention, in order to better reduce the interference of abnormal data, as Figure 2 and Figure 7 shown, when the RTK state is unavailable, the number of satellites, covariance, timestamp rollback or equality, the angle difference between two frames is too large, there is a large jump in the middle position between two frames, abnormal zero (timestamp is 0, position point is 0), the necessary input is empty, etc., the corresponding data information will also be correspondingly excluded.

[0054] Here, through this method, illegal time, position points, angles and other information can be excluded to ensure the accuracy of the input, thereby improving the overall matching continuity and accuracy of tying roads.

[0055] In an embodiment of the present invention, in order to avoid delays caused by multiple loadings of map data, as Figure 8 shown, the present invention selects to load all lane map data within a preset geographical range around the vehicle, such as within a 400m * 400m rectangular range. After traveling a certain straight distance, the map is cached in the map cache area.

[0056] As Figure 9 shown, in order to determine whether initialization is required currently, the surrounding map is obtained through a range. If initialization is required, lanes within a certain surrounding range are selected and screened to obtain the candidate lanes for the current frame.

[0057] As Figure 10 and Figure 11 shown, if initialization is not required, the candidate lanes for the current frame are obtained through the lane matching result of the previous frame: including the possible candidate lanes in the previous frame, the longitudinal successor lanes corresponding to the possible candidate lanes in the previous frame, the lateral lanes corresponding to the possible candidate lanes in the previous frame and taking effect when the lane boundary line type between the two lanes is a dotted line, and the longitudinal successor lanes of the lateral lanes corresponding to the possible candidate lanes in the previous frame.

[0058] Further, as Figure 12 shown, through the constraint of navigation data, the candidate lane range is constrained within the lane group set corresponding to the navigation data.

[0059] Here, in this way, the present invention performs route binding based on the map data structure and data format of the high-precision map. During the route binding process, the differences between the high-precision map and the navigation map are considered, and the transfer relationship between different candidate lanes is constructed through the characteristics of the high-precision map, thereby further improving the route binding accuracy.

[0060] In one embodiment of the present invention, the construction of the transfer probability based on the distance between the current position point and the previous position point, and the path similarity from the current candidate lane position to the previous candidate lane position includes: Obtain and based on the GPS raw data point p at the previous moment (moment i - 1) i-1 to the GPS raw data point p at the next moment (moment i) i distance d i-1,i , the matching point on the candidate lane at the previous moment (moment i - 1) to the matching point on the candidate lane at the next moment (moment i) the shortest path R on the earth's surface between them i-1,i , the weight λ of the similarity between the distance between GPS raw data points at adjacent moments and the shortest path on the earth's surface between matching points of candidate lanes at adjacent moments, the distance d from the GPS raw data point p at the previous moment (moment i - 1) i-1 to the corresponding candidate lane distance d i-1 , and the distance d from the GPS raw data point p at the next moment (moment i) i to the corresponding candidate lane distance d i , to construct the transfer probability. Here, moments i - 1 and i are used to indicate the relationship between adjacent moments and are not particularly limited.

[0061] Specifically, the transfer probability is defined as:

[0062] where d i-1,i represents the Euclidean distance from the GPS raw data point p at moment i - 1 to the GPS raw data point p at moment i i-1 , and the meaning of R i is the matching point on the candidate lane at moment i - 1 to the candidate road at moment i i-1,i ​​The shortest path on the earth's surface between matching points. The shortest distance between two points on the roads on the earth's surface can be obtained by using a shortest path algorithm, such as using algorithms like Dijkstra's algorithm to calculate. In the present invention, the A* algorithm is used for calculation. λ is the weight representing the similarity between the Euclidean distance between GPS raw data points at adjacent times and the shortest path distance on the earth's surface between candidate road matching points at adjacent times. Because d i-1 represents the distance from the GPS raw data point p at the (i - 1)th moment i-1 to the candidate road , so 1 - λ represents the weight of the change trend of the distance error at adjacent times. λ in the formula represents the influence of the transfer probability T of the distance similarity and the change trend of the distance error respectively.

[0063] Here, in calculating the transition probability matrix in the present invention, the A* algorithm based on a high-precision map is used to calculate the distance similarity between the historical trajectory and the distance from the previous candidate point to the current candidate point, which more realistically reflects the driving mode of the vehicle, thereby improving the road binding accuracy.

[0064] Specifically, as Figure 13 shown, the HMM model provided by the present invention can solve three specific problems in the lane-level road binding problem, which involves using the forward algorithm to evaluate the probability of the observation sequence, suitable for judging the matching degree between the observation sequence and the model; the Viterbi algorithm for the decoding problem, suitable for inferring the most likely hidden state sequence, and the forward-backward algorithm for calculating the posterior probability of the state, suitable for model parameter learning (such as the Baum-Welch algorithm) or state estimation. The HMM model includes: the model evaluation problem, the learning problem, and the decoding problem, where: The model evaluation problem is to calculate the probability P=(O|λ) of the observation sequence O=(o 1, o2,…,o T ) appearing under the hidden Markov model λ when the parameters of the determined hidden Markov model λ=(A,B,π) and the observation sequence O are also determined, and select the model λ with the highest probability from them.

[0065] The learning problem is to train and learn the parameters of the hidden Markov model λ=(A,B,π) under a determined observation sequence O=(o 1, o2,…,o T ) so that the trained model parameters make the probability of observing the sequence O the largest, that is, P=(O|λ) is the largest.

[0066] The decoding problem (also called the prediction problem) is to find the most likely hidden state sequence in the case of the known observation sequence O=(o 1, o2,…,o T), use the determined hidden Markov model λ=(A,B,π) to calculate the most likely hidden sequence I=(i1,i2,…,i T ) and the probability P(I|O) in this case.

[0067] In the present invention, the emission probability is calculated for the current location point and all the current candidate roads. The emission probability is used to indicate when the actual road is When and observe p i The probability of reaching the original GPS data point. The observation probability mainly considers the influence of distance error and the influence of direction error on it. Therefore, the distance error is defined as:

[0068] Where E d It is expressed as the distance error, and d represents the candidate road The projection point To GPS original data point p i σ represents the standard deviation of GPS measurement, which can be determined through a large number of experiments.

[0069] In the present invention, in order to further improve the accuracy of HMM binding, so that vehicles can still match more accurately at complex intersections, the angle difference between the vehicle heading angle and the road direction is added to the algorithm. This allows the binding method of the present invention to still ensure the accuracy of matching when encountering complex intersections with multiple similar candidate roads. The method of considering the angle difference is as follows: Figure 14 As shown, the angle between the candidate road direction and the vehicle driving direction is θ, and the direction error is defined as: E θ =cosθ.

[0070] Direction error E θ Used to indicate vehicle direction and candidate roads The similarity of the directions. When the directions are the same, the angle θ is 0°, and the direction error is 1. When the angle between the directions exceeds 90°, the direction error E θ is a negative number, when the vehicle direction is in the same direction as the candidate road The direction similarity is very low, then the current candidate road It is most likely not the current GPS original data point p i The road to match.

[0071] Finally, the improved observation probability can be obtained by combining the above three formulas: E=E d ·E θ .

[0072] In one embodiment of the present invention, obtaining the vehicle matching path based on the emission probability and the transition probability includes: Obtaining the probability that each candidate lane is selected at the previous moment (moment i - 1); wherein, at the initial moment (i = 1 moment, i.e., in the initial state), the probability that each candidate lane is selected is the observation probability corresponding to the initial moment, which is Based on the emission probability at the next moment (moment i), obtaining the observation probability corresponding to the next moment (moment i), that is, at subsequent moments, i.e., at moments when i > 1, the e i The probability that each candidate road is selected is: Wherein, p i Represents the GPS raw data point at moment i, e i Represents the set of candidate roads at this time, V i,k Represents the probability corresponding to the most likely road sequence as the first i hidden states Wherein Is the transition probability and Is the observation probability; based on the observation probability corresponding to the next moment (moment i), the probability that each candidate lane is selected at the previous moment (moment i - 1), and the transition probability corresponding to the previous moment (moment i - 1), obtaining the probability that each candidate lane is selected at the next moment (moment i); Obtaining the matching road with the maximum probability at the target moment (moment T); Performing trajectory backtracking in the order from the target moment (moment T) to the initial moment (i = 1 moment) to obtain the vehicle matching path.

[0073] Wherein, obtaining the matching road section with the maximum probability and recording the road matched at the previous moment At i = T moment, the matching road with the maximum probability is: Finally, performing trajectory backtracking on the obtained optimal matching path: x i = ψ i (e i ), i = T - 1, T - 2,..., 1; Thus, calculating the optimal vehicle driving path X = x1 → x2 →... → x T .

[0074] Specifically, to meet the real-time matching requirements of the algorithm, the present invention further uses the online Viterbi algorithm and combines the strategy of two outputs. When calculating candidate road segments for each GPS position point, the candidate road with the maximum current joint probability is used as the first output, and the converged global sub-optimal trajectory is used as the second output.

[0075] In an embodiment of the present invention, the obtaining of the probabilities that each candidate lane was selected at the previous moment includes: Obtaining the observation probability corresponding to the previous moment based on the distance error and direction error at the previous moment, where the direction error includes the angle difference between the vehicle heading angle and the road direction at the previous moment; at subsequent moments, based on the observation probability corresponding to the previous moment (the (i - 1)-th moment), the transition probability corresponding to the moment before the previous moment (the (i - 2)-th moment) of the previous moment (the (i - 1)-th moment), and the probabilities that each candidate lane was selected at the moment before the previous moment (the (i - 2)-th moment) of the previous moment (the (i - 1)-th moment), obtaining the probabilities that each candidate lane was selected at the previous moment (the (i - 1)-th moment). The obtaining method is similar to obtaining the probabilities that each candidate lane was selected at the subsequent moment (the i-th moment) based on the observation probability corresponding to the subsequent moment (the i-th moment), the probabilities that each candidate lane was selected at the previous moment (the (i - 1)-th moment), and the transition probability corresponding to the previous moment (the (i - 1)-th moment), and will not be elaborated here.

[0076] In an embodiment of the present invention, the obtaining of the observation probability corresponding to the subsequent moment based on the emission probability at the subsequent moment includes: Constructing the emission probability based on the distance error and direction error at the subsequent moment (the i-th moment), where the distance error is obtained based on the actual distance between the GPS observation point and the candidate lane, and the direction error is obtained based on the difference between the ideal distances and the similarity between the current vehicle direction and the candidate lane direction; Obtaining the observation probability corresponding to the subsequent moment (the i-th moment) based on the emission probability at the subsequent moment.

[0077] In the present invention, the construction methods of the initial probability, emission probability matrix, and transition probability matrix in the HMM model can use the exponential function to replace the Gaussian function; the parameter weights in the HMM emission probability matrix and transition probability matrix can be adjusted and optimized through machine learning; in the process of constructing the transition matrix, the path planning algorithm A* can be improved to bidirectional A* to further improve the real-time performance.

[0078] In the present invention, the concepts of moments expressed as i, i - 1, i - 2, etc. are only for illustration and a relative time relationship. The purpose is to enable those skilled in the art to further understand the technical content of the present invention, rather than simply defining or determining its connotation solely from the layout. Those skilled in the art can also infer the technical solution to be protected by the present invention without including the above concepts. To better illustrate the technical solution to be protected by the present invention, the present invention specifically provides the following pseudocode Figure 15 and Figure 16 and the corresponding visualization comparison of the binding path results Figure 17 and Figure 18 .

[0079] Herein, through this method, the present invention can 1) Meet the accuracy requirements of high-precision map binding paths; 2) Better consider the differences between high-precision maps and previous navigation maps, and give full play to the high-precision and multi-element advantages of high-precision maps; 3) While improving the accuracy, ensure the real-time performance of the binding path algorithm in the vehicle online scenario; 4) Have historical results, have chronological memory, and can maintain trajectory tracking; 5) When a matching error occurs, be able to quickly correct the error and match to the correct lane; 6) And through the map caching technology, avoid unnecessary performance overhead caused by repeated access to the map multiple times.

[0080] As Figure 19 shown, the second aspect of the present invention provides a lane-level binding path system, including: A first acquisition module, at least used to acquire and construct an emission probability based on the current position point and the current candidate lane; A second acquisition module, at least used to acquire and construct a transition probability based on the distance between the current position point and the previous position point, and the path similarity between the current candidate lane position and the previous candidate lane position; A generation module, at least used to generate a vehicle matching path based on the emission probability and the transition probability.

[0081] As Figure 20 shown, the present invention also provides an electronic device, including: At least one processor; and at least one memory communicatively connected to the processor, wherein: the memory stores program instructions executable by the processor, and the processor can execute the above-mentioned lane-level binding path method by calling the program instructions.

[0082] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above-described lane-level route binding method is implemented.

[0083] It can be understood that the computer-readable storage medium may include: any entity or device capable of carrying a computer program, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disc, a computer memory, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), and a software distribution medium, etc. The computer program includes computer program code. The computer program code may be in the form of source code, object code, an executable file, or some intermediate form, etc. The computer-readable storage medium may include: any entity or device capable of carrying computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disc, a computer memory, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), and a software distribution medium, etc.

[0084] In some embodiments of the present invention, the electronic device may include a controller or a processor. The controller is a single-chip microcomputer chip that integrates a processor, a memory, a communication module, etc. The processor may refer to the processor included in the controller. The processor may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0085] Any process or method description shown in the flowchart or described in other ways herein can be understood as representing a module, segment, or part of code including one or more executable instructions for implementing a specific logical function or process. And the scope of the preferred embodiments of the present invention includes additional implementations, where the functions may be executed in a substantially simultaneous manner or in an order opposite to that shown or discussed, according to the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention belong.

[0086] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0087] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A lane-level binding method, characterized in that: include: Obtain and construct the launch probability based on the current position point and the current candidate lane; Obtain and construct the transition probability based on the distance between the current position point and the previous position point and the path similarity from the current candidate lane position to the previous candidate lane position; Generate vehicle matching paths based on emission probabilities and transition probabilities.

2. The lane-level binding method according to claim 1, characterized in that: The obtaining and constructing the emission probability based on the current position point and the current candidate lane includes: The emission probability is obtained and constructed based on the original trajectory sequence and the original data points therein, and the road data and the road information therein.

3. The lane-level binding method according to claim 1, characterized in that: The obtaining and constructing the transition probability based on the distance between the current position point and the previous position point and the path similarity from the current candidate lane position to the previous candidate lane position includes: The transition probability is constructed based on the distance from the GPS raw data point at the previous moment to the GPS raw data point at the next moment, the shortest path on the earth's surface between the matching point on the candidate lane at the previous moment and the matching point on the candidate lane at the next moment, the weight of the similarity between the distance between the GPS raw data points at adjacent moments and the shortest path on the earth's surface between the matching points of the candidate lanes at adjacent moments, the distance from the GPS raw data point at the previous moment to the corresponding candidate lane, and the distance from the GPS raw data point at the next moment to the corresponding candidate lane.

4. The lane-level binding method according to claim 1, characterized in that: The obtaining of the vehicle matching path based on the emission probability and the transition probability includes: Get the probability of each candidate lane being selected at the previous moment; Based on the emission probability at the next moment, the observation probability corresponding to the next moment is obtained; Obtaining the probability of each candidate lane being selected at the next moment based on the observation probability corresponding to the next moment, the probability of each candidate lane being selected at the previous moment, and the transition probability corresponding to the previous moment; Obtain the matching road with the highest probability at the target time; The trajectory is backtracked in the order from the target time to the initial time to obtain the vehicle matching path.

5. The lane-level binding method according to claim 4, characterized in that: The obtaining of the probability of each candidate lane being selected at the previous moment includes: Obtaining the observation probability corresponding to the previous moment based on the distance error and direction error at the previous moment, wherein the direction error includes the angle difference between the vehicle heading angle and the road direction at the previous moment; Based on the observation probability corresponding to the previous moment, the transition probability corresponding to the moment before the previous moment, and the probability of each candidate lane being selected in the moment before the previous moment, the probability of each candidate lane being selected in the previous moment is obtained.

6. The lane-level binding method according to claim 4, characterized in that: The obtaining the observation probability corresponding to the next moment based on the emission probability at the next moment includes: The emission probability at the next moment is constructed based on the distance error and direction error at the next moment, wherein the distance error is obtained based on the actual distance between the GPS observation point and the candidate lane, and the direction error is obtained based on the difference between the ideal distance and the similarity between the current vehicle direction and the candidate lane direction; The observation probability corresponding to the next moment is obtained based on the emission probability at the next moment.

7. The lane-level binding method according to claim 4, characterized in that: The probability of each candidate lane being selected at the initial moment is the observation probability corresponding to the initial moment.

8. A lane-level road binding system, characterized in that: include: A first acquisition module, at least for acquiring and constructing a launch probability based on a current position point and a current candidate lane; A second acquisition module is at least used to acquire and construct a transition probability based on a distance between a current position point and a previous position point and a path similarity from a current candidate lane position to a previous candidate lane position; The generation module is at least used to generate a vehicle matching path based on the emission probability and the transition probability.

9. An electronic device, characterized in that: include: at least one processor; and at least one memory communicatively connected to the processor, wherein: the memory stores program instructions executable by the processor, and the processor calls the program instructions to execute the lane-level road binding method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a computer, the lane-level road binding method as described in any one of claims 1 to 7 is executed.

Images