Vehicle state estimation method, device, equipment and storage medium

Abstract

The present invention belongs to the technical field of environmental perception, and provides a method, device, equipment and storage medium for estimating the state of a vehicle, by determining multiple motion models in a vehicle maneuvering scene; for each motion model, determining the interactive state of the vehicle output by the motion model and other motion models in the multiple motion models; obtaining the predicted state of the vehicle according to the interactive state of the vehicle and the state transition information of the motion model; determining the probability of the measurement data being associated with the extended target area according to the predicted state of the vehicle, the measurement data of the vehicle and the prior probability of the measurement data in the extended target area of ​​the vehicle; obtaining the target state of the vehicle according to the probabilities of multiple motion models, the probability of the measurement data being associated with the extended target area and the predicted state of the vehicle output by each motion model. This method solves the problem of uncertainty in the motion model caused by the vehicle as an extended target when performing maneuvering motion, reduces tracking errors and improves tracking accuracy.

Description

Technical Field

[0001] The present invention relates to the field of environmental perception technology, and in particular to a vehicle state estimation method, device, equipment and storage medium. Background Art

[0002] Intelligent vehicles are vehicles equipped with perception, reasoning and execution mechanisms. They rely on sensor systems and powerful communication networks to exchange information between people, vehicles and roads, and actively analyze the driving environment for safety and plan the drivable area. In the perception system, state estimation is an important technical link in target tracking. As the bandwidth and resolution of vehicle-mounted radars are getting higher and higher, the resolution unit size is getting smaller and smaller. When the physical size of the target is larger than the resolution unit size, the target often occupies multiple resolution units, resulting in multiple measurements. In other words, the target is extensible. Tracking of extended targets such as vehicles is a problem that needs to be solved urgently.

[0003] For the tracking problem of extended targets such as vehicles, the existing state estimation method is mainly the data-region association (DRA) method. The DRA method assumes the shape of the vehicle target to be a rectangle, divides the extended rectangle into five regions, and models the scattering center for each region. However, the DRA method assumes that the target moves in a certain model, but in actual scenarios, the target's motion model may change over time, that is, it may make maneuvers. For example, vehicles often make sudden braking, acceleration, and sharp turns. At this time, a single dynamic model cannot well describe the maneuvering behavior of the target. Once the target makes a maneuver, the tracking performance of the DRA method will be greatly reduced, which is likely to cause a sudden increase in tracking errors. Summary of the invention

[0004] The present invention provides a vehicle state estimation method, device, equipment and storage medium, which are used to solve the defect of the prior art that the target tracking error of the vehicle in a maneuvering scene becomes larger, reduce the tracking error, and improve the tracking accuracy of the vehicle.

[0005] In a first aspect, the present invention provides a method for estimating a vehicle state, comprising:

[0006] Determine multiple motion models for vehicle maneuvering scenarios;

[0007] For each of the motion models, determining an interaction state of the vehicle output by the motion model interacting with other motion models in the plurality of motion models;

[0008] Obtaining a predicted state of the vehicle according to the interaction state of the vehicle and the state transition information of the motion model;

[0009] Determining a probability that the measurement data is associated with the extended target area according to the predicted state of the vehicle, the measurement data of the vehicle, and a priori probability that the measurement data is in the extended target area of ​​the vehicle;

[0010] The target state of the vehicle is obtained according to the probabilities of the multiple motion models, the probability of the measurement data being associated with the extended target area, and the predicted state of the vehicle output by each of the motion models.

[0011] Optionally, obtaining the target state of the vehicle according to the probabilities of the multiple motion models, the probability of the measurement data being associated with the extended target area, and the predicted state of the vehicle output by each of the motion models may include:

[0012] Performing weighted processing on the predicted state of the vehicle output by each of the motion models according to the probability of association between the measurement data and the extended target area, to obtain the target state of the vehicle output by each of the motion models;

[0013] The target state of the vehicle output by each of the motion models is weighted according to the probabilities of the multiple motion models to obtain the target state of the vehicle.

[0014] Optionally, the probabilities of the multiple motion models are obtained by updating the prior probabilities of the multiple motion models according to the probability density distribution function; the prior probabilities of the multiple motion models are obtained by multiplying the probabilities of the multiple motion models at the previous moment by the model transfer probability; the model transfer probability is determined based on a preset Markov transfer probability matrix, and the Markov transfer probability matrix represents the probability of mutual transfer between the multiple motion models.

[0015] Optionally, determining an interaction state of the vehicle output by interaction between the motion model and other motion models among the multiple motion models includes:

[0016] The interaction state of the vehicle is obtained according to the target state of the vehicle at the last moment, the probability of the motion model and the model transition probability of converting the other motion models into the target motion model.

[0017] Optionally, the measurement data includes at least one of the following: a distance between a measurement point and a measurement device of the vehicle, a radial Doppler velocity between the measurement point and the measurement device of the vehicle, and an azimuth between the measurement point and a line of sight of the measurement device of the vehicle.

[0018] Optionally, determining the probability of the measurement data being associated with the extended target area according to the predicted state of the vehicle, the measurement data of the vehicle, and the prior probability of the measurement data being in the extended target area of ​​the vehicle includes:

[0019] Determining a probability density distribution function of a measurement prediction value corresponding to the measurement data according to the predicted state of the vehicle and the measurement data of the vehicle;

[0020] The probability of the measurement data being associated with the extended target area is obtained according to the probability density distribution function of the measurement prediction value corresponding to the measurement data and the prior probability of the measurement data being in the extended target area of ​​the vehicle.

[0021] Optionally, before obtaining the probability of the measurement data being associated with the extended target area according to the probability density distribution function of the measurement prediction value corresponding to the measurement data and the prior probability of the measurement data in the extended target area of ​​the vehicle, the method further includes:

[0022] According to the predicted state of the vehicle, an extended target area of ​​the vehicle is calculated; the extended target area is a rectangular wave gate range, and the rectangular wave gate range includes five sub-areas.

[0023] Optionally, the prior probability of the measurement data in the extended target area of ​​the vehicle is obtained according to the predicted state of the vehicle and a preset regional distribution probability of the measurement data.

[0024] Optionally, the multiple motion models include at least one of the following: a uniform velocity motion CV model, a uniform acceleration motion CA model and a coordinated turning motion CT model.

[0025] Optionally, after obtaining the target state of the vehicle, the method may further include:

[0026] The target state of the vehicle at time k+1 is determined according to the target state of the vehicle at time k and the state transfer information of the target motion model at time k; k is an integer greater than or equal to 0.

[0027] In a second aspect, the present invention provides a vehicle state estimation device, comprising:

[0028] A model determination module, used to determine multiple motion models in vehicle maneuvering scenarios;

[0029] An interaction prediction module, for determining, for each of the motion models, an interaction state of the vehicle output by the motion model interacting with other motion models in the plurality of motion models; and obtaining a predicted state of the vehicle according to the interaction state of the vehicle and state transition information of the motion model;

[0030] The target estimation module is used to determine the probability of the measurement data being associated with the extended target area based on the predicted state of the vehicle, the measurement data of the vehicle and the prior probability of the measurement data in the extended target area of ​​the vehicle; and obtain the target state of the vehicle based on the probabilities of the multiple motion models, the probability of the measurement data being associated with the extended target area and the predicted state of the vehicle output by each of the motion models.

[0031] In a third aspect, the present invention provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, a method for estimating the vehicle state as described in any one of the above is implemented.

[0032] In a fourth aspect, the present invention provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a vehicle state estimation method as described in any one of the above.

[0033] In a fifth aspect, the present invention provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the vehicle state estimation methods described above.

[0034] The present invention provides a method, device, equipment and storage medium for estimating a vehicle state. The method first determines multiple motion models in a vehicle maneuvering scenario, and for each motion model, determines the interaction state of the vehicle output by the motion model interacting with other motion models in the multiple motion models. Then, based on the interaction state of the vehicle and the state transition information of the motion model, the predicted state of the vehicle is obtained. Furthermore, based on the predicted state of the vehicle, the measurement data of the vehicle and the prior probability of the measurement data in the extended target area of ​​the vehicle, the probability of the measurement data being associated with the extended target area is determined. Finally, based on the probabilities of the multiple motion models, the probability of the measurement data being associated with the extended target area and the predicted state of the vehicle output by each motion model, the target state of the vehicle is obtained. Because the state estimation method uses multiple interactive motion models to predict the state of the vehicle, that is, the influencing factors of the interactive outputs between multiple motion models on the target state of the vehicle are considered, and the probability of the measurement data being associated with the extended target area is determined through the predicted state of the vehicle, the measurement data of the vehicle and the prior probability of the measurement data in the extended target area of ​​the vehicle. Then, the predicted state of the vehicle output by each motion model is updated using the probabilities of the multiple motion models, and finally the target state of the vehicle is estimated, thereby solving the problem of motion model uncertainty caused by the extended target of the vehicle performing maneuvering motion, reducing tracking errors and improving tracking accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0036] Figure 1 It is one of the flow charts of the method for estimating the vehicle state provided by the present invention;

[0037] Figure 2 This is the second flow chart of the method for estimating the vehicle state provided by the present invention;

[0038] Figure 3 is a schematic diagram of the rectangular wave gate range of the extended target area of ​​the vehicle;

[0039] Figure 4a One of the schematic diagrams of the principle of the prior probability of the measurement data in the extended target area of ​​the vehicle;

[0040] Figure 4b They are respectively the second schematic diagram of the principle of the prior probability of the measurement data in the extended target area of ​​the vehicle;

[0041] Figure 5 The third flowchart of the method for estimating the vehicle state provided by the present invention;

[0042] Figure 6 It is a structural schematic diagram of a vehicle state estimation device provided by the present invention;

[0043] Figure 7 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] First, the terms involved in the embodiments of the present invention are explained:

[0046] Extended target: In a high-resolution sensor, the echo signal of each moving target may occupy multiple distance resolution units at the same time, that is, each target may generate multiple measurement data. Such a target is called an extended target.

[0047] Maneuverability: refers to the fact that vehicle targets often perform maneuvering behaviors such as sudden braking, acceleration, and sharp turns.

[0048] Secondly, the application scenarios of the vehicle state estimation method provided by the embodiment of the present invention are introduced.

[0049] Autonomous driving is an important application in smart transportation, which can realize functions such as emergency braking, pedestrian detection, collision warning, lane change assistance, automatic parking and adaptive cruise control. Perception is a prerequisite for autonomous driving and an important link. As smart cars become more and more intelligent, autonomous driving has higher and higher requirements for perception. The on-board millimeter-wave radar used in the perception link plays an important role in environmental perception and can accurately measure the information of targets around the vehicle. In addition, the on-board millimeter-wave radar can effectively penetrate fog, mist, rain and dust, is not constrained by bad weather and complex roads, and can work all day and all day. These advantages make it one of the important sensors in autonomous driving, and also make on-board radar an important research object. In the perception system, state estimation is an important technical link in target tracking, which refers to determining the state of targets such as vehicles based on observation and prediction.

[0050] In the prior art, the vehicle state estimation method is mainly based on the random hypersurface model (RHM) and the random matrix model (RMM).

[0051] Among them, the random hypersurface model (RHM) mainly uses the random hypersurface model to perceive and describe the vehicle, uses the hypersurface to describe its arbitrary extended shape, and performs state prediction based on the motion model during state estimation. The shortcomings are:

[0052] (1) Using a single model to describe the measurement spatial distribution of vehicles in the extended target area, a single model cannot describe the measurement spatial distribution of vehicles in this extended target area;

[0053] (2) Assuming that the target moves in a certain motion model, a priori assumptions are made about the vehicle's motion model. However, in actual traffic scenarios, vehicles often perform sudden braking, acceleration, and sharp turns. If the vehicle performs the above-mentioned maneuvers, the existing vehicle state estimation method cannot solve the problem of motion model uncertainty, which can easily lead to a sudden increase in tracking errors.

[0054] Among them, the random matrix model (RMM) assumes that the extended target is an ellipse and uses a random matrix to model the extended state of the extended target. The shortcomings of this method are that it cannot accurately describe the extended target, only using an ellipse cannot meet the application of most scenarios, and the scattering centers are considered to be uniformly distributed on the extended target. In fact, the measurement space distribution of vehicle targets is very complex, and the uniformity assumption is often not valid.

[0055] Based on the above-mentioned deficiencies, the present invention provides a method for estimating the vehicle state which can solve the problem of uncertainty in the motion model caused by the mobility of the vehicle as an extended target when performing maneuvering motion, reduce the tracking error, and improve the robustness of vehicle tracking of the maneuvering extended target.

[0056] Combine the following Figure 1-Figure 7 The technical solution for estimating the vehicle state of the present invention is described in detail. The following specific embodiments may be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0057] Figure 1 FIG. 1 is a flow chart of a method for estimating a vehicle state provided by the present invention, such as Figure 1 As shown, the method includes:

[0058] Step 101, determining multiple motion models in a vehicle maneuvering scenario;

[0059] Specifically, in actual scenarios, vehicles often perform various maneuvers, such as acceleration, turning, sudden braking, etc. Common target motion models include constant velocity (CV) motion model, constant acceleration (CA) motion model, and coordinated turning (CT) motion model. In this step, multiple motion models in the vehicle maneuvering scenario are first determined. For example, the multiple motion models are constant velocity CV motion model and constant acceleration (CA). The corresponding target state vectors at time k can be expressed as:

[0060]

[0061] in represents the target state vector at time k, x k ,y k represents the position of the target state vector at time k, and Indicates the coordinate positions of the vertices of the rectangular extended target in the target coordinate system.

[0062]

[0063] in represents the target state vector at time k, x k ,y k represents the position of the target state vector at time k, and Indicates the coordinate positions of the vertices of the rectangular extended target in the target coordinate system.

[0064] Step 102: for each motion model, determining an interaction state of a vehicle outputted by the motion model in interaction with other motion models in the plurality of motion models;

[0065] Specifically, after determining multiple motion models in a vehicle maneuvering scenario, the interactive states of the multiple motion models' interactive outputs can be calculated. In order to better understand the process of interactive output, one motion model can be assumed as the main motion model.

[0066] After the motion model is determined, the interaction state of the vehicle output by the motion model and other motion models in the motion model can be obtained based on the interaction between the motion model and other motion models in the multiple motion models.

[0067] Step 103: Obtain the predicted state of the vehicle according to the interaction state of the vehicle and the state transition information of the motion model;

[0068] Specifically, after obtaining the interactive state of the vehicle output by the motion model and other motion models in the motion model in step 102, the interactive state of the vehicle can be further updated based on the state transition information of the motion model to obtain the predicted state of the vehicle. The state transition information can be, for example, a state transition matrix corresponding to the motion model. Exemplarily, the predicted state of the vehicle can be expressed as:

[0069]

[0070] in, represents the predicted state of the vehicle at time k, F k-1 represents the state transfer matrix of the motion model (jth motion model) at time k-1, represents the interaction state of the vehicle at time k.

[0071] Step 104 : Determine the probability of the measurement data being associated with the extended target area according to the predicted state of the vehicle, the measurement data of the vehicle, and the prior probability of the measurement data being in the extended target area of ​​the vehicle.

[0072] Specifically, the vehicle measurement data refers to the measurement data of the vehicle's measurement device, such as a vehicle-mounted radar, at a measurement point in the extended target area. The specific information of the measurement data includes, for example, the distance d between the measurement point and the vehicle's measurement device, the radial Doppler velocity D between the measurement point and the vehicle's measurement device, and the azimuth angle β between the measurement point and the vehicle's measurement device. The extended target area refers to the extended area of ​​the vehicle target, such as a rectangular extended area. When the extended shape of the vehicle target is a rectangular extended area, the rectangular extended area can be further divided into five sub-areas.

[0073] Among them, the prior probability refers to the probability obtained based on past experience and analysis, and it often appears as the "cause" in the problem of "seeking the effect from the cause". The posterior probability refers to the probability estimate that is closer to the actual situation after correcting the original prior probability based on new information, and is the "cause" in the problem of "seeking the cause from the effect". In this method, the prior probability of the measurement data in the extended target area of ​​the vehicle can be derived based on the predicted state of the vehicle and the regional distribution probability of the pre-set measurement data.

[0074] Further, the prior probability of the measurement data in the extended target area of ​​the vehicle can be obtained based on the latest state of the vehicle, that is, the predicted state of the vehicle, combined with the pre-set regional distribution probability of the measurement data. After the predicted state of the vehicle is obtained in step 103, the probability of the measurement data being associated with the extended target area can be obtained based on the measurement data of the vehicle and the prior probability of the measurement data in the extended target area of ​​the vehicle.

[0075] Specifically, the probability density distribution function of the measurement prediction value corresponding to the measurement data can be calculated based on the predicted state of the vehicle and the measurement data of the vehicle, such as the likelihood function of the measurement prediction value corresponding to the measurement data. Then, based on the likelihood function of the measurement prediction value corresponding to the measurement data and the prior probability of the measurement data in the extended target area of ​​the vehicle, the probability of the measurement data being associated with the extended target area is determined, which can be understood as the posterior probability, which is the corrected data.

[0076] Step 105 : Obtain a target state of the vehicle according to the probabilities of the multiple motion models, the probability of the measurement data being associated with the extended target area, and the predicted state of the vehicle output by each motion model.

[0077] Specifically, the probability of multiple motion models can be understood as the probability of multiple motion models appearing in a vehicle maneuvering scene at a certain moment. Correspondingly, the prior probability of multiple motion models can be obtained by the probability of multiple motion models at the previous moment and the model transition probability. Furthermore, the predicted state of the vehicle output by each motion model can be subjected to secondary weighting processing according to the probability of the measurement data being associated with the extended target area and the probability of multiple motion models, thereby obtaining the target state of the vehicle.

[0078] In the method provided by the embodiment of the present invention, multiple motion models in a vehicle maneuvering scenario are first determined, and for each motion model, the interaction state of the vehicle output by the motion model interacting with other motion models in the multiple motion models is determined, and then the predicted state of the vehicle is obtained based on the interaction state of the vehicle and the state transition information of the motion model. Furthermore, based on the predicted state of the vehicle, the measurement data of the vehicle and the prior probability of the measurement data in the extended target area of ​​the vehicle, the probability of the measurement data being associated with the extended target area is determined, and finally the target state of the vehicle is obtained based on the probabilities of the multiple motion models, the probability of the measurement data being associated with the extended target area and the predicted state of the vehicle output by each motion model. Because the state estimation method uses multiple interactive motion models to predict the state of the vehicle, that is, the influencing factors of the interactive outputs between multiple motion models on the target state of the vehicle are considered, and the probability of the measurement data being associated with the extended target area is determined through the predicted state of the vehicle, the measurement data of the vehicle and the prior probability of the measurement data in the extended target area of ​​the vehicle. Then, the predicted state of the vehicle output by each motion model is updated using the probabilities of the multiple motion models, and finally the target state of the vehicle is estimated, thereby solving the problem of motion model uncertainty caused by the extended target of the vehicle performing maneuvering motion, reducing tracking errors and improving tracking accuracy.

[0079] Optionally, obtaining the target state of the vehicle according to the probabilities of the multiple motion models, the probability of the measurement data being associated with the extended target area, and the predicted state of the vehicle output by each motion model may include:

[0080] The predicted state of the vehicle output by each motion model is weighted according to the probability of association between the measurement data and the extended target area to obtain the target state of the vehicle output by each motion model;

[0081] The target state of the vehicle output by each motion model is weighted according to the probabilities of the multiple motion models to obtain the target state of the vehicle.

[0082] Specifically, after obtaining the predicted state of the vehicle output by each motion model, the predicted state of the vehicle output by each motion model can be firstly subjected to first-level probability weighting according to the probability of association between the measurement data and the extended target area to obtain the target state of the vehicle output by each motion model, as shown in the following example:

[0083]

[0084]

[0085] in, represents the target state of the vehicle given the j-th motion model output, is the probability of the measurement data being associated with the extended target area, represents the error covariance matrix corresponding to the target state of the vehicle given the j-th motion model output.

[0086] Furthermore, the target state of the vehicle output by each motion model can be subjected to secondary probability weighting processing according to the probabilities of the multiple motion models to obtain the target state of the vehicle, as shown in the following example:

[0087]

[0088]

[0089] in, represents the target state of the vehicle, Represents the probabilities of multiple motion models.

[0090] It is understandable that the probabilities of the above-mentioned multiple motion models can be obtained by updating the prior probabilities of the multiple motion models according to the probability density distribution function of the measurement prediction value corresponding to the measurement data, such as the likelihood function.

[0091] The method provided in the embodiment of the present invention more completely describes the coupling relationship between the vehicle's mobility characteristics and the extended target area. The method of this embodiment can well model the complex measurement space distribution on the vehicle's extended target area, effectively solve the problem of the vehicle's extended target maneuvering movement, and at the same time utilize the strong correlation characteristics between the extended target's morphology and maneuvering mode to significantly improve the target tracking performance.

[0092] Optionally, the probabilities of multiple motion models are obtained by updating the prior probabilities of multiple motion models according to a probability density distribution function; the prior probabilities of multiple motion models are obtained by multiplying the probabilities of multiple motion models at the previous moment by the model transfer probability; the model transfer probability is determined based on a preset Markov transition probability matrix, and the Markov transition probability matrix represents the probability of mutual transfer between multiple motion models.

[0093] Specifically, the model transfer probability can first be determined according to a preset model transfer matrix, such as a Markov transfer probability matrix, wherein the Markov transfer probability matrix represents the probability of mutual transfer between multiple motion models, and can be customized according to actual conditions. Exemplarily, the Markov transfer probability matrix can be expressed as:

[0094]

[0095] The rows and columns represent multiple preset motion models in order, for example, the first row, the second row, and the third row represent the uniform acceleration motion CA model, the coordinated turning motion CT model, and the uniform speed motion CV model, respectively; similarly, the first column, the second column, and the third column also represent the uniform acceleration motion CA model, the coordinated turning motion CT model, and the uniform speed motion CV model, respectively. The value "0.8" in the first row and the first column of the Markov transition probability matrix indicates that the probability of the uniform acceleration motion CA model transferring to the uniform acceleration motion CA model is 0.8, and the value "0.05" in the first row and the second column of the Markov transition probability matrix indicates that the probability of the uniform acceleration motion CA model transferring to the uniform acceleration motion CA model is 0.05.

[0096] Furthermore, the prior probabilities of the multiple motion models can be obtained by multiplying the probabilities of the multiple motion models at the previous moment by the model transition probability. The prior probabilities of the multiple motion models are expressed as follows:

[0097]

[0098] in, represents the prior probability of multiple motion models, represents the probability of multiple motion models at time k-1, π ij It represents the model transition probability of other motion models transferring to the given motion model at time k-1.

[0099] Furthermore, the prior probabilities of the multiple motion models may be updated according to the probability density distribution function of the measurement prediction value corresponding to the measurement data, such as the likelihood function, to obtain the probabilities of the multiple motion models. The process example is as follows:

[0100]

[0101]

[0102] in, represents the prior probability of multiple motion models, represents the probability of multiple motion models, Represents the likelihood function of the measurement prediction value corresponding to the measurement data.

[0103] In the method provided by the embodiment of the present invention, the model transfer probability is determined based on a preset model transfer matrix, such as a Markov transfer probability matrix, and the prior probabilities of multiple motion models are obtained by multiplying the probabilities of multiple motion models at the previous moment by the model transfer probability. Then, the prior probabilities of multiple motion models are updated according to the probability density distribution function to obtain the probabilities of multiple motion models. Subsequently, the predicted state of the vehicle output by each motion model is further weighted according to the probabilities of the multiple motion models, and the influence of the interaction of multiple models on the target state of the vehicle is fully considered. The estimation of the target state of the vehicle is finally obtained, which is more accurate and has better tracking performance.

[0104] Optionally, determining an interaction state of a vehicle outputted by the motion model in interaction with other motion models in the plurality of motion models may include:

[0105] The interaction state of the vehicle is obtained according to the target state of the vehicle at the previous moment, the probability of the motion model and the model transition probability of other motion models converted into the motion model.

[0106] Specifically, in the above step 102, the interaction state of the vehicle can be obtained according to the target state of the vehicle at the previous moment, the probability of the motion model and the model transition probability of other motion models converted into the motion model. For example, the motion model is assumed to be a uniform CV motion model. In this step, in addition to calculating the state of the motion model output, it is also necessary to calculate the output of the motion model after interaction with other motion models such as the uniform acceleration CA motion model and the coordinated turning CT motion model.

[0107] Exemplarily, the interaction state of the vehicle can be expressed as:

[0108]

[0109] in, represents the interaction state of the vehicle at time k given the jth motion model, represents the target state of the vehicle at time k-1, π ij represents the model transition probability, Represents the probability of multiple motion models at time k-1.

[0110] In the method provided by the embodiment of the present invention, the interactive state of the vehicle is obtained through the target state of the vehicle at the previous moment, the probability of the motion model and the model transition probability of other motion models converted into the motion model. The interactive state of the vehicle fully considers the uncertainty of the motion model of the vehicle in the maneuvering scene, and estimates the state of the vehicle through the interactive output of multiple motion models, which is more suitable for the maneuvering scene, so that the final target state of the vehicle is more accurate.

[0111] Optionally, the measurement data may include at least one of the following: a distance between the measurement point and a measurement device of the vehicle, a radial Doppler velocity between the measurement point and the measurement device of the vehicle, and an azimuth between the measurement point and a line of sight of the measurement device of the vehicle.

[0112] Specifically, refer to Figure 2 As shown, the measurement data of the vehicle may include, for example, the distance d between the measurement point and the scene vehicle measurement device, the radial Doppler velocity D between the measurement point and the vehicle's measurement device, and the azimuth angle β between the measurement point and the sight axis of the vehicle's measurement device. The extended target area of ​​the vehicle includes five sub-areas, namely, four sub-areas around the vehicle in the figure, represented by different filling background colors, and a sub-area at the center of the vehicle.

[0113] The vehicle's measuring device may be a vehicle-mounted radar, such as a vehicle-mounted millimeter-wave radar, a vehicle-mounted infrared / laser radar. The vehicle-mounted millimeter-wave radar can effectively penetrate fog, smoke, rain and dust, is not constrained by bad weather and complex roads, and can work around the clock. These advantages make it one of the important sensors in autonomous driving, and also make vehicle-mounted radar an important research object.

[0114] In the method provided in the embodiment of the present invention, the vehicle's measurement data includes data in multiple dimensions, including the distance between the measurement point and the vehicle's measurement device, the radial Doppler velocity between the measurement point and the vehicle's measurement device, and the azimuth angle between the measurement point and the vehicle's measurement device's line of sight, which can facilitate subsequent more accurate prediction of the vehicle's target state and improve prediction accuracy.

[0115] Optionally, determining the probability of the measurement data being associated with the extended target area according to the predicted state of the vehicle, the measurement data of the vehicle, and the prior probability of the measurement data being in the extended target area of ​​the vehicle may include:

[0116] Determine a probability density distribution function of a measurement prediction value corresponding to the measurement data according to the predicted state of the vehicle and the measurement data of the vehicle;

[0117] The probability of the measurement data being associated with the extended target area is obtained according to the probability density distribution function of the measurement prediction value corresponding to the measurement data and the prior probability of the measurement data in the extended target area of ​​the vehicle.

[0118] Specifically, it should be understood that, because the extended target area includes multiple sub-areas, for example, the rectangular extended area includes five sub-areas, the associated probability represents the probability of the specific sub-area that the measurement data falls into. Among them, the prior probability of the extended target area of ​​the vehicle and the probability of the extended target area association can be understood as the prior probability and the posterior probability of the association between the measurement data and the extended target area, respectively.

[0119] The prior probability of the extended target area of ​​the vehicle refers to the probability of the vehicle's extended target area appearing according to the current prediction state and the pre-set measurement data. For example, the prior probability of the i-th measurement data in the extended target area of ​​the vehicle is:

[0120]

[0121] in, represents the prior probability of the i-th measurement data in the extended target area of ​​the vehicle, θ near ,θ′ near ,θ far and θ′ far Indicates the position angle relationship between the measurement point and the vehicle measurement device based on the rectangular wave gate range.

[0122] Furthermore, the probability of the measurement data being associated with the extended target area can be understood as a posterior probability, that is, based on the prior probability of the extended target area of ​​the above vehicle, it is updated by the measurement data of the vehicle at the current moment. For example, the probability of the measurement data being associated with the extended target area can be expressed as: An example of a process of obtaining the measurement data and the probability of the measurement data being associated with the extended target area is as follows:

[0123] In a specific embodiment, firstly, a probability density distribution function of a measurement prediction value corresponding to the measurement data, such as a likelihood function of the measurement prediction value corresponding to the measurement data, can be determined according to the predicted state of the vehicle and the measurement data of the vehicle. The process of obtaining the likelihood function of the measurement prediction value corresponding to the measurement data is exemplified as follows:

[0124] Assume that the vehicle generates N k A measurement, represented by a vector:

[0125]

[0126] Among them, the measurement data The information includes the distance d between the measuring point and the measuring device of the vehicle, the radial Doppler velocity D between the measuring point and the measuring device of the vehicle, and the azimuth angle β between the measuring point and the line of sight of the measuring device of the vehicle.

[0127] The measurement model is, for example:

[0128]

[0129] Among them, H i About ( If i<5, If i=5), i from 1 to 5 indicates that the measurement model parameters of the five regions are different. The specific form is:

[0130]

[0131]

[0132]

[0133]

[0134]

[0135] In the formula, (or ) represents a random variable uniformly distributed in the interval [0,1], that is, the scattering centers are uniformly distributed in each region.

[0136] A 1 =[[0,1] T , [-1,0] T ],

[0137] A 2 =[A 5 , O 2×4 ],

[0138]

[0139]

[0140]

[0141] in represents the i-th n-dimensional basis vector.

[0142] Depending on the different sub-areas of the extended target area associated with the measurement data, the form of H is different, and the corresponding measurement equations are different, which can achieve the purpose of partition modeling measurement distribution. And given motion model Under the condition of , the measurement equation can be:

[0143]

[0144] in, It is the measurement data The measurement model of is a target measurement model for multiple measurements. In order to obtain the likelihood function of the measurement prediction value corresponding to the measurement data, the state vector can be expanded to:

[0145]

[0146] Therefore, the above measurement data is updated as follows:

[0147]

[0148] Among them, UT represents the weight The sampling point is The traceless transformation of yes The lth sampling point of .

[0149] Furthermore, the likelihood function of the measured prediction value corresponding to the measured data can be obtained as:

[0150]

[0151] in, Represents the likelihood function of the measurement prediction value corresponding to the measurement data, Z k represents the measured predicted value, It represents the mean and error covariance of the measurement prediction values ​​corresponding to the measurement data obtained by unscented transformation.

[0152] Furthermore, the probability (posterior probability) that the measurement data is associated with the extended target area can be obtained based on the probability density distribution function of the measurement prediction value corresponding to the measurement data and the prior probability of the measurement data in the extended target area of ​​the vehicle. For example, the process of calculating the probability (posterior probability) that the measurement data is associated with the extended target area based on the likelihood function of the measurement prediction value corresponding to the measurement data is as follows:

[0153]

[0154]

[0155] in, represents the probability of the measurement data being associated with the extended target area, Represents the likelihood function of the measurement prediction value corresponding to the measurement data, Represents the prior probability that the measurement data is in the extended target area of ​​the vehicle.

[0156] In the method provided in an embodiment of the present invention, a probability density distribution function, such as a likelihood function, of a measurement prediction value corresponding to the measurement data is determined based on the predicted state of the vehicle and the measurement data of the vehicle, and the probability of the measurement data being associated with the extended target area is obtained by further combining the prior probability of the measurement data in the extended target area of ​​the vehicle. The predicted state of the vehicle can be subsequently updated based on the probability of the measurement data being associated with the extended target area, so that the correlation between the measurement data and the extended target area is fully considered in the target state of the vehicle finally obtained, thereby solving the problem that the complex spatial distribution of the measurement caused by the scalability of the vehicle is difficult to describe.

[0157] Optionally, before obtaining the probability of the measurement data being associated with the extended target area according to the probability density function of the measurement prediction value corresponding to the measurement data and the prior probability of the measurement data being in the extended target area of ​​the vehicle, the method may further include:

[0158] According to the predicted state of the vehicle, the extended target area of ​​the vehicle is calculated; the extended target area is a rectangular wave gate range, and the rectangular wave gate range includes five sub-areas.

[0159] Specifically, the extended target area of ​​the vehicle may be calculated according to the predicted state of the vehicle, wherein the extended target area of ​​the vehicle is a rectangular wave gate range, and the rectangular wave gate range includes five sub-areas.

[0160] like Figure 3 As shown, the gate range of each sub-area of ​​the extended target area is a rectangular area. Exemplarily, the size of the rectangular area is calculated as follows:

[0161]

[0162]

[0163]

[0164]

[0165] Among them, t and is a scaling factor on (0,1) that can be used to scale the size of the gate range.

[0166] Furthermore, the vertices of the target area can be expanded to further predict the target state of the vehicle, and the gate range of each area can be calculated. A measurement may fall within the gate range of multiple sub-areas, and the nth data area combination generated can be expressed as

[0167] In the method provided in an embodiment of the present invention, the rectangular wave gate range of the extended target area of ​​the vehicle is calculated based on the predicted state of the vehicle, so that the predicted state of the vehicle can be further estimated based on the measurement data and the probability of the extended target area of ​​the vehicle. The final vehicle target state is more accurate and the tracking error is small.

[0168] Optionally, the prior probability of the measurement data in the extended target area of ​​the vehicle is obtained according to the predicted state of the vehicle and a preset regional distribution probability of the measurement data.

[0169] Specifically, the aforementioned prior probability of the measurement data in the extended target area of ​​the vehicle can be obtained based on the predicted state of the vehicle and the pre-set regional distribution probability of the measurement data. For example, the process of obtaining the prior probability of the measurement data in the extended target area of ​​the vehicle is as follows:

[0170] Figure 4a and Figure 4b Schematic diagram of the principle of the prior probability of the measurement data in the extended target area of ​​the vehicle. Figure 4a and Figure 4b As shown, represents the area associated with the i-th measurement data, P near , P far , P interior The pre-set probability satisfies:

[0171] P near +P far +P interior =1,

[0172] Among them, P interior Represents the probability that the measured data appears in the internal area, P near It represents the probability that the measured data appears in the area close to the edge of the radar, P far Indicates the probability that the measurement data appears in the area far away from the edge of the radar.

[0173] If the i-th measurement data appears in the area close to the edge of the radar, then:

[0174] The prior probability of the i-th measurement data in the extended target area of ​​the vehicle is:

[0175]

[0176] in, represents the prior probability of the i-th measurement data in the extended target area of ​​the vehicle, θ near ,θ′ near It represents the position angle relationship between the extended vertex measurement point based on the rectangular wave gate range and the vehicle measurement device.

[0177] On the contrary, that is, the i-th measurement data appears in the area far away from the edge of the radar:

[0178] The prior probability of the i-th measurement data in the extended target area of ​​the vehicle is:

[0179]

[0180] in, represents the prior probability of the i-th measurement data in the extended target area of ​​the vehicle, θ far ,θ′ far It represents the position angle relationship between the extended vertex measurement point based on the rectangular wave gate range and the vehicle measurement device.

[0181] In the method provided by the embodiment of the present invention, the prior probability of the measured data in the extended target area of ​​the vehicle can be obtained by combining the predicted state of the vehicle with the regional distribution probability of the pre-set measured data. This method can achieve real-time updating of the prior probability of the measured data and the extended target area of ​​the vehicle according to the latest estimated state of the vehicle, that is, the predicted state of the vehicle, so that the probability of the measurement data and the extended target area obtained based on the prior probability of the extended target area of ​​the vehicle is more in line with the actual scenario.

[0182] Optionally, the multiple motion models include at least one of the following: a uniform velocity motion CV model, a uniform acceleration motion CA model, and a coordinated turning motion CT model.

[0183] Specifically, during normal driving, the vehicle generally maintains a uniform motion. In actual scenarios, when the vehicle performs maneuvers such as sudden braking, acceleration, and sharp turns, etc. Correspondingly, there can be multiple motion models in this method, such as a uniform motion CV model, a uniform acceleration CA model, and a coordinated turning motion CT model.

[0184] In the method provided by the embodiment of the present invention, by determining multiple motion models corresponding to the vehicle maneuvering scenario, the motion models include, for example, a uniform motion CV model, a uniform acceleration motion CA model, and a collaborative turning motion CT model, and finally using an interactive multi-model to predict and estimate the target state of the vehicle, the error of target tracking is greatly reduced and the tracking accuracy is improved.

[0185] Optionally, after obtaining the target state of the vehicle, the following steps may also be included:

[0186] The target state of the vehicle at time k+1 is determined according to the target state of the vehicle at time k and the state transition information of the target motion model at time k; k is an integer greater than or equal to 0.

[0187] Specifically, after the target state of the vehicle is obtained in this method, the target state of the vehicle at the current moment, for example, moment k, and the state transition information of each motion model assumed at moment k can be used to determine the target state of the vehicle at moment k+1.

[0188] Optionally, the target state of the vehicle at time k+1 can be obtained by combining observation noise such as Gaussian noise and its corresponding covariance matrix and driving matrix, that is, the influence of observation noise on the target state estimation is taken into account. Exemplarily, the target state of the vehicle at time k+1 is expressed by the following formula:

[0189] x k+1 =F k x k +G k w k ,

[0190] Among them, x k+1 represents the target state of the vehicle at time k+1, F k represents the state transfer matrix of a given motion model at time k, x k represents the target state of the vehicle at time k, w k represents Gaussian noise with a mean of 0, G k Represents the driving matrix corresponding to Gaussian noise.

[0191] In the method provided by the embodiment of the present invention, the target state of the vehicle at time k+1 can be determined through the target state of the vehicle at time k and the state transition information of multiple motion models at time k, thereby realizing real-time recursive estimation of the vehicle target state.

[0192] The vehicle state estimation device provided by the present invention is described below. The vehicle state estimation device described below and the vehicle state estimation method described above can refer to each other.

[0193] Figure 5 This is a second flow chart of the vehicle state estimation method provided by the present invention.

[0194] like Figure 5 As shown, the method includes:

[0195] Step 501: Select multiple motion models that can match the vehicle maneuvering scene, initialize the observation noise variance of each motion model, the target state of each motion model and the corresponding error covariance matrix.

[0196] Step 502: for each motion model, according to the target state or initial state of the vehicle at the last moment, calculate the interaction state of the vehicle output by the motion model and other motion models in the plurality of motion models;

[0197] Step 503: Calculate the rectangular gate ranges of the four sides and the five inner sub-areas of the vehicle's extended target area according to the predicted state of the vehicle, and associate the measurement data with the five sub-areas;

[0198] Step 504: Calculate the probability density distribution function of the measurement prediction value corresponding to the measurement data using the untraceable transformation method, and update the interaction state of the vehicle to obtain the predicted state;

[0199] Step 505, calculating the probability of data region association according to the probability density distribution function of the measurement prediction value corresponding to the measurement data and the prior probability of the measurement data occurrence region;

[0200] Step 506: update the probabilities of the multiple motion models according to the probability density distribution function of the measurement prediction value corresponding to the measurement data;

[0201] Step 507: According to the probability of the measurement data being associated with the extended target area and the probabilities of the multiple motion models, a two-level probability weighted processing is performed on the predicted state of the vehicle output by each motion model to obtain the target state of the vehicle.

[0202] Specifically, we first select multiple motion models that can match the vehicle maneuvering scenario, such as the constant velocity (CV) motion model, the constant acceleration (CA) motion model, and the cooperative turning (CT) motion model. Furthermore, we can initialize the observation noise variance of each motion model, the target state of each motion model, and the corresponding error covariance matrix.

[0203] Furthermore, for each motion model, the interaction state of the vehicle output by the motion model interacting with other motion models in the multiple motion models can be calculated according to the target state or initial state of the vehicle at the previous moment.

[0204] Furthermore, according to the predicted state of the vehicle, the rectangular gate ranges of the four sides and the five inner sub-areas of the vehicle extended target area are calculated, and the measurement data are associated with the five sub-areas to obtain, for example, multiple data area combinations.

[0205] Furthermore, an untraceable transformation may be used to calculate a probability density distribution function of the measurement prediction value corresponding to the measurement data, such as a likelihood function, and the likelihood function of the measurement prediction value corresponding to the measurement data may be used to update the interaction state of the vehicle to obtain the predicted state of the vehicle.

[0206] Furthermore, the probability of data region association can be calculated based on the probability density distribution function of the measurement prediction value corresponding to the measurement data, such as the likelihood function, and the prior probability of the measurement data occurrence area, which can be understood as the posterior probability.

[0207] Furthermore, the prior probabilities of the multiple motion models are updated according to the probability density distribution function of the measurement prediction value corresponding to the measurement data to obtain the probabilities of the multiple motion models.

[0208] Finally, according to the probability of the measurement data being associated with the extended target area and the probabilities of multiple motion models, the predicted state of the vehicle output by each motion model is subjected to a two-level probability weighted processing to obtain the target state of the vehicle.

[0209] The method provided in the embodiment of the present invention solves the problem that the complex spatial distribution of the vehicle extended target measurement is difficult to describe, and at the same time solves the problem of uncertainty in the motion model when the vehicle extended target is maneuvering, thereby improving the robustness and accuracy of tracking the maneuvering extended target.

[0210] Figure 6 The vehicle state estimation device 600 provided by the present invention is as follows: Figure 6 As shown, the vehicle state estimation device 600 includes:

[0211] A model determination module 610 is used to determine multiple motion models in a vehicle maneuvering scenario;

[0212] An interaction prediction module 620 is used to determine, for each of the motion models, an interaction state of the vehicle output by the motion model interacting with other motion models in the multiple motion models; and obtain a predicted state of the vehicle according to the interaction state of the vehicle and the state transition information of the motion model;

[0213] The target estimation module 630 is used to determine the probability of the measurement data being associated with the extended target area based on the predicted state of the vehicle, the measurement data of the vehicle and the prior probability of the measurement data in the extended target area of ​​the vehicle; and obtain the target state of the vehicle based on the probabilities of the multiple motion models, the probability of the measurement data being associated with the extended target area and the predicted state of the vehicle output by each of the motion models.

[0214] In the device provided by the embodiment of the present invention, the model determination module 610 first determines multiple motion models in the vehicle maneuvering scenario, and the interaction prediction module 620 determines, for each motion model, the interaction state of the vehicle output by the interaction of the motion model with other motion models in the multiple motion models, and then obtains the predicted state of the vehicle based on the interaction state of the vehicle and the state transfer information of the motion model. Furthermore, the target estimation module 630 determines the probability of the measurement data being associated with the extended target area based on the predicted state of the vehicle, the measurement data of the vehicle and the prior probability of the measurement data in the extended target area of ​​the vehicle. Finally, the target state of the vehicle is obtained based on the probabilities of multiple motion models, the probability of the measurement data being associated with the extended target area and the predicted state of the vehicle output by each motion model. Because the state estimation device uses multiple interactive motion models to predict the state of the vehicle, that is, the influencing factors of the interactive outputs between the multiple motion models on the target state of the vehicle are considered, and the probability of the measurement data being associated with the extended target area is determined through the predicted state of the vehicle, the measurement data of the vehicle and the prior probability of the measurement data in the extended target area of ​​the vehicle. Then, the predicted state of the vehicle output by each motion model is updated using the probabilities of the multiple motion models, and finally the target state of the vehicle is estimated, thereby solving the problem of motion model uncertainty caused by the extended target of the vehicle performing maneuvering motion, reducing tracking errors and improving tracking accuracy.

[0215] Optionally, the target estimation module 630 is specifically configured to:

[0216] Performing weighted processing on the predicted state of the vehicle output by each of the motion models according to the probability of association between the measurement data and the extended target area, to obtain the target state of the vehicle output by each of the motion models;

[0217] The target state of the vehicle output by each of the motion models is weighted according to the probabilities of the multiple motion models to obtain the target state of the vehicle.

[0218] Optionally, the probabilities of the multiple motion models are obtained by updating the prior probabilities of the multiple motion models according to the probability density distribution function; the prior probabilities of the multiple motion models are obtained by multiplying the probabilities of the multiple motion models at the previous moment by the model transfer probability; the model transfer probability is determined based on a preset Markov transfer probability matrix, and the Markov transfer probability matrix represents the probability of mutual transfer between the multiple motion models.

[0219] Optionally, the interaction prediction module 620 is specifically used for:

[0220] The interaction state of the vehicle is obtained according to the target state of the vehicle at the last moment, the probability of the motion model, and the model transition probability of converting the other motion models into the target motion model.

[0221] Optionally, the measurement data may include at least one of the following: a distance between a measurement point and a measurement device of the vehicle, a radial Doppler velocity between the measurement point and the measurement device of the vehicle, and an azimuth between the measurement point and the line of sight of the measurement device of the vehicle.

[0222] Optionally, the target estimation module 630 is specifically configured to:

[0223] Determining a probability density distribution function of a measurement prediction value corresponding to the measurement data according to the predicted state of the vehicle and the measurement data of the vehicle;

[0224] The probability of the measurement data being associated with the extended target area is obtained according to the probability density distribution function of the measurement prediction value corresponding to the measurement data and the prior probability of the measurement data being in the extended target area of ​​the vehicle.

[0225] Optionally, the vehicle state estimation device 600 further includes an extended target area calculation module;

[0226] The extended target area calculation module is used to:

[0227] According to the predicted state of the vehicle, an extended target area of ​​the vehicle is calculated; the extended target area is a rectangular wave gate range, and the rectangular wave gate range includes five sub-areas.

[0228] Optionally, the prior probability of the measurement data in the extended target area of ​​the vehicle is obtained according to the predicted state of the vehicle output by the motion model and a preset regional distribution probability of the measurement data.

[0229] Optionally, the multiple motion models include at least one of the following: a uniform velocity motion CV model, a uniform acceleration motion CA model and a coordinated turning motion CT model.

[0230] Optionally, the vehicle state estimation device 600 further includes a target recursive module;

[0231] The target recursion module is used to:

[0232] The target state of the vehicle at time k+1 is determined according to the target state of the vehicle at time k and the state transfer information of the target motion model at time k; k is an integer greater than or equal to 0.

[0233] Figure 7 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 7 As shown, the electronic device may include: a processor 710, a communication interface 720, a memory 730 and a communication bus 740, wherein the processor 710, the communication interface 720 and the memory 730 communicate with each other through the communication bus 740. The processor 710 may call the logic instructions in the memory 730 to execute the vehicle state estimation method, which includes:

[0234] Determine multiple motion models for vehicle maneuvering scenarios;

[0235] For each of the motion models, determining an interaction state of the vehicle output by the motion model interacting with other motion models in the plurality of motion models;

[0236] Obtaining a predicted state of the vehicle according to the interaction state of the vehicle and the state transition information of the motion model;

[0237] Determining a probability that the measurement data is associated with the extended target area according to the predicted state of the vehicle, the measurement data of the vehicle, and a priori probability that the measurement data is in the extended target area of ​​the vehicle;

[0238] The target state of the vehicle is obtained according to the probabilities of the multiple motion models, the probability of the measurement data being associated with the extended target area, and the predicted state of the vehicle output by each of the motion models.

[0239] In addition, the logic instructions in the above-mentioned memory 730 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0240] On the other hand, the present invention further provides a computer program product, the computer program product includes a computer program, the computer program can be stored in a non-transitory computer-readable storage medium, when the computer program is executed by a processor, the computer can execute the vehicle state estimation method provided by the above methods, the method includes:

[0241] Determine multiple motion models for vehicle maneuvering scenarios;

[0242] For each of the motion models, determining an interaction state of the vehicle output by the motion model interacting with other motion models in the plurality of motion models;

[0243] Obtaining a predicted state of the vehicle according to the interaction state of the vehicle and the state transition information of the motion model;

[0244] Determining a probability that the measurement data is associated with the extended target area according to the predicted state of the vehicle, the measurement data of the vehicle, and a priori probability that the measurement data is in the extended target area of ​​the vehicle;

[0245] The target state of the vehicle is obtained according to the probabilities of the multiple motion models, the probability of the measurement data being associated with the extended target area, and the predicted state of the vehicle output by each of the motion models.

[0246] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for estimating the vehicle state provided by the above methods is implemented, and the method includes:

[0247] Determine multiple motion models for vehicle maneuvering scenarios;

[0248] For each of the motion models, determining an interaction state of the vehicle output by the motion model interacting with other motion models in the plurality of motion models;

[0249] Obtaining a predicted state of the vehicle according to the interaction state of the vehicle and the state transition information of the motion model;

[0250] Determining a probability that the measurement data is associated with the extended target area according to the predicted state of the vehicle, the measurement data of the vehicle, and a priori probability that the measurement data is in the extended target area of ​​the vehicle;

[0251] The target state of the vehicle is obtained according to the probabilities of the multiple motion models, the probability of the measurement data being associated with the extended target area, and the predicted state of the vehicle output by each of the motion models.

[0252] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative effort.

[0253] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiment.

[0254] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for estimating a vehicle state, characterized in that: include: Determine multiple motion models in a vehicle maneuvering scenario; the multiple motion models include at least one of the following: a uniform velocity motion CV model, a uniform acceleration motion CA model, and a coordinated turning motion CT model; For each of the motion models, determining an interaction state of the vehicle output by the motion model interacting with other motion models in the plurality of motion models; Obtaining a predicted state of the vehicle according to the interaction state of the vehicle and the state transition information of the motion model; Determining a probability that the measurement data is associated with the extended target area according to the predicted state of the vehicle, the measurement data of the vehicle, and a priori probability that the measurement data is in the extended target area of ​​the vehicle; Obtaining a target state of the vehicle according to the probabilities of the multiple motion models, the probability of the measurement data being associated with the extended target area, and the predicted state of the vehicle output by each of the motion models; The probabilities of the multiple motion models are obtained by updating the prior probabilities of the multiple motion models according to the probability density distribution function; the prior probabilities of the multiple motion models are obtained based on the multiplication of the probabilities of the multiple motion models at the previous moment and the model transition probability; the model transition probability is determined based on a preset Markov transition probability matrix, and the Markov transition probability matrix represents the probability of mutual transition between the multiple motion models; The rows and columns in the Markov transition probability matrix represent a plurality of preset motion models in order respectively; the values ​​in the Markov transition probability matrix represent the probability that the motion model corresponding to the row where the values ​​are located is transferred to the motion model corresponding to the column where the values ​​are located; The determining the interaction state of the vehicle output by the interaction between the motion model and other motion models in the plurality of motion models includes: Obtaining the interaction state of the vehicle according to the target state of the vehicle at the last moment, the probability of the motion model, and the model transition probability of converting the other motion models into the target motion model; The step of obtaining the interaction state of the vehicle according to the target state of the vehicle at the last moment, the probability of the motion model, and the model transition probability of converting the other motion models into the target motion model comprises: in, represents the interaction state of the vehicle at time k given the jth motion model, represents the target state of the vehicle at time k-1, π ij represents the model transition probability, Represents the probability of multiple motion models at time k-1; The obtaining of the target state of the vehicle according to the probabilities of the plurality of motion models, the probability of the measurement data being associated with the extended target area, and the predicted state of the vehicle output by each of the motion models comprises: Performing weighted processing on the predicted state of the vehicle output by each of the motion models according to the probability of association between the measurement data and the extended target area, to obtain the target state of the vehicle output by each of the motion models; Performing weighted processing on the target state of the vehicle output by each of the motion models according to the probabilities of the multiple motion models to obtain the target state of the vehicle; The measurement data includes at least one of the following: a distance between a measurement point and a measurement device of the vehicle, a radial Doppler velocity between the measurement point and the measurement device of the vehicle, and an azimuth between the measurement point and a boresight of the measurement device of the vehicle; The determining, based on the predicted state of the vehicle, the measurement data of the vehicle, and the prior probability of the measurement data being in the extended target area of ​​the vehicle, the probability of the measurement data being associated with the extended target area includes: Determining a probability density distribution function of a measurement prediction value corresponding to the measurement data according to the predicted state of the vehicle and the measurement data of the vehicle; The probability of the measurement data being associated with the extended target area is obtained according to the probability density distribution function of the measurement prediction value corresponding to the measurement data and the prior probability of the measurement data being in the extended target area of ​​the vehicle.

2. The method for estimating a vehicle state according to claim 1, characterized in that: Before obtaining the probability of the measurement data being associated with the extended target area according to the probability density distribution function of the measurement prediction value corresponding to the measurement data and the prior probability of the measurement data in the extended target area of ​​the vehicle, the method further includes: According to the predicted state of the vehicle, an extended target area of ​​the vehicle is calculated; the extended target area is a rectangular wave gate range, and the rectangular wave gate range includes five sub-areas.

3. The vehicle state estimation method according to claim 1 or 2, characterized in that: The prior probability of the measurement data in the extended target area of ​​the vehicle is obtained according to the predicted state of the vehicle and a preset regional distribution probability of the measurement data.

4. The vehicle state estimation method according to claim 1 or 2, characterized in that: After obtaining the target state of the vehicle, the method further includes: The target state of the vehicle at time k+1 is determined according to the target state of the vehicle at time k and the state transfer information of the target motion model at time k; k is an integer greater than or equal to 0.

5. A vehicle state estimation device, executing the vehicle state estimation method according to any one of claims 1 to 4, characterized in that: include: A model determination module, used to determine multiple motion models in vehicle maneuvering scenarios; An interaction prediction module, for determining, for each of the motion models, an interaction state of the vehicle output by the motion model interacting with other motion models in the plurality of motion models; and obtaining a predicted state of the vehicle according to the interaction state of the vehicle and state transition information of the motion model; The target estimation module is used to determine the probability of the measurement data being associated with the extended target area based on the predicted state of the vehicle, the measurement data of the vehicle and the prior probability of the measurement data in the extended target area of ​​the vehicle; and obtain the target state of the vehicle based on the probabilities of the multiple motion models, the probability of the measurement data being associated with the extended target area and the predicted state of the vehicle output by each of the motion models.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the vehicle state estimation method according to any one of claims 1 to 4 is implemented.

7. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the vehicle state estimation method according to any one of claims 1 to 4 is implemented.

Images