Method for realizing vehicle trajectory prediction in open-pit mine by combining random forest and kalman

Abstract

The present application relates to the unmanned technical field of open-pit mine area, and particularly relates to a method for realizing trajectory prediction of open-pit mine vehicle by combining random forest and Kalman, which comprises the following steps: obtaining current motion state information of mine vehicle by perception device; obtaining candidate reference path based on current motion state information and prior open-pit mine area high-precision map information, and obtaining path curvature of candidate reference path; processing current motion state information and path curvature based on random forest, and obtaining observation value of longitudinal velocity and yaw angle velocity of reference path; establishing state transition model and measurement model of extended Kalman filter, and taking observation value of longitudinal velocity and yaw angle velocity as measurement value of extended Kalman filter; performing state prediction and measurement update of extended Kalman filter, and finally obtaining prediction trajectory of mine vehicle at future time; the present application can improve trajectory prediction performance of open-pit mine vehicle.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of unmanned technology in open-pit mine area, and particularly relates to a method for realizing trajectory prediction of open-pit mine vehicle by combining random forest and Kalman. BACKGROUND

[0002] As a key component of intelligent transportation system construction, unmanned transportation in open-pit mine is not only one of the best scenarios for the wide application of unmanned technology, but also an important driving force for the development of unmanned transportation industry. The driving safety of unmanned mine truck in open-pit mine is particularly crucial, which is the core link to ensure the safe and efficient operation of the entire mine unmanned transportation system. The trajectory prediction of vehicles in open-pit mine area aims to predict the driving trajectory of other manned vehicles around the unmanned vehicle in the mine road scene, and to provide accurate prediction trajectory information for the unmanned vehicle to support its downstream modules such as traffic decision, obstacle avoidance planning and control execution.

[0003] The invention patent application with publication number CN118953406A discloses a vehicle trajectory prediction method based on attention mechanism, and the invention patent with publication number CN114495036B discloses an automatic driving vehicle trajectory prediction method and system based on double-cross Transformer. The technical solutions of the above include that after the historical data of the target vehicle and the surrounding vehicles are obtained, the interaction behaviors between the ego vehicle and other vehicles in the scene are encoded by the method of deep learning, which also includes the encoding of vehicle and lane interaction features, vehicle trajectory feature encoding and global environment feature encoding. These features need to be learned by a neural network model based on a large amount of scene data. Such methods are mainly applied in urban road environment, and are used for trajectory prediction of complex multi-modal interaction of a large number of passenger vehicles. If these methods are directly applied to trajectory prediction in open-pit mine area, there will be significant adaptability challenges. One of the main reasons is that in the urban road scene, there is clear lane line information, while in the open-pit mine area, there is no clear lane constraint for vehicle operation. How to effectively use the mine area map information to improve the trajectory prediction accuracy of the target vehicle is a difficulty to be solved. The second main reason is that data-driven prediction methods usually rely on a large amount of labeled scene data for training and learning, which consumes a lot of computing power and takes a long time for real-time calculation. There is no public motion trajectory dataset available for use at present. SUMMARY

[0004] In view of the above problems, the present application provides a method for realizing trajectory prediction of open-pit mine vehicle by combining random forest and Kalman, which solves the technical problems of vehicle trajectory prediction relying on labeled data and consuming a large amount of computing power in the prior art.

[0005] The application provides a method for realizing open-pit mine vehicle trajectory prediction by combining random forest and Kalman, comprising the following steps:

[0006] Step S1, acquiring current motion state information of a mine vehicle by a perception device, wherein the motion state information comprises position, heading angle, longitudinal speed, longitudinal acceleration and yaw rate of the mine vehicle;

[0007] Step S2, acquiring a candidate reference path based on the current motion state information and prior high-precision mine map information, and acquiring path curvature of the candidate reference path, wherein the candidate reference path is a path that the mine vehicle can travel in future time;

[0008] Step S3, processing the current motion state information and the path curvature based on random forest to acquire observation values of longitudinal speed and yaw rate of the mine vehicle;

[0009] Step S4, establishing a state transition model and a measurement model of extended Kalman filter, and taking the observation values of the longitudinal speed and the yaw rate as measurement values of the longitudinal speed and the yaw rate of the extended Kalman filter;

[0010] performing state prediction and measurement update of the extended Kalman filter to finally acquire a predicted trajectory of the mine vehicle in future time.

[0011] Preferably, in the step S1, the motion state information of the target vehicle at the current t time is represented as: x t =[x t ,y t ,θ t ,v t ,a t ,ω t ] T , wherein x t , y t , θ t , v t , a t and ω t represent east position coordinate, north position coordinate, heading angle, longitudinal speed, longitudinal acceleration and yaw rate at the current t time respectively.

[0012] Preferably, the step S2 specifically comprises:

[0013] Step S2-1, acquiring a candidate reference path set SET(Path i ) that can be possibly traveled by the target vehicle in front according to the current motion state information and prior high-precision mine map information, wherein Path i represents the i-th path;

[0014] Step S2-2: Divide the preset future time period into multiple time steps, denoted as t+kΔt, where t is the current time, Δt is the time step size, and k is the index of the time step; based on the assumption that the target vehicle moves at a constant speed, obtain the target vehicle position coordinates (px) for each time step. k ,py k ), as the pre-aiming point for the corresponding k time steps. in Let K be the coordinates of the preview point at time step k;

[0015] Step S2-3: Based on the position coordinates of the pre-aiming point and the candidate reference path set SET(Path) i The path in the target area is matched to obtain the reference path closest to the target point. The path curvature κ of the target point at time step k is determined by the reference path closest to the target point. k .

[0016] Preferably, step S3 specifically includes:

[0017] Step S3-1: Based on the target vehicle's motion state and the path curvature, establish the input of the random forest at time k.

[0018] Step S3-2: Input the data into the random forest for processing. The N trained decision trees of the random forest generate corresponding prediction results. in, This represents the prediction result generated by the Nth decision tree. v RF,N ,ω RF,N These are the longitudinal velocity and yaw rate of the target vehicle generated by the Nth decision tree, respectively.

[0019] Step S3-3: Calculate the average of the outputs of N decision trees to obtain the longitudinal velocity and yaw rate observations at time step k.

[0020] Preferably, step S4 specifically includes:

[0021] Step S4-1: Establish the state transition model of the extended Kalman filter. The state transition model calculates the state vector of the target vehicle at the current time step based on the nonlinear vehicle kinematics model to obtain the state vector of the target vehicle at the next time step.

[0022] Step S4-2: Establish a measurement model for the extended Kalman filter, wherein the measurement model uses the observed values ​​of longitudinal velocity and yaw rate as the measured values ​​of longitudinal velocity and yaw rate for the extended Kalman filter.

[0023] Step S4-3, obtaining a prior state estimation vector of a next time step by using the state transition model, and calculating a prior error covariance matrix;

[0024] Step S4-4, calculating a Kalman gain from the prior error covariance matrix, obtaining a state estimation vector of the next time step from the prior state estimation vector of the next time step, and obtaining an error covariance matrix of the next time step from the prior error covariance matrix of the next time step;

[0025] Step S4-5, returning to step S4-3 until the calculation of all time steps in the future time period is completed;

[0026] Step S4-6, taking the state estimation vectors of k+1 to k+m time steps in the future time period as a predicted trajectory Traj k+1,Kk+m .

[0027] Preferably, in the step S4-1, the state transition model is expressed as:

[0028] x k+1 =f(x k )+q k =[f1,f2,f3,f4,f5,f6]+q k

[0029] f1:

[0030] f2:

[0031] f3:θ k+1 =θ k +ω k Δt

[0032] f4:v k+1 =v k +a k Δt

[0033] f5:a k+1 =a k

[0034] f6:ω k+1 =ω k

[0035] wherein x k is a state vector of the target vehicle at the k time step, v k is a longitudinal speed of the target vehicle at the k time step, ω k is a yaw rate of the target vehicle at the k time step, f(·) is a vehicle state transition model based on a nonlinear vehicle kinematics model, q k is process noise, and has a mean of 0 and a variance of Qk a normal distribution, i.e. q k ~ N(0, Q k ), Q k = diag{0, 0, q dθ , q dv , q dω , q da}, q dθ , q dv , q dω , q da are the process noises of the heading angle, the longitudinal velocity, the yaw rate, and the longitudinal acceleration, respectively.

[0036] Preferably, the step S4-3 specifically comprises:

[0037] The state of the vehicle at the next time step is predicted using the state transition equation, and the prior error covariance matrix is calculated, expressed as:

[0038]

[0039] wherein represents the prior state estimation vector of the target vehicle at the k time step, represents the known posterior state estimation vector of the target vehicle at the k-1 time step, is the prior error covariance matrix of the target vehicle at the k time step, is the posterior error covariance matrix of the target vehicle at the k-1 time step, F k is the Jacobian matrix of the state transition equation at the k time step, is the transpose of F k , Q k-1 is the process noise covariance matrix at the k-1 time step.

[0040] Preferably, the step S4-4 specifically comprises:

[0041] The Kalman gain is calculated, and the state estimation vector and the error covariance matrix are updated, expressed as:

[0042]

[0043] wherein K k is the Kalman gain at the k time step, R k represents the measurement noise covariance matrix, R k = diag{σ v,k 2 , σ ω,k 2}, σ v,k 2 , σ ω,k 2 are the measurement noises of the heading angle, the longitudinal velocity, the yaw rate, and the longitudinal acceleration, respectively.k ,ω k the variance of the state measurement value, represents the state estimation vector of the target vehicle at the k time step, is the state measurement value, and the state prediction result of the random forest is used as the measurement value, that is, wherein, respectively represent the longitudinal speed and the yaw rate of the target vehicle at the k time step, is the posterior error covariance matrix at the k time step, and I is the unit matrix.

[0044] Compared with the prior art, the present application has at least the following beneficial effects:

[0045] (1) The present application uses high-precision map information and current motion state to generate multiple candidate reference paths and calculate the corresponding path curvature. These information provides rich input features for the random forest, making the model better adapt to changes in complex environments. In this way, when the vehicle drives in complex terrains such as mine areas, the prediction model can be adjusted in real time to make more accurate predictions for different road conditions.

[0046] (2) The present application combines the advantages of random forest and extended Kalman filter to significantly improve the accuracy and stability of vehicle trajectory prediction. The random forest model can effectively identify the driving intention of the vehicle, learn and analyze the complex relationship between vehicle speed, direction and road curvature, thereby improving the prediction accuracy of future driving speed and turning rate, and better adapting to trajectory prediction in complex scenes such as mine areas.

[0047] (3) The present application effectively reduces the accumulation of errors by combining random forest and extended Kalman filter, ensuring the stability of the prediction results. By establishing state transition model and measurement model, extended Kalman filter can effectively estimate and update the state when dealing with nonlinear systems. The prediction result not only has high precision, but also ensures the stability of the prediction result. BRIEF DESCRIPTION OF DRAWINGS

[0048] The accompanying drawings are for the purpose of illustrating specific embodiments and are not to be considered limiting of the application.

[0049] Figure 1 The flowchart of the method for realizing open-pit mine vehicle trajectory prediction by combining random forest and Kalman provided by the present application.

[0050] Figure 2 The schematic diagram for obtaining the input information of the candidate reference path of the target vehicle and the random forest model provided by the present application.

[0051] Figure 3 The result graph of long-time trajectory prediction of the target vehicle on the specific mine road provided by the present application. Detailed Implementation

[0052] To better understand the above-described objectives, features, and advantages of the present invention, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments of the present invention and the features thereof can be combined with each other. Furthermore, the present invention can be implemented in other ways different from those described herein; therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0053] This invention provides a method for predicting the trajectory of vehicles in open-pit mines by combining random forest and Kalman filtering. It primarily relies on path information of reference lane lines in a prior high-precision map of the open-pit mine area. Based on the random forest method, it comprehensively considers the vehicle's driving state and the curvature of the reference path to predict the vehicle's future speed and yaw rate, achieving vehicle state estimation. Then, it uses the state estimation results combined with extended Kalman filtering for trajectory prediction, thereby obtaining the long-term trajectory prediction result of the target vehicle.

[0054] To illustrate the effectiveness of the method proposed in this invention, the following detailed description of the above technical solution is provided through a specific embodiment, such as... Figure 1 As shown, a method for predicting the trajectory of open-pit mine vehicles by combining random forest and Kalman scattering is disclosed. The specific implementation steps are as follows:

[0055] Step S1: The current motion status information of the mining vehicles is obtained by the sensing device. The motion status information includes the position, heading angle, longitudinal speed, longitudinal acceleration and yaw rate of the mining vehicles.

[0056] The target vehicle is the mining vehicle whose trajectory is predicted. The current motion state information of the target vehicle can be acquired through various sensing devices. Specifically, it can be acquired through the perception module equipped on the autonomous vehicle itself, collected by perception units deployed on the roadside, or obtained from other vehicles via vehicle-to-vehicle (V2V) communication. In some embodiments, information from these multiple sources can be fused to obtain more accurate and reliable target vehicle motion state information.

[0057] The motion state information of the target vehicle at time t is represented as: x t =[x t ,y t ,θ t ,v t ,a t ,ω t ] T , where x t ,y t ,θt ,v t ,a t ,ω t These represent the eastward and northward position coordinates, heading angle, longitudinal velocity, longitudinal acceleration, and yaw rate at time t, respectively.

[0058] By acquiring and updating the above status information in real time, the movement status of the target vehicle on the mining area roads can be accurately grasped, providing reliable data support for subsequent trajectory prediction.

[0059] Step S2: Based on the current motion state information and the prior high-precision map information of the open-pit mine, obtain candidate reference paths and obtain the path curvature of the candidate reference paths. The candidate reference paths are the paths that mining vehicles may travel in the future.

[0060] like Figure 2 As shown, based on the real-time status information of the target vehicle and the prior high-precision map information of the open-pit mine, a set of possible candidate reference paths SET(Path) is obtained for the target vehicle. i ). Among them, Path i Let i represent the i-th path, where each path is a feasible driving trajectory.

[0061] In some embodiments, the candidate reference path set can be obtained as follows: based on the current location of the target vehicle, location matching is performed in a high-precision map to obtain the current road segment where the vehicle is located. Then, starting from the current road segment, a search is conducted in the road network topology of the high-precision map, and all drivable paths obtained are used as candidate paths. Furthermore, the candidate paths can be initially screened by combining the vehicle's current heading angle, speed, and other state information to eliminate paths that obviously conflict with the current vehicle's motion state, ultimately obtaining a reasonable candidate reference path set.

[0062] Let the current time be t. Based on the assumption that the target vehicle moves at a constant speed, obtain the target vehicle's position at each time step t+kΔt within the next 5 seconds, where Δt is the time step size, k is the index of the time step, Δt = 0.1s, k = 1, 2, ..., 50. This yields the target vehicle's position coordinates (px) at each time step within the next 5 seconds. k ,py k ). The k-time step position coordinates (px) k ,py k The aiming point is determined as k time steps. coordinates, where Let K be the coordinates of the aiming point at time step k.

[0063] For each pre-aiming point, based on the position of the pre-aiming point and the candidate reference path set SET(Path) iThe path in the path matching step S2 is matched with the reference path to obtain a reference path closest to the position of the preview point, and the path curvature K of the k-time-step preview point is determined according to the reference path closest to the position of the preview point. k .

[0064] In some embodiments, the path curvature can be obtained from the high-precision mine area map according to the reference path.

[0065] In step S3, the current motion state information and the path curvature are processed based on the random forest to obtain observation values of the longitudinal speed and the yaw rate of the mine vehicle.

[0066] The random forest of the present application is composed of a plurality of trained decision trees, N is the total number of decision trees, and the trained decision trees can obtain predicted longitudinal speed and yaw rate based on input target vehicle information.

[0067] Based on the target vehicle motion state and the path curvature, the input quantity of the random forest at the k-time-step is established The current vehicle state, time step and curvature information of the reference path are integrated.

[0068] The N decision trees of the random forest generate corresponding prediction results Among them, denotes the prediction result generated by the Nth decision tree, v RF,N ,ω RF,N are the longitudinal speed and yaw rate of the target vehicle generated by the Nth decision tree, respectively. Finally, the output result of the random forest is the average value of the output results of the N decision trees, denoted as variance

[0069]

[0070] By inputting the input quantity of the random forest at the k-time-step into the random forest, the longitudinal speed and the yaw rate at the k-time-step are obtained as the observation values at the k-time-step The variance of the random forest at the k-time-step is taken as the observation noise R k = diag{σ v,k 2 ,σ ω,k 2}, wherein σ v,k 2 ,σ ω,k 2 are the variances of , and diag{·} denotes a diagonal matrix.

[0071] Step S4, a state transition model and a measurement model of the extended Kalman filter are established, and the observed values of the longitudinal velocity and the yaw angle velocity are taken as the measurement values of the longitudinal velocity and the yaw angle velocity of the extended Kalman filter;

[0072] The state prediction and measurement update of the extended Kalman filter are performed, and finally the predicted trajectory of the mine vehicle at the future time is obtained.

[0073] In this step, the driving speed and the yaw angle velocity of the target vehicle at the future time step obtained by the state estimation based on the random forest method are taken as the measurement values, the other state quantities are updated by the extended Kalman filter, the estimated values of the other state quantities of the target vehicle in the prediction time domain are obtained, and then the long-time prediction trajectory of the target vehicle is obtained. The specific steps include:

[0074] (1) Constructing a state transition model

[0075] The state transition model expression of the extended Kalman filter is:

[0076] x k+1 =f(x k )+q k =[f1,f2,f3,f4,f5,f6]+q k

[0077] Wherein, x k is the state vector of the target vehicle at the k time step, v k is the longitudinal velocity of the target vehicle at the k time step, ω k is the yaw angle velocity of the target vehicle at the k time step, f(·) is the vehicle state transition model based on the nonlinear vehicle kinematics model, q k is the process noise, which is subject to normal distribution with mean value 0 and variance Q k , i.e. q k ~N(0,Q k ), Q k =diag{0,0,q dθ ,q dv ,q dω ,q da}, q dθ , q dv , q dγ , q da are the process noises of the heading angle, the longitudinal velocity, the yaw angle velocity and the longitudinal acceleration respectively.

[0078] The vehicle state transition equation f(x k ) based on the nonlinear vehicle kinematics model is:

[0079] f1:

[0080] f2:

[0081] f3:θ k+1 = θ k + ω k Δt

[0082] f4: v k+1 = v k + a k Δt

[0083] f5: a k+1 = a k

[0084] f6: ω k+1 = ω k

[0085] (2) Constructing measurement model

[0086] The measurement model expression of extended Kalman filter is:

[0087] z k = Hx k + γ k

[0088] wherein, is the measurement matrix, γk is the measurement noise, γ k ~ N(0, R k ) indicates that the measurement noise satisfies the normal distribution with variance σ v,k 2 ,σ ω,k 2 R k = diag{σ v,k 2 ,σ ω,k 2}.

[0089] (3) State prediction process

[0090] In the state prediction process, the state transition equation is used to predict the state of the vehicle at the next time step, and the prior error covariance matrix is calculated, the expression is:

[0091]

[0092] wherein, represents the prior state estimation vector of the target vehicle at k time step, represents the known posterior state estimation vector of the target vehicle at k-1 time step, is the prior error covariance matrix of the target vehicle at k time step, is the posterior error covariance matrix of the target vehicle at k-1 time step, Fk The Jacobian matrix of the state transition equation at time step k is obtained by solving the first-order partial differential of the state transition equation with respect to the state vector of the target vehicle. For F k The transpose of Q k-1 Let be the process noise covariance matrix at time step k-1.

[0093] (4) Measurement update process

[0094] During the measurement update process, the Kalman gain is calculated, and the state estimation vector and error covariance matrix are updated, as expressed by:

[0095]

[0096] Among them, K k R is the Kalman gain at time step k. k R represents the measurement noise covariance matrix. k =diag{σ v,k 2 ,σ ω,k 2}, σ v,k 2 ,σ ω,k 2 v k ,ω k variance This represents the state estimation vector of the target vehicle at time step k. The state measurement is the result of the random forest's state prediction. in, Let represent the longitudinal velocity and yaw rate of the target vehicle at time step k, respectively. Let I be the posterior error covariance matrix at time step k, and let I be the identity matrix.

[0097] By continuously iterating through the state prediction and measurement update processes, the predicted trajectory Traj formed by the motion state of the target vehicle at time steps k+1 to k+m can be obtained. k+1,Kk+m .

[0098] like Figure 3 The figure shown is a result of the present invention for long-term trajectory prediction of target vehicles on roads in a specific mining area.

[0099] Through the method described in the embodiment, the vehicle trajectory prediction in the open-pit mine area is realized, the advantages of the random forest and the extended Kalman filter are combined, compared with the vehicle trajectory prediction method based on the physical kinematic model, the driving state of the vehicle and the prior map information are comprehensively considered, the driving intention is recognized through the random forest, and the relationship between the vehicle speed, the direction and the road curvature is learned, the prediction accuracy of the future driving speed and the turning rate of the vehicle is effectively improved, and the precision, stability and adaptability of the vehicle trajectory prediction are significantly improved. Meanwhile, the combination of the random forest and the extended Kalman filter reduces the error accumulation, ensures the stability of the prediction result, and provides strong support for the unmanned trajectory prediction technology in the complex environment such as the mine area.

[0100] The detailed description of the application, although describing each action or step in a specific order, should be understood as requiring such actions or steps to be performed in the specific order shown or in a sequential order, or requiring all illustrated actions or steps to be performed to achieve the desired results. In certain circumstances, multi-tasking and parallel processing can be advantageous. Similarly, although the above discussion contains many specific implementation details, these should not be interpreted as limiting the scope of the disclosure. Certain features described in the context of separate embodiments can also be combined in a single implementation. Conversely, various features described in the context of a single implementation can also be implemented separately or in any suitable subcombination. The above description is merely the preferred specific implementation of the application, but the scope of protection of the application is not limited thereto, and any changes or replacements within the scope of the technology disclosed by the application can be easily thought of by those skilled in the art, and should be covered within the scope of protection of the application.

[0101] The above description is merely the preferred specific implementation of the application, but the scope of protection of the application is not limited thereto, and any changes or replacements within the scope of the technology disclosed by the application can be easily thought of by those skilled in the art, and should be covered within the scope of protection of the application.

Claims

1. A method for predicting the trajectory of open-pit mine vehicles by combining random forest and Kalman scattering, characterized in that, Includes the following steps: Step S1: The current motion status information of the mining vehicles is obtained by the sensing device. The motion status information includes the position, heading angle, longitudinal speed, longitudinal acceleration and yaw rate of the mining vehicles. Step S2: Based on the current motion state information and the prior high-precision map information of the open-pit mine, obtain candidate reference paths and obtain the path curvature of the candidate reference paths. The candidate reference paths are the paths that mining vehicles may travel in the future. Step S3: Process the current motion state information and the path curvature based on random forest to obtain the observed values ​​of the longitudinal speed and yaw rate of the mining vehicle. Step S4: Establish the state transition model and measurement model of the extended Kalman filter, and use the observed values ​​of the longitudinal velocity and yaw rate as the measured values ​​of the longitudinal velocity and yaw rate of the extended Kalman filter. Extended Kalman filtering is used for state prediction and measurement updates to ultimately obtain the predicted trajectories of mining vehicles in the future.

2. The method for predicting open-pit mine vehicle trajectories by combining random forest and Kalman mechanics according to claim 1, characterized in that, In step S1, the motion state information of the target vehicle at the current time t is represented as: x t =[x t ,y t ,θ t ,v t ,a t ,ω t ] T , where x t ,y t ,θ t ,v t ,a t ,ω t These represent the eastward and northward position coordinates, heading angle, longitudinal velocity, longitudinal acceleration, and yaw rate at time t, respectively.

3. The method for predicting open-pit mine vehicle trajectories by combining random forest and Kalman mechanics according to claim 2, characterized in that, Step S2 specifically includes: Step S2-1: Based on the current motion state information and the prior high-precision map information of the open-pit mine, obtain the set of possible candidate reference paths SET(Path) for the target vehicle. i ), where Path i This represents the i-th path; Step S2-2: Divide the preset future time period into multiple time steps, denoted as t+kΔt, where t is the current time, Δt is the time step size, and k is the index of the time step; based on the assumption that the target vehicle moves at a constant speed, obtain the target vehicle position coordinates (px) for each time step. k ,py k ), as the pre-aiming point for the corresponding k time steps. in Let K be the coordinates of the preview point at time step k; Step S2-3: Based on the position coordinates of the pre-aiming point and the candidate reference path set SET(Path) i The path in the target area is matched to obtain the reference path closest to the target point. The path curvature κ of the target point at time step k is determined by the reference path closest to the target point. k .

4. The method for predicting open-pit mine vehicle trajectories by combining random forest and Kalman mechanics according to claim 3, characterized in that, Step S3 specifically includes: Step S3-1: Based on the target vehicle's motion state and the path curvature, establish the input of the random forest at time k. Step S3-2: Input the data into the random forest for processing. The N trained decision trees of the random forest generate corresponding prediction results. in, This represents the prediction result generated by the Nth decision tree. v RF,N ,ω RF,N These are the longitudinal velocity and yaw rate of the target vehicle generated by the Nth decision tree, respectively. Step S3-3: Calculate the average of the outputs of N decision trees to obtain the longitudinal velocity and yaw rate observations at time step k.

5. The method for predicting open-pit mine vehicle trajectories by combining random forest and Kalman mechanics according to claim 4, characterized in that, Step S4 specifically includes: Step S4-1: Establish the state transition model of the extended Kalman filter. The state transition model calculates the state vector of the target vehicle at the current time step based on the nonlinear vehicle kinematics model to obtain the state vector of the target vehicle at the next time step. Step S4-2: Establish a measurement model for the extended Kalman filter, wherein the measurement model uses the observed values ​​of longitudinal velocity and yaw rate as the measured values ​​of longitudinal velocity and yaw rate for the extended Kalman filter. Step S4-3: Use the state transition model to obtain the prior state estimation vector for the next time step, and calculate the prior error covariance matrix; Step S4-4: Calculate the Kalman gain from the prior error covariance matrix; obtain the state estimation vector for the next time step from the prior state estimation vector for the next time step; obtain the error covariance matrix for the next time step from the prior error covariance matrix for the next time step. Step S4-5: Return to step S4-3 until the calculation of all time steps in the future time period is completed; Step S4-6: Use the state estimate from time step k+1 to k+m in the future time period as the predicted trajectory Traj. k +1,Kk+m .

6. The method for predicting open-pit mine vehicle trajectories by combining random forest and Kalman mechanics according to claim 5, characterized in that, In step S4-1, the state transition model expression is: x k+1 =f(x k )+q k =[f1,f2,f3,f4,f5,f6]+q k f3:θ k+1 =θ k +oh k Δt f4:v k+1 =v k +a k Δt f5: a k+1 =a k f6:oh k+1 =ω k Where, x k v is the state vector of the target vehicle at time step k. k Let ω be the longitudinal velocity of the target vehicle at time step k. k Let f(·) be the yaw rate of the target vehicle at time step k, and let f(·) be the vehicle state transition model based on the nonlinear vehicle kinematics model. k It is process noise, with a mean of 0 and a variance of Q. k The normal distribution, i.e., q k ~N(0,Q) k ), Q k =diag{0,0,q dθ ,q dv ,q dω ,q da }, q dθ ,q dv ,q dω ,q da These are the process noises for heading angle, longitudinal velocity, yaw rate, and longitudinal acceleration, respectively.

7. The method for predicting open-pit mine vehicle trajectories by combining random forest and Kalman mechanics according to claim 6, characterized in that, Step S4-3 specifically includes: The state transition equation is used to predict the vehicle's state at the next time step, and the prior error covariance matrix is ​​calculated as follows: in, This represents the prior state estimation vector of the target vehicle at time step k. This represents the posterior state estimate vector of the target vehicle at time step k-1, which is known. Let be the prior error covariance matrix of the target vehicle at time step k. Let F be the posterior error covariance matrix of the target vehicle at time step k-1. k Let be the Jacobian matrix of the state transition equation at time step k. For F k The transpose of Q k-1 Let be the process noise covariance matrix at time step k-1.

8. The method for predicting open-pit mine vehicle trajectories by combining random forest and Kalman mechanics according to claim 7, characterized in that, Step S4-4 specifically includes: Calculate the Kalman gain and update the state estimation vector and error covariance matrix, as expressed by: Among them, K k R is the Kalman gain at time step k. k R represents the measurement noise covariance matrix. k =diag{σ v,k 2 ,σ ω,k 2 }, σ v,k 2 ,σ ω,k 2 v k ,ω k variance This represents the state estimation vector of the target vehicle at time step k. The state measurement is the result of the random forest's state prediction. in, Let represent the longitudinal velocity and yaw rate of the target vehicle at time step k, respectively. Let I be the posterior error covariance matrix at time step k, and let I be the identity matrix.

Images