CN115123217A

CN115123217A - Mining area obstacle vehicle driving track generation method and device and computer equipment

Info

Publication number: CN115123217A
Application number: CN202211068215.8A
Authority: CN
Inventors: 魏辉; 张宏韬; 高�玉
Original assignee: Qingdao Vehicle Intelligence Pioneers Inc
Current assignee: Qingdao Vehicle Intelligence Pioneers Inc
Priority date: 2022-09-02
Filing date: 2022-09-02
Publication date: 2022-09-30
Anticipated expiration: 2042-09-02
Also published as: CN115123217B

Abstract

The invention discloses a method and a device for generating a driving track of a mine obstacle vehicle and computer equipment, relates to the technical field of automation, and mainly aims to improve the generation precision of the driving track of the obstacle vehicle. The method comprises the following steps: obtaining obstacle vehicles in a mine area within a preset distance range from an unmanned vehicle and driving parameter information corresponding to the obstacle vehicles; inputting the running parameter information into a preset running direction prediction model for direction prediction to obtain a first predicted running direction corresponding to the obstacle vehicle; and generating a driving track corresponding to the obstacle vehicle by adopting a corresponding track generation model according to the first predicted driving direction.

Description

Mine obstacle vehicle driving track generation method and device and computer equipment

Technical Field

The application relates to the technical field of automation, in particular to a method and a device for generating a driving track of a mine area obstacle vehicle and computer equipment.

Background

Compared with a common highway, most of the roads in the mining area are non-public roads, information such as road boundaries, lane lines and traffic indicator lamps is not available, the road conditions in the mining area are complex, various uphill slopes, downhill slopes and curves exist, when the curves go uphill slopes, downhill crossings and intersections, collision risks between unmanned vehicles and obstacle vehicles such as manned trucks and auxiliary vehicles can occur, and in order to avoid collision, the driving tracks of the obstacle vehicles need to be predicted.

Conventionally, a travel locus of an obstacle vehicle is generally predicted based on a set position coordinate function of the obstacle vehicle with respect to time. However, since the mine area has no road boundary and has various curves, the driving direction of the obstacle vehicle may change at any time, and the movement track of the obstacle vehicle is also complex.

Disclosure of Invention

The invention provides a method and a device for generating a driving track of an obstacle vehicle in a mining area and computer equipment, and mainly aims to improve the generation precision of the driving track of the obstacle vehicle.

According to a first aspect of the present invention, there is provided a driving track generation method for a mine obstacle vehicle, comprising:

obtaining obstacle vehicles in a mine area within a preset distance range from an unmanned vehicle and driving parameter information corresponding to the obstacle vehicles;

inputting the running parameter information into a preset running direction prediction model for direction prediction to obtain a first predicted running direction corresponding to the obstacle vehicle;

and generating a driving track corresponding to the obstacle vehicle by adopting a corresponding track generation model according to the first predicted driving direction.

According to a second aspect of the present invention, there is provided a driving route generating device for a mine obstacle vehicle, comprising:

the system comprises an acquisition unit, a display unit and a control unit, wherein the acquisition unit is used for acquiring obstacle vehicles in a mine area within a preset distance range from an unmanned vehicle and driving parameter information corresponding to the obstacle vehicles;

the prediction unit is used for inputting the running parameter information into a preset running direction prediction model to carry out direction prediction so as to obtain a first predicted running direction corresponding to the obstacle vehicle;

and the generating unit is used for generating a driving track corresponding to the obstacle vehicle by adopting a corresponding track generation model according to the first predicted driving direction.

According to a third aspect of the present invention, there is provided a computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of:

According to a fourth aspect of the present invention, there is provided a computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the following steps when executing the program:

Compared with the current mode of directly utilizing a position coordinate function related to time to generate the driving track of the obstacle vehicle, the driving track generation method, the driving track generation device and the driving track generation computer equipment for the obstacle vehicle in the mining area can acquire the obstacle vehicle in the predetermined distance range with the unmanned vehicle in the mining area and the driving parameter information corresponding to the obstacle vehicle; inputting the running parameter information into a preset running direction prediction model for direction prediction to obtain a first predicted running direction corresponding to the obstacle vehicle; and finally, generating a driving track corresponding to the obstacle vehicle by adopting a corresponding track generation model according to the first predicted driving direction. Because the driving tracks of the obstacle vehicle in different directions are different, the driving direction of the obstacle vehicle is predicted by using the preset driving direction prediction model, and on the basis, the driving track meeting the vehicle kinematic constraint is generated by using the track generation model, so that the generation precision of the driving track of the obstacle vehicle can be ensured, the subsequent accurate judgment of whether the unmanned vehicle has collision risk with the obstacle vehicle is facilitated, and the unmanned vehicle makes driving behavior decision to avoid the risk.

Drawings

The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the invention without limiting the invention. In the drawings:

FIG. 1 is a flow chart of a method for generating a driving track of a mine obstacle vehicle according to an embodiment of the invention;

FIG. 2 is a flow chart of another method for generating a driving track of a mine obstacle vehicle according to an embodiment of the present invention;

FIG. 3 is a schematic view of a driving direction of an obstacle vehicle according to an embodiment of the present invention;

FIG. 4 is a schematic view of another obstacle vehicle driving direction provided by an embodiment of the invention;

fig. 5 is a schematic structural diagram illustrating a driving track generating device for a mine obstacle vehicle according to an embodiment of the present invention;

fig. 6 is a schematic structural diagram illustrating another driving track generating device for mine obstacle vehicles according to an embodiment of the present invention;

FIG. 7 is a block diagram illustrating an example of a physical structure of a computer device according to an embodiment of the present invention;

FIG. 8 is a schematic diagram illustrating a chip according to an embodiment of the present invention;

fig. 9 is a schematic structural diagram of a terminal according to an embodiment of the present invention.

Detailed Description

The invention will be described in detail hereinafter with reference to the accompanying drawings in conjunction with embodiments. It should be noted that, in the present application, the embodiments and features of the embodiments may be combined with each other without conflict.

At present, as a mining area has no road boundary and various curves, the driving direction of the obstacle vehicle may change at any time, the motion track of the obstacle vehicle is also relatively complex, and the driving track of the obstacle vehicle cannot be effectively generated by the simple position coordinate function related to time in the prior art.

In order to solve the above problem, an embodiment of the present invention provides a driving track generation method for a mine obstacle vehicle, as shown in fig. 1, the method including:

101. the method comprises the steps of obtaining obstacle vehicles in a mine area within a preset distance range from unmanned vehicles and driving parameter information corresponding to the obstacle vehicles.

Wherein, unmanned vehicle includes mining haulage vehicle, and this mining haulage vehicle specifically includes: mine trucks, wide cars, articulated mine cars, and the like. The obstacle vehicle is a manned vehicle, the manned vehicle comprises a manned truck and an auxiliary vehicle, and the auxiliary vehicle specifically comprises a sprinkler, a command vehicle, an engineering emergency vehicle and the like. The preset distance range can be set according to the road condition of the mining area and the actual running condition of the unmanned vehicle, for example, barrier vehicles within the ranges of 120 meters, 150 meters, 200 meters and the like from the unmanned vehicle are set and obtained. The preset distance is not further limited, and can be set according to the requirements of application scenes in practical application. Further, the driving parameter information includes the current speed, acceleration, position information, lateral distance from the center line of the road, heading angle, angular velocity, distance from the unmanned vehicle, and the like of the obstacle vehicle.

The embodiment of the invention is mainly applied to the scene of generating the driving track of the obstacle vehicle in the mining area. The execution main body of the embodiment of the invention is a device or equipment capable of generating the driving track of the obstacle vehicle, and can be specifically arranged on the side of a server or the side of a vehicle end.

For the embodiment of the invention, when detecting the obstacle vehicle corresponding to the unmanned vehicle, the sensing device mounted on the unmanned vehicle can detect the obstacle vehicle within the predetermined distance range, and in addition, the cooperative positioning device mounted on other vehicles and the detection device mounted on two sides of the road can also assist the sensing device on the unmanned vehicle in detecting the obstacle vehicle, so that the obstacle vehicle which is within the predetermined distance range and in front of the unmanned vehicle can be obtained by the three detection modes. It should be noted that the obstacle vehicles in the implementation of the present invention are vehicles that are traveling dynamically, the number of obstacle vehicles corresponding to the finally detected unmanned vehicle may be one, two, or more, and when two or more detected obstacle vehicles are detected, it is necessary to predict the traveling direction and generate a trajectory for each obstacle vehicle.

Furthermore, when detecting an obstacle vehicle corresponding to the unmanned vehicle, it is also necessary to obtain driving parameter information corresponding to the obstacle vehicle, where the speed, position information, lateral distance from the center line of the road, and course angle corresponding to the obstacle vehicle in the driving parameter information are detected by the sensing device on the unmanned vehicle, the cooperative positioning device on other vehicles, and the detection devices installed on both sides of the road, and when the unmanned vehicle is far away from the obstacle vehicle and cannot sense the driving parameter information of the obstacle vehicle, the protocol positioning device may send the driving parameter information of the obstacle vehicle to the unmanned vehicle through cloud communication, and when the road conditions are complicated and the cooperative positioning device cannot detect the parameter information, the detection devices installed on both sides of the road may perform auxiliary detection, thereby obtaining the current speed, and course angle of the obstacle vehicle through mutual supplement of three detection modes, Location information, lateral distance from the road centerline, and heading angle. Further, the current acceleration and angular velocity of the obstacle vehicle can be calculated through the detected speed and heading angle, and the distance between the obstacle vehicle and the unmanned vehicle can be calculated according to the current position information of the unmanned vehicle and the current position information of the obstacle vehicle.

102. And inputting the running parameter information into a preset running direction prediction model for direction prediction to obtain a first predicted running direction corresponding to the obstacle vehicle.

The preset driving direction prediction model may be a preset neural network model, the preset neural network model is specifically mlp (Multi-Layer probability) neural network model, the first predicted driving direction is specifically the next action state of the obstacle vehicle, and the first predicted driving direction includes a left turn, a right turn and a straight line.

For the embodiment of the present invention, in order to predict the first predicted traveling direction corresponding to the obstacle vehicle, step 102 specifically includes: inputting the driving parameter information into the preset neural network model for direction prediction to obtain probability values of driving of the obstacle vehicle in all directions; and screening out a maximum probability value from the probability values of the obstacle vehicle in the traveling directions, and determining the direction corresponding to the maximum probability value as a first predicted traveling direction of the obstacle vehicle. On this basis, the input of the driving parameter information into the preset neural network model for direction prediction to obtain probability values of driving of the obstacle vehicle in all directions includes: normalizing the driving parameter information to obtain processed driving parameter information; inputting the processed driving parameter information to the first hidden layer through the input layer for feature extraction to obtain a first driving parameter feature corresponding to the obstacle vehicle; inputting the first driving parameter characteristics to the second hidden layer for characteristic extraction to obtain second driving parameter characteristics corresponding to the obstacle vehicle; and inputting the second driving parameter characteristics to the output layer for classification to obtain probability values of driving of the obstacle vehicle in all directions.

Specifically, because the number of layers of the neural network model is too large, which results in too large training calculation amount, and the number of layers is too small, which affects the prediction accuracy of the model, a mlp neural network model adopted in the embodiment of the present invention has 4 layers, which includes an input layer, a first hidden layer, a second hidden layer, and an output layer, where different layers are all connected, and because the information dimension of the driving parameter information is 7, the number of neurons in the input layer is set to 7, the number of neurons in the first hidden layer is set to 7, the number of neurons in the second hidden layer is set to 7, and because the classification result of the output layer includes a straight line, a left turn, and a right turn, the number of neurons in the output layer is set to 3. The specific construction process of the model is detailed in step 202.

Further, when the driving direction of the obstacle vehicle is predicted by using the preset neural network model, for different driving parameter information, because the magnitude of the numerical value may have great difference, normalization processing needs to be performed first, and the numerical value is mapped into the range of [0,1], and the specific formula is as follows:

（1-1）

wherein, x is any one of the running parameter information of speed, acceleration, position information, transverse distance from the central line of the road, course angle, angular speed and distance with the unmanned vehicle, x' is any one of the running parameter information after normalization processing, and MIN and MAX are the minimum value and the maximum value corresponding to any one of the running parameter information.

Further, after the driving parameter information after the normalization processing is obtained, the processed driving parameter information is input to the hidden layers (the first hidden layer and the second hidden layer) through the input layer for feature extraction, and the second driving parameter features output by the second hidden layer are input to the output layer for classification, so that the probability value of the straight driving of the obstacle vehicle, the probability value of the left turn and the probability value of the right turn are obtained. And finally, determining the driving direction corresponding to the maximum probability value as a first predicted driving direction corresponding to the obstacle vehicle.

It should be noted that, when the driving direction of the obstacle vehicle is predicted, the road environment of the mining area needs to be comprehensively considered, so as to reduce unnecessary calculation. As shown in fig. 3, the vehicle indicated by the dashed square box is an unmanned vehicle, the vehicle indicated by the solid square box is a manned vehicle, that is, an obstacle vehicle, and the unmanned vehicle travels along a preset track path, and at this time, the lower obstacle vehicle may only turn left or go straight, and no right turn is possible, so after obtaining the probability values of the obstacle vehicle in various directions, it is only necessary to compare the probability value of the obstacle vehicle turning left with the probability value of the obstacle vehicle going straight, and then determine whether the obstacle vehicle is going straight or turning left.

103. And generating a driving track corresponding to the obstacle vehicle by adopting a corresponding track generation model according to the first predicted driving direction.

The track generation model comprises a vehicle kinematics model and a sampling track planner. For the embodiment of the present invention, when the first predicted traveling direction is the straight traveling of the obstacle vehicle, a vehicle kinematic model is used to generate a traveling track corresponding to the obstacle vehicle. As an optional implementation manner, the specific process of generating the corresponding driving track of the obstacle vehicle by using the vehicle kinematics model includes: sampling any point on a straight line along the direction of the target course angle according to the target course angle of the obstacle vehicle at the current position to obtain an abscissa and an ordinate corresponding to any sampling point; determining a driving straight line between the current position of the obstacle vehicle and the arbitrary sampling point according to the abscissa and the ordinate corresponding to the current position of the obstacle vehicle and the abscissa and the ordinate corresponding to the arbitrary sampling point; and determining the driving straight line as a driving track corresponding to the obstacle vehicle.

Fig. 4 shows a typical intersection where the vehicle indicated by the dashed square box is an unmanned vehicle, the vehicle indicated by the solid square box is a manned vehicle, i.e., an obstacle vehicle, and if the first predicted driving direction is the straight-ahead movement of the obstacle vehicle, the vehicle kinematics model is used to generate the driving track of the obstacle vehicle. Specifically, since two points can determine a straight line track, any point on a straight line along the target course angle direction is sampled, and according to the abscissa and the ordinate corresponding to the any sampling point and the abscissa and the ordinate corresponding to the current position of the obstacle vehicle, a straight driving line including the current position and any sampling point is determined, and the straight driving line is determined as the driving track of the obstacle vehicle.

Further, when the first predicted driving direction is a left turn or a right turn of the obstacle vehicle, a sampling trajectory planner is used to generate a driving trajectory corresponding to the obstacle vehicle, and for a specific process of generating the driving trajectory by using the sampling trajectory planner, see step 203.

According to the method for generating the driving track of the obstacle vehicle in the mining area, the driving direction of the obstacle vehicle is predicted by using the preset driving direction prediction model, and on the basis, the driving track meeting the vehicle kinematic constraint is generated by using the track generation model, so that the generation precision of the driving track of the obstacle vehicle can be ensured, the subsequent accurate judgment of whether the unmanned vehicle has collision risk with the obstacle vehicle is facilitated, and the unmanned vehicle can make driving behavior decision to avoid the risk.

Further, in order to better explain the process of generating the driving track of the obstacle vehicle, as a refinement and an extension of the above embodiment, an embodiment of the invention provides another method for generating the driving track of the obstacle vehicle in a mine, as shown in fig. 2, the method includes:

201. obtaining obstacle vehicles in a mine area within a preset distance range from an unmanned vehicle and driving parameter information corresponding to the obstacle vehicles.

For the embodiment of the present invention, the specific process of obtaining the obstacle vehicles and the corresponding driving parameter information in the mining area is the same as that in step 101, and is not described herein again.

202. And inputting the running parameter information into a preset running direction prediction model for direction prediction to obtain a first predicted running direction corresponding to the obstacle vehicle.

For the embodiment of the present invention, before predicting the driving direction of the obstacle vehicle, a preset driving direction prediction model needs to be constructed, and as an optional implementation manner, for a specific construction process of the preset driving direction prediction model, the method includes: acquiring historical driving parameter information and historical driving directions corresponding to the obstacle vehicle; constructing an initial driving direction prediction model by using a preset neural network algorithm;

inputting the historical driving parameter information into the initial driving direction prediction model for direction prediction to obtain a second predicted driving direction corresponding to the obstacle vehicle; constructing a running direction loss function corresponding to the obstacle vehicle based on a second predicted running direction and a historical running direction corresponding to the obstacle vehicle; continuously performing iterative updating on initial weight parameter values in the initial driving direction prediction model on the basis of the driving direction loss function, and outputting weight parameter values of corresponding iterative levels when the driving direction loss function is minimum; and determining the preset driving direction prediction model based on the weight parameter values of the corresponding iteration levels.

Further, the continuously iteratively updating the initial weight parameter values in the initial driving direction prediction model based on the driving direction loss function, and outputting the weight parameter values of the corresponding iteration level when the driving direction loss function is minimum includes: calculating the partial derivative of the weight parameter in the running direction loss function to obtain a partial derivative function related to the weight parameter; updating the initial weight parameter value based on the partial derivative function to obtain the weight parameter value after the iteration update in the current round; and repeating the process of iteratively updating the initial weight parameter values until the weight parameter values of the corresponding iteration levels are output when the driving direction loss function is minimum.

Wherein, the full connection form between different layers in the mlp neural network model can be expressed as:

（1-2）

wherein x is ₁ …x _n For the information or features entered, w ₁ …w _n As a weight parameter, θ _m Is a bias parameter. In addition, the present inventionIn the implementation, a sigmoid activation function is adopted, and the specific formula of the activation function is as follows:

（1-3）

specifically, when a preset driving direction prediction model is constructed, a training sample set is firstly obtained, the training sample set comprises historical driving parameter information and historical driving directions corresponding to obstacle vehicles, and then the historical driving parameter information is subjected to normalization processing by using a formula (1-1), so that the normalized historical driving parameter information is obtained. Then, a preset neural network algorithm is utilized to construct an initial driving direction prediction model, and specifically, the initial driving direction prediction model comprises an input layer i _i First hidden layer h1 _i Second hidden layer h2 _i And an output layer o _i Wherein the input layer i _i Comprising 7 neurons, a first hidden layer h1 _i Comprising 7 neurons, a second hidden layer h2 _i Comprising 7 neurons, output layer o _i The method comprises 3 neurons, wherein the number n =7 × 7+7 × 3=119 of weight parameters in an initial driving direction prediction model is determined according to the number of each layer of neurons, and a bias parameter theta is further included in the initial driving direction prediction model _m M =1,2,3, and initial values are given after setting the weight parameter and the bias parameter, respectively.

Further, after the initial traveling direction prediction model is constructed, the normalized historical traveling parameter information (historical speed, historical acceleration, historical position information, historical lateral distance from the center line of the road, historical heading angle, historical angular velocity, distance from the unmanned vehicle, and the like) is input into the initial behavior direction prediction model for direction prediction, specifically, through the input layer i _i Inputting the historical driving parameter information to the first hidden layer h1 _i Calculating by adopting a formula (1-2), and substituting the calculation result of the first hidden layer into a formula (1-3) to obtain f (h 1) _i ) Then f (h 1) _i ) Input as input features to the second hidden layer h2 _i Same miningCalculating by using the formula (1-2), and substituting the calculation result of the second hidden layer into the formula (1-3) to obtain f (h 2) _i ) Then f (h 2) _i ) Input to output layer o as input features _i Calculating by adopting a formula (1-2), and substituting the calculation result of the output layer into a formula (1-3) to obtain f (o) _i )，f（o _i ) Probability values of straight, left-turn and right-turn of the obstacle vehicle are predicted for the initial driving direction prediction model, and a second predicted driving direction corresponding to the obstacle vehicle can be determined based on the probability values.

For example, if the probability value output by the initial driving direction prediction model is [1, 0, 0], it is determined that the second predicted driving direction of the obstacle vehicle is a left turn; if the probability value output by the initial driving direction prediction model is [0, 0,1], determining that the second predicted driving direction of the obstacle vehicle is a right turn; and if the probability value output by the initial driving direction prediction model is [0,1, 0], determining that the second predicted driving direction of the obstacle vehicle is straight.

Further, a running direction loss function corresponding to the obstacle vehicle is constructed according to the second predicted running direction and the historical running direction corresponding to the obstacle vehicle, and the specific formula of the running direction loss function is as follows:

（1-4）

wherein l is a traveling direction loss function representing a deviation value between the expected output and the actual output, and y represents a difference between the expected output and the actual output as the loss function value becomes smaller _i And f (x) is a historical driving direction (actual output) corresponding to the historical driving parameter information of the obstacle vehicle, and f (x) is a second predicted driving direction (expected output) corresponding to the historical driving parameter information, wherein N represents the number of samples in the training sample set. In the process of iteratively updating the initial weight parameter value by using the driving direction loss function, the second predicted driving direction is continuously changed because the initial weight parameter value is continuously updated, and when the driving direction loss function value is minimum, the iteration is stopped,and outputting the weight parameter value of the corresponding iteration level.

Specifically, in the embodiment of the present invention, a Batch Gradient Descent (BGD) method is used to solve the minimum value of the travel direction loss function, and then the weight parameter value w is adjusted in the reverse direction ₁ …w _n The update formula of the weight parameter value is as follows:

（1-5）

wherein,

for learning rate, the value range (0,1) is taken, i and i +1 represent the ith and (i + 1) th iterations respectively,

for the weight w of the loss function ₁ …w _n The deviation is calculated and the deviation is calculated,

namely, it is

（1-6）

Where m is the number of samples in the training sample set, f (x) is the expected output by the activation function, yj is the actual output, and x is the input historical driving parameter information.

Therefore, the initial weight parameter values can be continuously updated and iterated through the formulas (1-5) and (1-6), when the loss function value of the driving direction is minimum, iteration is stopped, the weight parameter values of the corresponding iteration levels are output, and therefore the preset driving direction prediction model can be determined according to the weight parameter values of the corresponding iteration levels.

203. And generating a driving track corresponding to the obstacle vehicle by adopting a corresponding track generation model according to the first predicted driving direction.

For the implementation of the invention, when the first predicted driving direction is the left turn or the right turn of the obstacle vehicle, a sampling trajectory planner is adopted to generate the driving trajectory corresponding to the obstacle vehicle. As an optional implementation manner, the method for generating the driving track of the obstacle vehicle by using the sampling track planner includes: sampling the drivable area of the obstacle vehicle to obtain abscissa, ordinate, course angle and curvature value corresponding to a plurality of sampling points respectively; determining a plurality of driving curves which meet vehicle kinematic constraints between the current position of the obstacle vehicle and the plurality of sampling points based on the abscissa, the ordinate, the course angle and the curvature value respectively corresponding to the sampling trajectory planner and the plurality of sampling points; and screening at least one driving curve from the plurality of driving curves according to the road environment where the obstacle vehicle is located, and taking the at least one driving curve as a driving track corresponding to the obstacle vehicle.

Specifically, if the first predicted travel direction is a left turn of the obstacle vehicle, a series of points are sampled in a travelable area on the left side of the road; if the first predicted driving direction is the right turning of the obstacle vehicle, a certain series of points are sampled in a drivable area on the right side of the road, namely, the current position of the obstacle vehicle is taken as a starting point, the sampling points are taken as an end point, and a plurality of driving curves meeting the kinematic constraint of the vehicle are generated. During specific sampling, acquiring an abscissa, an ordinate, a course angle and a curvature value of the obstacle vehicle at each sampling point in the future, and then determining a plurality of driving curves between the current position of the obstacle vehicle and the plurality of sampling points according to the abscissa, the ordinate, the course angle and the curvature value which are respectively corresponding to the curvature function about the arc length of the curve and the plurality of sampling points in the sampling track planner, wherein the driving curves are substantially cubic polynomial spiral lines, and the curvature function about the arc length of the curve is as follows:

（1-7）

wherein, the k(s) curvature function is a cubic polynomial function about the curve arc length, s is the curve arc length, and a, b, c and d are unknown curvature coefficients to be solved. Therefore, only by determining the curvature coefficients a, b, c and d through the abscissa, the ordinate, the course angle and the curvature value corresponding to any one sampling point, a driving curve between the current position of the obstacle vehicle and the sampling point can be determined, and when a plurality of sampling points exist, a plurality of driving curves can be determined. And then screening at least one driving curve from the plurality of driving curves according to the actual environment of the road to be used as the driving track of the obstacle vehicle, and if the driving curve with obvious obstacles and the obstacle vehicle can not be driven is excluded from the plurality of driving curves.

Further, the sampling trajectory planner includes an abscissa function, an ordinate function, a course angle function and a curvature function about the arc length of the curve, and the specific formula is as follows:

（1-8）

where x(s) is an abscissa function with respect to the arc length of the curve, y(s) is an ordinate function with respect to the arc length of the curve, and θ(s) is a heading angle function with respect to the arc length of the curve. For any one sampling point (target sampling point) in the plurality of sampling points, the abscissa, the ordinate, the course angle and the curvature value corresponding to the target sampling point and the abscissa, the ordinate, the course angle and the curvature value corresponding to the current position of the obstacle vehicle are known and are substituted into the equation (1-8) to solve the equation, so that the curvature coefficients a, b, c and d can be determined, but because the expression comprises Fresnel integral, the curvature coefficients a, b, c and d are difficult to directly solve, and therefore the embodiment of the invention adopts a Newton root method to solve the curvature coefficients a, b, c and d of the cubic polynomial.

Based on this, the method comprises: aiming at a target sampling point in the plurality of sampling points, solving a curvature coefficient corresponding to a curvature function about the curve arc length in the sampling track planner by using a Newton root method according to an abscissa, an ordinate, a course angle and a curvature value corresponding to the target sampling point; and determining a driving curve between the current position of the obstacle vehicle and the target sampling point, wherein the driving curve meets vehicle kinematic constraint, based on the curvature coefficient.

Further, for a specific process of solving curvature coefficients by using a newton root method, as an optional embodiment, the method includes: determining a curvature coefficient function for a start curvature value, a first curvature, a second curvature, an end curvature value, and a curve arc length from the start curvature value, the end curvature value at the target sampling point, and the curvature function for a curve arc length of the obstacle vehicle at the current position, respectively, wherein the first curvature and the second curvature are curvatures at 1/3 curve arc length and 2/3 curve arc length, respectively; determining an abscissa function, an ordinate function, and a course angle function with respect to a start curvature value, a first curvature, a second curvature, a curve arc length, and an end curvature value, respectively, based on the curvature coefficient function and the sampling trajectory planner; acquiring a first initial value, a second initial value and a third initial value corresponding to a first curvature, a second curvature and a curve arc length respectively, and substituting the first initial value, the second initial value, the third initial value, the starting point curvature value and the ending point curvature value into the abscissa function, the ordinate function, the course angle function and the curvature coefficient function respectively to obtain an abscissa, an ordinate, a course angle and a curvature value corresponding to a target insertion point between the current position of the obstacle vehicle and the target sampling point; correspondingly subtracting the abscissa, the ordinate, the course angle and the curvature value corresponding to the target sampling point from the abscissa, the ordinate, the course angle and the curvature value corresponding to the target insertion point to obtain an abscissa difference value, an ordinate difference value, a course angle difference value and a curvature difference value between the target sampling point and the target insertion point, and calculating the difference sum of the abscissa difference value, the ordinate difference value, the course angle difference value and the curvature difference value; continuously updating and iterating the first initial value, the second initial value and the third initial value based on the sum of the difference values until the abscissa, the ordinate, the course angle and the curvature value corresponding to the target insertion point are respectively the same as the abscissa, the ordinate, the course angle and the curvature value corresponding to the target sampling point, and outputting a first curvature value, a second curvature value and a curve arc length after final iteration updating; and determining a curvature coefficient according to the first curvature value, the second curvature value, the curve arc length and the curvature coefficient function after the final iteration updating.

Further, the continuously updating and iterating the first initial value, the second initial value, and the third initial value based on the sum of the difference values until the abscissa, the ordinate, the course angle, and the curvature value corresponding to the target insertion point are respectively the same as the abscissa, the ordinate, the course angle, and the curvature value corresponding to the target sampling point, and outputting the finally iteratively updated first curvature value, second curvature value, and curve arc length includes: calculating Jacobian matrixes respectively corresponding to the first curvature, the second curvature and the curve arc length based on the abscissa function, the ordinate function and the course angle function; calculating inverse matrixes of Jacobian matrixes corresponding to the first curvature, the second curvature and the curve arc length respectively, and multiplying the inverse matrixes corresponding to the first curvature, the second curvature and the curve arc length respectively by the difference sum to obtain a first curvature variation, a second curvature variation and a curve arc length variation of the iteration; updating the first initial value, the second initial value and the third initial value respectively based on the first curvature variation, the second curvature variation and the curve arc length variation; and repeating the iterative updating process of the first initial value, the second initial value and the third initial value until the sum of the difference values is 0, and outputting a first curvature value, a second curvature value and a curve arc length after final iterative updating.

Specifically, assume that the total arc length of the curve is S _f The curvature at the starting position 0 of the arc length of the curve is P ₀ The curvature at the arc of the third curve is P ₁ The curvature at the arc of the two-thirds curve is P ₂ The curvature at the end of the arc length of the curve is P ₃ From this, the following expression can be obtained:

（1-9）

further, it can be found by simultaneous equations (1-7) and (1-9):

（1-10）

wherein, P ₀ Is the curvature at the current position of the obstacle vehicle, i.e., the curvature at the curve arc length start position 0; p is ₃ Is the curvature at the target sampling point, i.e. the curvature at the end point of the arc length of the curve; p ₀ And P ₃ Are all known, S _f 、P ₁ And P ₂ Is unknown, as can be seen from equations (1-10), as long as S is solved _f 、P ₁ And P ₂ The curvature coefficients a, b, c, d can be determined.

Based on this, the embodiment of the present invention will solve S _f 、P ₁ And P ₂ Is described as a root-finding problem as follows:

（1-12）

wherein x is _des Representing the abscissa, ordinate, course angle and curvature value, x, corresponding to the target sampling point _p (s _f ) The method comprises the following steps of representing an abscissa, an ordinate, a course angle and a curvature value corresponding to a target insertion point between the current position of the unmanned vehicle and a target sampling point, wherein the abscissa, the ordinate, the course angle and the curvature value corresponding to the target insertion point can be changed continuously in the iteration process, when the abscissa, the ordinate, the course angle and the curvature value corresponding to the target insertion point are equal to the abscissa, the ordinate, the course angle and the curvature value corresponding to the target sampling point, the target insertion point is coincided with the target sampling point, and the iteration is stopped, wherein the specific formula is as follows:

（1-13）

wherein x is _pi (s _f ) Representing the abscissa, the ordinate, the course angle and the curvature value corresponding to the target insertion point in the ith round of iteration process, deltax representing the difference sum of the abscissa difference, the ordinate difference, the course angle difference and the curvature difference between the target sampling point and the target insertion point, and P _i Represents S in the ith iteration process _f 、P ₁ And P ₂ Value of (A), P _i+1 Represents S in the (i + 1) th iteration process _f 、P ₁ And P ₂ Δ P represents a first curvature variation (P) ₁ Variation), second curvature variation (P) ₂ Variation) and curve arc length variation (S) _f Amount of change), J _pi (x _pi (s _f ) ) represents S _f 、P ₁ And P ₂ Respectively corresponding Jacobian matrices, J _pi (x _pi (s _f )) ^-1 To represent the inverse of the Jacobian matrix, dx _pi (s _f ) Represents with respect to S _f 、P ₁ And P ₂ The abscissa function, the ordinate function and the course angle function of the target vector are respectively corresponding to S _f 、P ₁ And P ₂ The derivation of the partial derivative can be determined by the simultaneous equations (1-10) and (1-8) _f 、P ₁ And P ₂ Abscissa function, ordinate function and course angle function of d _pi As a parameter, 0.001 was set.

Particularly, the formula is utilized to solve S _f 、P ₁ And P ₂ Firstly, S is respectively given according to actual experience _f 、P ₁ And P ₂ After the first initial value, the second initial value and the third initial value are substituted into a value formula (1-10), an initial curvature coefficient is determined, S =0.1 is then set, and the initial curvature coefficient is substituted into a value formula (1-8), so as to obtain an abscissa, an ordinate, a navigation angle and a curvature value corresponding to the target insertion point, wherein the curve arc length S corresponding to the target insertion point is increased by 0.1 every iteration. Further, the abscissa, the ordinate, the navigation angle and the curvature value corresponding to the target insertion point and the abscissa, the ordinate, the navigation angle and the curvature value corresponding to the target sampling point are substituted into the formula (1-13) to calculate the difference sum Δ x.

Further, with respect to P, based on the abscissa function, the ordinate function and the heading angle function ₁ And the first initial value, the second initial value and the third initial value, calculating P ₁ The inverse of the corresponding Jacobian matrix, and apply P to the inverse of the Jacobian matrix ₁ The inverse matrix of the corresponding Jacobian matrix is multiplied by the difference sum delta x to obtain P of the iteration of the current round ₁ The variation can obtain P of the iteration in the same way ₂ Amount of change and S _f Yellowing amount. Then respectively pairing S based on the variables _f 、P ₁ And P ₂ Is updated, thereby completing the cost wheel pair S _f 、P ₁ And P ₂ The update iteration process of (1). Repeating the above pair S _f 、P ₁ And P ₂ Until Deltax is 0, namely the abscissa, ordinate, course angle and curvature value corresponding to the target insertion point are the same as those corresponding to the target sampling point, stopping updating the iterative process, and obtaining the final output S _f 、P ₁ And P ₂ And S to be finally output _f 、P ₁ And P ₂ The value of (a) is substituted into the formula (1-10), so that a driving curve between the current position of the obstacle vehicle and the target sampling point, which satisfies the vehicle kinematic constraint, can be determined. By executing the above process for each sampling point, a plurality of driving curves can be generated, so that at least one driving curve can be screened out from the plurality of driving curves according to the road environment to serve as the driving track of the obstacle vehicle.

According to the embodiment of the invention, the driving track meeting the vehicle kinematic constraint is generated according to the predicted driving direction of the obstacle vehicle, so that the generation precision of the driving track can be ensured, and powerful data support can be provided for the subsequent collision prediction.

204. And acquiring a running track corresponding to the unmanned vehicle, and judging whether the running track of the unmanned vehicle is crossed with the running track of the obstacle vehicle.

In the embodiment of the present invention, since the driving trajectory corresponding to the unmanned vehicle is usually planned in advance, the driving trajectory of the unmanned vehicle (own vehicle) can be directly obtained, and the driving decision can be made according to the driving trajectory of the own vehicle and the driving trajectory of the obstacle vehicle. Specifically, whether the running track of the unmanned vehicle is crossed with the running track of the obstacle vehicle or not is judged, if so, the collision risk between the unmanned vehicle and the obstacle vehicle is shown, and the unmanned vehicle needs to be controlled to decelerate or stop; if no intersection exists, the unmanned vehicle and the obstacle vehicle are not in collision risk.

205. And if the running track of the unmanned vehicle is crossed with the running track of the obstacle vehicle, determining that the unmanned vehicle has a collision risk, and controlling the unmanned vehicle to decelerate or stop.

For the embodiment of the invention, when a driving decision is made for an unmanned vehicle, an intersection point of a driving track of the unmanned vehicle and a driving track of an obstacle vehicle is determined, then a distance between the current position of the unmanned vehicle and the intersection point is calculated, and if the distance is less than a preset distance, the unmanned vehicle is controlled to stop; and if the distance is greater than or equal to a preset distance, controlling the unmanned vehicle to decelerate. The preset distance can be set according to the actual running condition of the unmanned vehicle.

Further, as a specific implementation of fig. 1, an embodiment of the present invention provides a driving trajectory generating device for a mine obstacle vehicle, as shown in fig. 5, the device includes: an acquisition unit 31, a prediction unit 32, and a generation unit 33.

The obtaining unit 31 may be configured to obtain obstacle vehicles in a mine area within a predetermined distance range from the unmanned vehicle, and driving parameter information corresponding to the obstacle vehicles.

The predicting unit 32 may be configured to input the driving parameter information into a preset driving direction prediction model to perform direction prediction, so as to obtain a first predicted driving direction corresponding to the obstacle vehicle.

The generating unit 33 may be configured to generate a driving track corresponding to the obstacle vehicle by using a corresponding track generation model according to the first predicted driving direction.

In a specific application scenario, the preset driving direction prediction model is a preset neural network model, and the prediction unit 32, as shown in fig. 6, includes: a prediction module 321 and a first determination module 322.

The prediction module 321 may be configured to input the driving parameter information into the preset neural network model to perform direction prediction, so as to obtain probability values of the obstacle vehicle driving in each direction.

The first determining module 322 may be configured to screen a maximum probability value from the probability values of the obstacle vehicle traveling in the respective directions, and determine a direction corresponding to the maximum probability value as a first predicted traveling direction of the obstacle vehicle.

Further, the preset neural network model includes an input layer, a first hidden layer, a second hidden layer, and an output layer, and the prediction module 321 may be specifically configured to perform normalization processing on the driving parameter information to obtain processed driving parameter information; inputting the processed driving parameter information to the first hidden layer through the input layer for feature extraction to obtain a first driving parameter feature corresponding to the obstacle vehicle; inputting the first driving parameter characteristics to the second hidden layer for characteristic extraction to obtain second driving parameter characteristics corresponding to the obstacle vehicle; and inputting the second driving parameter characteristics to the output layer for classification to obtain probability values of driving of the obstacle vehicle in all directions.

In a specific application scenario, the apparatus further includes: a building unit 34 and an updating unit 35.

The obtaining unit 31 may be further configured to obtain historical driving parameter information and historical driving directions corresponding to the obstacle vehicle.

The constructing unit 34 may be configured to construct an initial driving direction prediction model by using a preset neural network algorithm.

The prediction unit 32 may be further configured to input the historical driving parameter information into the initial driving direction prediction model to perform direction prediction, so as to obtain a second predicted driving direction corresponding to the obstacle vehicle.

The constructing unit 34 may be further configured to construct a traveling direction loss function corresponding to the obstacle vehicle based on the second predicted traveling direction and the historical traveling direction corresponding to the obstacle vehicle.

The updating unit 35 may be configured to continuously perform iterative updating on the initial weight parameter value in the initial driving direction prediction model based on the driving direction loss function, and when the driving direction loss function is minimum, output a weight parameter value of a corresponding iterative level; and determining the preset driving direction prediction model based on the weight parameter values of the corresponding iteration levels.

Further, the updating unit 35 may be specifically configured to calculate a partial derivative of the weight parameter in the traveling direction loss function, so as to obtain a partial derivative function related to the weight parameter; updating the initial weight parameter value based on the partial derivative function to obtain the weight parameter value after the iteration update in the current round; and repeating the process of iteratively updating the initial weight parameter values until the weight parameter values of the corresponding iteration levels are output when the driving direction loss function is minimum.

In a specific application scenario, the generating unit 33 may be specifically configured to generate a driving track corresponding to the obstacle vehicle by using a sampling track planner if the first predicted driving direction is a left turn or a right turn of the obstacle vehicle; and if the first predicted driving direction is that the obstacle vehicle moves straight, generating a driving track corresponding to the obstacle vehicle by adopting a vehicle kinematics model.

Further, the generating unit 33 includes: a sampling module 331, a second determination module 332, and a screening module 333.

The sampling module 331 may be configured to sample a travelable area of the obstacle vehicle to obtain an abscissa, an ordinate, a course angle, and a curvature value corresponding to the plurality of sampling points, respectively.

The second determining module 332 may be configured to determine a plurality of driving curves between the current position of the obstacle vehicle and the plurality of sampling points, where the driving curves satisfy vehicle kinematic constraints, based on the abscissa, the ordinate, the heading angle, and the curvature value corresponding to the sampling trajectory planner and the plurality of sampling points, respectively.

The screening module 333 may be configured to screen at least one driving curve from the plurality of driving curves according to a road environment where the obstacle vehicle is located, where the at least one driving curve is used as a driving track corresponding to the obstacle vehicle.

Further, the second determining module 332 includes: a solution submodule and a determination submodule.

The solving submodule can be used for solving a curvature coefficient corresponding to a curvature function about the curve arc length in the sampling track planner by utilizing a Newton root method according to an abscissa, an ordinate, a course angle and a curvature value corresponding to a target sampling point in the plurality of sampling points.

The determining submodule may be configured to determine, based on the curvature coefficient, a driving curve between the current position of the obstacle vehicle and the target sampling point, which satisfies vehicle kinematics constraints.

Further, the solving submodule may be specifically configured to determine a curvature coefficient function with respect to a start curvature value, a first curvature, a second curvature, an end curvature value, and a curve arc length according to a start curvature value of the obstacle vehicle at the current position, an end curvature value at the target sampling point, and the curvature function with respect to the curve arc length, wherein the first curvature and the second curvature are curvatures at 1/3 curve arc length and 2/3 curve arc length, respectively; determining an abscissa function, an ordinate function, and a course angle function with respect to a start curvature value, a first curvature, a second curvature, a curve arc length, and an end curvature value, respectively, based on the curvature coefficient function and the sampling trajectory planner; acquiring a first initial value, a second initial value and a third initial value corresponding to a first curvature, a second curvature and a curve arc length respectively, and substituting the first initial value, the second initial value, the third initial value, the starting point curvature value and the ending point curvature value into the abscissa function, the ordinate function, the course angle function and the curvature coefficient function respectively to obtain an abscissa, an ordinate, a course angle and a curvature value corresponding to a target insertion point between the current position of the obstacle vehicle and the target sampling point; correspondingly subtracting the abscissa, the ordinate, the course angle and the curvature value corresponding to the target sampling point from the abscissa, the ordinate, the course angle and the curvature value corresponding to the target insertion point to obtain an abscissa difference value, an ordinate difference value, a course angle difference value and a curvature difference value between the target sampling point and the target insertion point, and calculating the difference sum of the abscissa difference value, the ordinate difference value, the course angle difference value and the curvature difference value; continuously updating and iterating the first initial value, the second initial value and the third initial value based on the sum of the difference values until the abscissa, the ordinate, the course angle and the curvature value corresponding to the target insertion point are respectively the same as the abscissa, the ordinate, the course angle and the curvature value corresponding to the target sampling point, and outputting a first curvature value, a second curvature value and a curve arc length after final iteration updating; and determining a curvature coefficient according to the first curvature value, the second curvature value, the curve arc length and the curvature coefficient function after the final iteration update.

Further, the solving submodule may be further configured to calculate, based on the abscissa function, the ordinate function, and the course angle function, jacobian matrices corresponding to the first curvature, the second curvature, and the curve arc length, respectively; calculating inverse matrixes of Jacobian matrixes corresponding to the first curvature, the second curvature and the curve arc length respectively, and multiplying the inverse matrixes corresponding to the first curvature, the second curvature and the curve arc length respectively by the difference sum to obtain a first curvature variation, a second curvature variation and a curve arc length variation of the iteration; updating the first initial value, the second initial value and the third initial value respectively based on the first curvature variation, the second curvature variation and the curve arc length variation; and repeating the iterative updating process of the first initial value, the second initial value and the third initial value until the sum of the difference values is 0, and outputting a first curvature value, a second curvature value and a curve arc length after final iterative updating.

Further, the sampling module 331 may be further configured to sample any point on a straight line along the direction of the target course angle according to the target course angle of the obstacle vehicle at the current position, so as to obtain an abscissa and an ordinate corresponding to any sampling point.

The second determining unit 332 may be further configured to determine a driving straight line between the current position of the obstacle vehicle and the arbitrary sampling point according to an abscissa and an ordinate corresponding to the current position of the obstacle vehicle and an abscissa and an ordinate corresponding to the arbitrary sampling point; and determining the driving straight line as a driving track corresponding to the obstacle vehicle.

In a specific application scenario, the apparatus further includes: a determination unit 36 and a control unit 37.

The determination unit 36 may be configured to acquire a travel track corresponding to the unmanned vehicle, and determine whether the travel track of the unmanned vehicle intersects with the travel track of the obstacle vehicle.

The control unit 37 may be configured to determine that the unmanned vehicle has a collision risk if the traveling track of the unmanned vehicle intersects with the traveling track of the obstacle vehicle, and control the unmanned vehicle to decelerate or stop.

Further, the control unit 37 may be specifically configured to determine an intersection of a travel trajectory of the unmanned vehicle and a travel trajectory of the obstacle vehicle; calculating a distance between a current location of the unmanned vehicle and the intersection; if the distance is smaller than the preset distance, controlling the unmanned vehicle to stop; and if the distance is greater than or equal to a preset distance, controlling the unmanned vehicle to decelerate.

It should be noted that other corresponding descriptions of the functional modules involved in the device for generating the driving track of the obstacle vehicle in the mining area according to the embodiment of the present invention may refer to the corresponding descriptions of the method shown in fig. 1, and are not described herein again.

Based on the method shown in fig. 1, correspondingly, the embodiment of the invention also provides a computer-readable storage medium, on which a computer program is stored, and the program is executed by a processor to implement the mine area obstacle vehicle driving track generation method in the embodiment.

Based on the above embodiments of the method shown in fig. 1 and the apparatus shown in fig. 5, an embodiment of the present invention further provides an entity structure diagram of a computer device, as shown in fig. 7, where the computer device includes: a processor 41, a memory 42, and a computer program stored on the memory 42 and operable on the processor, wherein the memory 42 and the processor 41 are both arranged on a bus 43, and the processor 41 implements the method for generating a driving trace of a mine obstacle vehicle in the above-described embodiment when the processor 41 executes the program.

According to the technical scheme, the obstacle vehicle in the mine area within the preset distance range from the unmanned vehicle and the driving parameter information corresponding to the obstacle vehicle are obtained; inputting the running parameter information into a preset running direction prediction model for direction prediction to obtain a first predicted running direction corresponding to the obstacle vehicle; and finally, generating a driving track corresponding to the obstacle vehicle by adopting a corresponding track generation model according to the first predicted driving direction. Because the driving tracks of the obstacle vehicle in different directions are different, the driving direction of the obstacle vehicle is predicted by using the preset driving direction prediction model, and on the basis, the driving track meeting the vehicle kinematic constraint is generated by using the track generation model, so that the generation precision of the driving track of the obstacle vehicle can be ensured, the subsequent accurate judgment of whether the unmanned vehicle has collision risk with the obstacle vehicle is facilitated, and the unmanned vehicle makes driving behavior decision to avoid the risk.

Fig. 8 is a schematic structural diagram of a chip according to an embodiment of the present invention, and as shown in fig. 8, the chip 500 includes one or more than two (including two) processors 510 and a communication interface 530. The communication interface 530 is coupled to the at least one processor 510, and the at least one processor 510 is configured to execute a computer program or instructions to implement the method for generating a driving trajectory for a mine obstacle vehicle as in the above-described embodiments.

Preferably, the memory 540 stores the following elements: an executable module or a data structure, or a subset thereof, or an expanded set thereof.

In an embodiment of the invention, memory 540 may include both read-only memory and random access memory and provide instructions and data to processor 510. A portion of memory 540 may also include non-volatile random access memory (NVRAM).

In an embodiment of the present invention, memory 540, communication interface 530, and memory 540 are coupled together by bus system 520. The bus system 520 may include a power bus, a control bus, a status signal bus, and the like, in addition to the data bus. For ease of description, the various buses are labeled as bus system 520 in FIG. 8.

The method described in the embodiments of the present application may be applied to the processor 510, or implemented by the processor 510. Processor 510 may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the above method may be performed by integrated logic circuits of hardware or instructions in the form of software in the processor 510. The processor 510 may be a general-purpose processor (e.g., a microprocessor or a conventional processor), a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an FPGA (field-programmable gate array) or other programmable logic device, discrete gate, transistor logic device or discrete hardware component, and the processor 510 may implement or execute the methods, steps and logic blocks disclosed in the embodiments of the present invention.

Fig. 9 is a schematic structural diagram of a terminal according to an embodiment of the present invention, and as shown in fig. 9, the terminal 600 includes the driving trajectory generation device 100 for the mine obstacle vehicle.

The terminal 600 may perform the method described in the above embodiment by the mine obstacle vehicle travel track generation apparatus 100. It can be understood that the implementation manner of the terminal 600 controlling the driving track generating device 100 of the obstacle vehicle in the mining area may be set according to an actual application scenario, and the embodiment of the present application is not particularly limited.

The terminal 600 includes but is not limited to: the vehicle can implement the method provided by the application through the vehicle-mounted terminal, the vehicle-mounted controller, the vehicle-mounted module, the vehicle-mounted component, the vehicle-mounted chip, the vehicle-mounted unit, the vehicle-mounted radar or the camera.

The terminal in the embodiment of the invention is used as a control or adjustment system for executing non-electric variables, predicts the running direction of the obstacle vehicle, and generates the running track corresponding to the obstacle vehicle on the basis of the prediction, thereby realizing the improvement of the running track generation precision of the obstacle vehicle.

It will be apparent to those skilled in the art that the modules or steps of the present invention described above may be implemented by a general purpose computing device, they may be centralized on a single computing device or distributed across a network of multiple computing devices, and alternatively, they may be implemented by program code executable by a computing device, such that they may be stored in a storage device and executed by a computing device, and in some cases, the steps shown or described may be performed in an order different than that described herein, or they may be separately fabricated into individual integrated circuit modules, or multiple ones of them may be fabricated into a single integrated circuit module. Thus, the present invention is not limited to any specific combination of hardware and software.

The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made without departing from the spirit and principle of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for generating a driving track of a mine obstacle vehicle is characterized by comprising the following steps:

2. The method according to claim 1, wherein the preset driving direction prediction model is a preset neural network model, and the inputting the driving parameter information into the preset driving direction prediction model for direction prediction to obtain a first predicted driving direction corresponding to the obstacle vehicle comprises:

inputting the driving parameter information into the preset neural network model for direction prediction to obtain probability values of driving of the obstacle vehicle in all directions;

and screening out a maximum probability value from the probability values of the obstacle vehicle in the traveling directions, and determining the direction corresponding to the maximum probability value as a first predicted traveling direction of the obstacle vehicle.

3. The method of claim 2, wherein the preset neural network model comprises an input layer, a first hidden layer, a second hidden layer and an output layer, and the inputting of the driving parameter information into the preset neural network model for direction prediction to obtain probability values of driving of the obstacle vehicle in all directions comprises:

carrying out normalization processing on the driving parameter information to obtain processed driving parameter information;

inputting the processed driving parameter information to the first hidden layer through the input layer for feature extraction to obtain a first driving parameter feature corresponding to the obstacle vehicle;

inputting the first driving parameter characteristics to the second hidden layer for characteristic extraction to obtain second driving parameter characteristics corresponding to the obstacle vehicle;

and inputting the second driving parameter characteristics to the output layer for classification to obtain probability values of driving of the obstacle vehicle in all directions.

4. The method according to claim 1, wherein before the inputting the driving parameter information into a preset driving direction prediction model for direction prediction to obtain a first predicted driving direction corresponding to the obstacle vehicle, the method further comprises:

acquiring historical driving parameter information and historical driving directions corresponding to the obstacle vehicle;

constructing an initial driving direction prediction model by using a preset neural network algorithm;

inputting the historical driving parameter information into the initial driving direction prediction model for direction prediction to obtain a second predicted driving direction corresponding to the obstacle vehicle;

constructing a running direction loss function corresponding to the obstacle vehicle based on a second predicted running direction and a historical running direction corresponding to the obstacle vehicle;

continuously carrying out iterative updating on initial weight parameter values in the initial driving direction prediction model on the basis of the driving direction loss function, and outputting weight parameter values of corresponding iterative levels when the driving direction loss function is minimum;

and determining the preset driving direction prediction model based on the weight parameter values of the corresponding iteration levels.

5. The method according to claim 4, wherein the continuously iteratively updating the initial weight parameter values in the initial driving direction prediction model based on the driving direction loss function, and outputting the weight parameter values of the corresponding iteration level when the driving direction loss function is minimum comprises:

calculating the partial derivative of the weight parameter in the running direction loss function to obtain a partial derivative function related to the weight parameter;

updating the initial weight parameter value based on the partial derivative function to obtain a weight parameter value after the iteration updating in the current round;

and repeating the process of iteratively updating the initial weight parameter values until the weight parameter values of the corresponding iteration levels are output when the driving direction loss function is minimum.

6. The method of claim 1, wherein generating a corresponding travel trajectory for the obstacle vehicle using a corresponding trajectory generation model based on the first predicted travel direction comprises:

if the first predicted driving direction is the left turn or the right turn of the obstacle vehicle, generating a driving track corresponding to the obstacle vehicle by adopting a sampling track planner;

and if the first predicted driving direction is that the obstacle vehicle moves straight, generating a driving track corresponding to the obstacle vehicle by adopting a vehicle kinematics model.

7. The method of claim 6, wherein generating the corresponding travel trajectory for the obstacle vehicle using a sample trajectory planner comprises:

sampling the travelable area of the obstacle vehicle to obtain abscissa, ordinate, course angle and curvature values corresponding to a plurality of sampling points respectively;

determining a plurality of driving curves which meet vehicle kinematic constraints between the current position of the obstacle vehicle and the plurality of sampling points based on the abscissa, the ordinate, the course angle and the curvature value respectively corresponding to the sampling trajectory planner and the plurality of sampling points;

and screening at least one driving curve from the plurality of driving curves according to the road environment where the obstacle vehicle is located, and taking the at least one driving curve as a driving track corresponding to the obstacle vehicle.

8. The method of claim 7, wherein determining a plurality of driving curves between the current position of the obstacle vehicle and the plurality of sampling points that satisfy vehicle kinematic constraints based on abscissa, ordinate, heading angle, and curvature values corresponding to the sampling trajectory planner and the plurality of sampling points, respectively, comprises:

aiming at a target sampling point in the plurality of sampling points, solving a curvature coefficient corresponding to a curvature function about the curve arc length in the sampling track planner by using a Newton root method according to an abscissa, an ordinate, a course angle and a curvature value corresponding to the target sampling point;

and determining a driving curve between the current position of the obstacle vehicle and the target sampling point, wherein the driving curve meets vehicle kinematic constraint, based on the curvature coefficient.

9. The method according to claim 8, wherein the solving the curvature coefficient corresponding to the curvature function about the curve arc length in the sampling trajectory planner by using a newton root method according to the abscissa, the ordinate, the course angle and the curvature value corresponding to the target sampling point comprises:

determining a curvature coefficient function for a start curvature value, a first curvature, a second curvature, an end curvature value, and a curve arc length from a start curvature value of the obstacle vehicle at the current position, an end curvature value at the target sampling point, and the curvature function for a curve arc length, wherein the first curvature and the second curvature are curvature at 1/3 curve arc length and 2/3 curve arc length, respectively;

determining an abscissa function, an ordinate function, and a course angle function with respect to a start curvature value, a first curvature, a second curvature, a curve arc length, and an end curvature value, respectively, based on the curvature coefficient function and the sampling trajectory planner;

acquiring a first initial value, a second initial value and a third initial value corresponding to a first curvature, a second curvature and a curve arc length respectively, and substituting the first initial value, the second initial value, the third initial value, the starting point curvature value and the ending point curvature value into the abscissa function, the ordinate function, the course angle function and the curvature coefficient function respectively to obtain an abscissa, an ordinate, a course angle and a curvature value corresponding to a target insertion point between the current position of the obstacle vehicle and the target sampling point;

correspondingly subtracting the abscissa, the ordinate, the course angle and the curvature value corresponding to the target sampling point from the abscissa, the ordinate, the course angle and the curvature value corresponding to the target insertion point to obtain an abscissa difference value, an ordinate difference value, a course angle difference value and a curvature difference value between the target sampling point and the target insertion point, and calculating the difference sum of the abscissa difference value, the ordinate difference value, the course angle difference value and the curvature difference value;

continuously updating and iterating the first initial value, the second initial value and the third initial value based on the sum of the difference values until the abscissa, the ordinate, the course angle and the curvature value corresponding to the target insertion point are respectively the same as the abscissa, the ordinate, the course angle and the curvature value corresponding to the target sampling point, and outputting a first curvature value, a second curvature value and a curve arc length after final iteration updating;

and determining a curvature coefficient according to the first curvature value, the second curvature value, the curve arc length and the curvature coefficient function after the final iteration updating.

10. The method according to claim 9, wherein continuously performing update iteration on the first initial value, the second initial value, and the third initial value based on the sum of the difference values until an abscissa, an ordinate, a heading angle, and a curvature value corresponding to a target insertion point are respectively the same as an abscissa, an ordinate, a heading angle, and a curvature value corresponding to the target sampling point, and outputting a final iteratively updated first curvature value, second curvature value, and curve arc length comprises:

calculating Jacobian matrixes respectively corresponding to the first curvature, the second curvature and the curve arc length based on the abscissa function, the ordinate function and the course angle function;

calculating inverse matrixes of Jacobian matrixes corresponding to the first curvature, the second curvature and the curve arc length respectively, and multiplying the inverse matrixes corresponding to the first curvature, the second curvature and the curve arc length respectively by the difference sum to obtain a first curvature variation, a second curvature variation and a curve arc length variation of the iteration;

updating the first initial value, the second initial value and the third initial value respectively based on the first curvature variation, the second curvature variation and the curve arc length variation;

and repeating the iterative updating process of the first initial value, the second initial value and the third initial value until the sum of the difference values is 0, and outputting the finally iteratively updated first curvature value, second curvature value and curve arc length.

11. The method of claim 6, wherein generating the corresponding travel path of the obstacle vehicle using the vehicle kinematics model comprises:

sampling any point on a straight line along the direction of the target course angle according to the target course angle of the obstacle vehicle at the current position to obtain an abscissa and an ordinate corresponding to any sampling point;

determining a driving straight line between the current position of the obstacle vehicle and the arbitrary sampling point according to the abscissa and the ordinate corresponding to the current position of the obstacle vehicle and the abscissa and the ordinate corresponding to the arbitrary sampling point;

and determining the driving straight line as a driving track corresponding to the obstacle vehicle.

12. The method according to any one of claims 1-11, further comprising:

acquiring a running track corresponding to the unmanned vehicle, and judging whether the running track of the unmanned vehicle is crossed with the running track of the obstacle vehicle;

and if the running track of the unmanned vehicle is crossed with the running track of the obstacle vehicle, determining that the unmanned vehicle has a collision risk, and controlling the unmanned vehicle to decelerate or stop.

13. The method of claim 12, wherein the controlling the unmanned vehicle to slow down or stop comprises:

determining an intersection of a travel trajectory of the unmanned vehicle and a travel trajectory of the obstacle vehicle;

calculating a distance between a current location of the unmanned vehicle and the intersection;

if the distance is smaller than the preset distance, controlling the unmanned vehicle to stop;

and if the distance is greater than or equal to a preset distance, controlling the unmanned vehicle to decelerate.

14. A mine obstacle vehicle travel track generation device is characterized by comprising:

the prediction unit is used for inputting the running parameter information into a preset running direction prediction model to perform direction prediction to obtain a first predicted running direction corresponding to the obstacle vehicle;

15. A chip comprising at least one processor and a communication interface, the communication interface coupled with the at least one processor, the at least one processor for executing a computer program or instructions to implement the method of mine obstacle vehicle travel trajectory generation of any of claims 1-13.

16. A terminal characterized in that the terminal comprises the mine obstacle vehicle running trajectory generating device according to claim 14.

17. A computer arrangement comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the computer program realizes the steps of the method of any of claims 1 to 13 when executed by the processor.

18. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 13.