Unmanned mine card safety protection method based on double-layer risk estimation

Abstract

The invention relates to an unmanned mine card safety protection method based on double-layer risk estimation, belongs to the technical field of unmanned driving safety, and solves the problem that a passive safety protection mode in the prior art can only relieve accident hazards after an accident occurs and cannot perform accident prevention and early warning. Obtaining a prediction track; s2, acquiring information of a task path and a sensing obstacle, and performing collision detection and safe distance calculation on the task path, the predicted trajectory and the sensing obstacle to obtain a safe distance; s3, performing safety risk field calculation according to the obtained obstacle sensing information to obtain a safety risk score; and S4, based on the obtained safety distance and the safety risk field, carrying out control quantity calculation on the vehicle speed, obtaining a control instruction, outputting the control instruction, and carrying out driving control on the vehicle.

Description

Technical Field

[0001] The present invention relates to the field of unmanned driving safety technology, and in particular to an unmanned mining truck safety protection method based on double-layer risk estimation. Background Art

[0002] Since the mining area is sufficiently closed, the scene is relatively simple, there is less interaction between vehicles and people, and the market demand is strong, so autonomous driving in mining areas has always been considered the easiest scene for autonomous driving to land. However, mining trucks are extremely large in size and weight. For example, Caterpillar's 797F mining truck has a rated weight of 623.7 tons. If an accident occurs, it will cause very serious safety hazards and economic losses. Therefore, high requirements are placed on the safety of unmanned driving in mining areas. Unmanned mining trucks are generally equipped with advanced sensors (such as lidar, millimeter-wave radar, camera, combined inertial navigation, etc.), controllers, and actuators. The sensors perceive the environmental information around the unmanned mining trucks. The perception program processes this raw information and converts it into specific protocol data and sends it downstream. The downstream driving safety-related modules receive the processed environmental information and the vehicle's own driving status (such as positioning, speed, turning angle, etc.), evaluate the safety hazards of the vehicle at present and in the future, and enable the vehicle to respond in time to ensure the safety of unmanned driving in mining areas.

[0003] At present, unmanned driving safety methods include passive safety protection and active safety protection. Passive safety protection focuses on safety protection devices, such as airbags and anti-collision structures, which can mitigate the impact after the danger occurs and reduce the risk of equipment damage and property loss.

[0004] Active safety protection generally refers to the protection at the driverless program level. Through on-board sensors, vehicle networking and algorithms, information about the vehicle's form, environment and other traffic participants can be detected in a timely manner, and possible risks in the future can be predicted in advance, so that the vehicle can be controlled to perform specific actions in advance to avoid possible accident risks.

[0005] The passive safety protection methods in the existing technology can only reduce the harm of accidents after they occur, but cannot prevent and warn accidents. The active safety protection in the existing technology has high technical costs and is mainly aimed at urban roads with traffic signals and many vehicles interacting around. It is not applicable to mining roads with almost no traffic signal facilities and few interactions with other vehicles.

[0006] Therefore, there exists in the art a safety protection method that can provide active safety protection for unmanned mining trucks traveling on mining roads with complex boundaries and does not rely on traffic signal facilities. Summary of the invention

[0007] In order to solve the above problems, the embodiments of the present invention provide a safety protection method for unmanned vehicles in mining areas based on double-layer risk estimation. Specifically, considering that the boundaries of mining areas are relatively complex, the operation routes are relatively simple, and mining vehicles are relatively heavy, the control accuracy is lower than that of passenger cars, and the control deviation is relatively large, a collision detection method that integrates the preset driving path and vehicle trajectory estimation is proposed. The vehicle driving path and the predicted future trajectory of the vehicle are respectively used to determine whether there is a risk of collision between the vehicle and the road boundary and the surrounding environment obstacles. At the same time, the driving risk field is calculated based on the perceived obstacles, and the response action of the vehicle is determined based on the distance between the vehicle and the risk position and the strength of the risk field, thereby ensuring the safe and efficient operation of unmanned mining vehicles.

[0008] According to an embodiment of the present invention, a method for protecting an unmanned mining truck based on double-layer risk estimation is provided, comprising the following steps:

[0009] Step S1, predicting the vehicle trajectory to obtain a predicted trajectory;

[0010] Step S2, obtaining information about the task path and perceived obstacles, performing collision detection and safety distance calculation on the task path and predicted trajectory and perceived obstacles to obtain a safety distance;

[0011] Step S3, calculating the safety risk field based on the acquired information of perceived obstacles to obtain a safety risk score;

[0012] Step S4, calculating the control amount of the vehicle speed based on the obtained safety distance and safety risk field, obtaining and outputting a control instruction, and controlling the driving of the vehicle;

[0013] Wherein, the step S1 specifically includes the following steps:

[0014] Step S1.1, vectorizing the map information to obtain the map information of the mining area, representing the elements in the map information as discrete points, and composing the coordinates of the discrete points into vector sets according to the geographical location and functional classification, to obtain the vectorized map information;

[0015] Step S1.2, using a graph neural network to encode the vectorized map information to obtain the encoded map information, wherein the vectorized map information is composed of nodes and edges, wherein each node represents the starting point and the end point of each vector in the vector set;

[0016] Step S1.3, using a history trajectory encoding network including an FPN network (feature pyramid network) to encode the vehicle history trajectory to obtain an encoded vehicle history trajectory;

[0017] Step S1.4, processing the features of the encoded map information and the features of the encoded historical trajectory of the vehicle through the prediction network to obtain k future trajectories;

[0018] Step S1.5, score the k future trajectories output by the prediction network and select the trajectory with the highest probability as the predicted trajectory.

[0019] Optionally, the step S1.3 specifically includes the following steps:

[0020] Step S1.3.1, obtain the historical trajectory of the ego vehicle, vectorize it, and obtain the vectorized historical trajectory of the ego vehicle:

[0021]

[0022] Among them, x represents the x-coordinate set of the historical trajectory, x t represents the x-coordinate of the vehicle at time t, y represents the y-coordinate set of the historical trajectory, and y t represents the y coordinate of the ego vehicle at time t, h represents the set of ego vehicle headings in the historical trajectory, and h t represents the heading of the vehicle at time t, t represents the current time, and τ represents the time step;

[0023] Step S1.3.2, use the vector of the above-obtained vectorized historical trajectory of the vehicle as the input of the historical trajectory encoding network, use CNN (convolutional neural network) for training, obtain the features extracted by CNN, and then use the FPN network to perform multi-scale feature extraction on the features extracted by CNN to obtain the encoded historical trajectory of the vehicle.

[0024] Optionally, the step S2 specifically includes the following steps:

[0025] Step S2.1, obtaining the predicted trajectory and task path, and generating the vehicle envelope of the ego vehicle;

[0026] Step S2.2, obtaining information about perceived obstacles, filtering the obstacles, removing obstacles that are far away and obstacles that have no collision risk, obtaining obstacles to be detected, and processing the information about the obstacles to be detected to obtain an obstacle envelope;

[0027] Step S2.3, obtaining the vehicle envelope and the obstacle envelope, performing collision detection, and obtaining the collision detection result;

[0028] Step S2.4, based on the collision detection result, obtain the safe distance and output it.

[0029] Optionally, the step S2.1 specifically includes the following steps:

[0030] Step S2.1.1, obtain the predicted trajectory and task path, traverse the path points of the measured trajectory and task path, use the vehicle speed as the basis for envelope expansion, and generate the vehicle envelope frame of the path point coordinate system;

[0031] Step S2.1.2, based on the relationship between the path point and the vehicle position, the coordinates of the vehicle envelope are converted from the path point coordinate system to the vehicle coordinate system:

[0032]

[0033] Among them, (x a ,y a ) is the coordinate of the path point in the vehicle coordinate system, (x1′, y1′) is the coordinate of the corner point 1 of the vehicle envelope in the path point coordinate system, is the rotation angle from the path point coordinate system to the vehicle coordinate system, (x1, y1) is the coordinate of the corner point 1 of the vehicle envelope after rotation in the vehicle coordinate system;

[0034] Step S2.1.3, repeat the above step S2.1.2, calculate the coordinates of each corner point of the vehicle envelope frame in the vehicle coordinate system, obtain the vehicle envelope frame converted from the path point coordinate system to the vehicle envelope frame in the vehicle coordinate system, and output the converted vehicle envelope frame in the vehicle coordinate system as the vehicle envelope frame.

[0035] Optionally, the step S2.3 specifically includes:

[0036] Step S2.3.1, obtain the vehicle envelope, take out an edge from the vehicle envelope, and obtain its normal vector, the normal vector is used as a projection axis, and each edge of the obstacle envelope and the vehicle envelope is obtained in a loop, and they are projected onto the projection axis to obtain the projections of the vehicle envelope and the obstacle envelope;

[0037] Step S2.3.2, check whether the projections of the vehicle envelope and the obstacle envelope overlap. If not, then determine that there is no collision between the vehicle envelope and the obstacle envelope, and take no collision as the collision detection result, and execute step S2.4; otherwise, there is overlap, and execute step 2.3.3;

[0038] Step S2.3.3, check the edges in the vehicle envelope box, select an edge from the edges that have not been compared and return to execute step S2.3.1; if all the edges of the vehicle envelope box have been checked, then check the edges in the obstacle envelope box, select an edge from the edges that have not been compared and return to execute step S2.3.1; if all the edges of the vehicle envelope box and the obstacle envelope box have been compared and there is no collision, then it is determined that there is a collision between the vehicle envelope box and the obstacle envelope box, and the collision is taken as the collision detection result.

[0039] Optionally, the step S2.4 specifically includes:

[0040] Obtain the collision detection result. If the collision detection result is a collision, that is, it is detected that the vehicle envelope has collided with an obstacle at a certain point on the task path or predicted trajectory, then the distance from the collision point to the current position of the vehicle is used as the safety distance;

[0041] If the collision detection result is no collision, the safety distance is set to the preset value;

[0042] The obtained safety distance and trajectory path are output to the downstream to indicate the risk level of the vehicle.

[0043] Optionally, the step S3 specifically includes:

[0044] Step S3.1, select a moving obstacle based on the information of the perceived obstacle, and calculate the kinetic energy field to obtain the moving obstacle O to any point (x e ,y e ) at:

[0045]

[0046] Among them, v o is the speed of the moving obstacle O, θ o The speed of the moving obstacle O to any point in the driving environment (x e ,y e ), k1, k2, k3 are constant terms, r is the vector distance between the obstacle point and any point in the driving environment, Represents the direction of the field strength, R o is the road condition influencing factor, M o is the virtual mass of the moving obstacle O;

[0047] Step S3.2, select a static obstacle based on the information of the perceived obstacle, calculate the potential energy field, and obtain the potential energy field of the static obstacle Q:

[0048]

[0049] Among them, R Q is the road condition influencing factor, M Q is the virtual mass of the static obstacle Q, k4 and k5 are constant terms respectively;

[0050] Step S3.3, combining the obtained kinetic energy field of the moving obstacle and the potential energy field of the static obstacle to obtain a safety risk score as the safety risk field.

[0051] Optionally, the step S4 specifically includes:

[0052] Step S4.1, calculating the target speed of the vehicle according to the safety distance and the safety risk score;

[0053] Step S4.2, based on the obtained target vehicle speed, the acceleration control amount is obtained by processing through the controller;

[0054] Step S4.3, using throttle calibration, the final throttle and brake openings are determined according to the obtained acceleration control amount, and are output as control instructions to control the driving of the vehicle.

[0055] Optionally, the step S4.1 specifically includes:

[0056] Step S4.1.1, obtain the security risk score threshold T score and safety distance threshold T dis ;

[0057] Step S4.1.2, obtain the target vehicle speed:

[0058]

[0059] Among them, score is the obtained safety risk score, v is the current vehicle speed, dis is the obtained safety distance, and k6 and k7 are constant parameters respectively.

[0060] Optionally, the step S4.2 specifically includes:

[0061] Construct a proportional-integral controller, including the time domain of the proportional-integral controller:

[0062]

[0063] Among them, k P is the proportional term coefficient, k I is the integral term coefficient, e(t′) is the speed error, the target speed minus the actual speed, t′ is the time, τ′ is the differential of time, and e(τ′) is the historical speed error;

[0064] The overall transfer function is:

[0065]

[0066] Where s represents the complex frequency domain unit;

[0067] The target vehicle speed is input into the proportional-integral controller constructed by equation (12) or (16) to obtain the acceleration control variable.

[0068] According to another embodiment of the present invention, an unmanned mining truck safety protection method based on double-layer risk estimation is provided, which reads or calculates the current mission path and future predicted trajectory of the unmanned vehicle respectively, evenly samples points on these two trajectories as detection points, and generates the vehicle envelope with the detection points as the center of the rear axle, and calculates whether the vehicle envelope overlaps with the envelope of the surrounding obstacles. If there is an overlap, it is considered that there is a collision risk; at the same time, based on the perceived obstacle information, the driving safety risk field intensity of the vehicle's current position is calculated. According to the above collision distance and risk score, it is decided whether the vehicle should slow down or emergency brake. The specific steps are as follows:

[0069] The first step is to predict the trajectory of the vehicle: trajectory prediction is performed based on deep learning methods. In order to comprehensively consider the known driving trajectory, road map boundaries, vehicle motion status and vehicle control information, a multimodal trajectory prediction model based on graph convolutional neural network and self-attention mechanism is proposed. The mining area road information is represented by a vector map and encoded by a convolutional neural network. The historical trajectory of the vehicle is encoded based on the self-attention mechanism, and then the fusion network is used to fuse the map features and the trajectory features of the vehicle, and the multimodal prediction trajectory is output. Finally, the selection network is used to calculate the trajectory with the highest confidence.

[0070] Step 2: Collision detection outputs safe distance: Use the vehicle's mission path and predicted trajectory to perform collision detection with the boundaries and obstacles of the surrounding environment. If a collision is detected, the distance from the nearest collision point to the vehicle is the safe distance.

[0071] Step 3: Construct a safety risk field based on the vehicle's surrounding environment and output a safety risk score: The safety field model divides the factors affecting traffic safety into three categories: the risk caused by moving obstacles around the vehicle is called the kinetic field; the risk caused by static obstacles in the environment is called the potential field; the risk caused by the driver's behavior is called the behavioral field; unifying these three fields becomes a unified "driving safety field" model.

[0072] Step 4: Control the vehicle speed based on the safety distance in the second step and the safety risk field output in the third step: First, calculate the target vehicle speed based on the safety distance and safety risk field score, and then use PID to output the longitudinal control quantity.

[0073] Among them, the specific method of the first step: vehicle trajectory prediction is:

[0074] 1) Complete the vectorization of map information and historical trajectory information: Map boundaries, task paths, historical trajectories, etc. can all be regarded as a series of lines. By discretizing these lines into ordered points, the vectorization of image information can be completed.

[0075] 2) Use graph neural networks to encode map information: The connectivity information between mining roads can be stored using a directed graph, and graph neural networks can be used to capture the interaction features between roads.

[0076] 3) Use FPN network to encode historical trajectories: The long-term trends and short-term changes of historical trajectories are important information for trajectory prediction. The FPN network can greatly improve the extraction of multi-scale information without changing the computational complexity of CNN.

[0077] 4) The prediction network processes map features and trajectory features: map features can reflect the overall trend of the vehicle's future trajectory, and historical trajectory features retain the slight differences in the trajectories of different vehicles at different times. These two features are fused and processed to output a multimodal predicted trajectory, from which the trajectory with the highest confidence is extracted.

[0078] Step 2: The specific method of collision detection and safety distance calculation is:

[0079] 1) Generate vehicle envelope: Take the mission path and predicted trajectory as baselines, and generate the envelope of the vehicle traveling on these two paths as the coverage area of ​​the vehicle on the map.

[0080] 2) Obstacle filtering: Filtering out obstacles that are unlikely to cause collision risks can effectively improve the computational efficiency of the program.

[0081] 3) Collision detection: Perform collision detection on the vehicle envelope and the obstacle envelope that may have a collision risk.

[0082] 4) Collision distance output: If it is detected that there is an obstacle around that conflicts with the vehicle envelope, the distance from the obstacle to the vehicle is output.

[0083] Step 3: The specific method for calculating the safety risk field is:

[0084] As mentioned above, the safety risk field is divided into three parts: kinetic field, potential field and behavioral field. Since the vehicle speed on mining roads is generally low and the road conditions are relatively simple, the impact of the behavioral field on vehicle safety is not considered, and only the kinetic field and potential field are discussed. That is:

[0085] E S =E v +E R

[0086] Where E S For driving safety, E v is the kinetic energy field, E R Therefore, it is necessary to calculate the kinetic energy field and potential energy field generated by the surrounding obstacles at the position of the vehicle, and then add them together to get the risk score of the current position of the vehicle, that is, the safety field.

[0087] Step 4: The specific method of controlling the output is:

[0088] The input quantities are the safety distance generated in the second step and the safety risk field generated in the third step. The longitudinal speed of the vehicle is controlled according to these two risk quantification data without considering detours and other actions.

[0089] 1) First, calculate the target speed of the current vehicle based on the collision distance and safety risk score.

[0090] 2) According to the target vehicle speed and the current actual vehicle speed, the speed error is obtained, and then the desired acceleration is output by the PI controller and feedback control. The feedback mode is set to a first-order system.

[0091] 3) Finally, calculate the throttle or brake control amount based on the vehicle's throttle-acceleration calibration table.

[0092] Compared with the prior art, the unmanned mining truck safety protection method based on double-layer risk estimation provided according to the embodiment of the present invention has at least the following beneficial effects:

[0093] 1. The unmanned mining truck safety protection method based on double-layer risk estimation of the present invention calculates the driving risk in real time according to the vehicle driving condition and surrounding environment information, and intelligently decides parking and speed reduction measures according to the size of the risk, thereby realizing active safety protection of unmanned vehicles in mining areas.

[0094] 2. The predicted trajectory is used to calculate the safe distance (the distance between the vehicle and the target object), the surrounding environment is used to perform risk scoring, and two quantitative indicators (safety distance and risk score) are combined to perform a double-layer risk estimation, ensuring the accuracy of risk estimation.

[0095] 3. The future trajectory of the vehicle is estimated based on the vectorized map and the historical trajectory of the vehicle, making full use of the road characteristics of the mining area and the deduction of the future trajectory of the vehicle to ensure the accuracy of the trajectory prediction of the vehicle.

[0096] 4. Use the environmental kinetic energy field and potential energy field to calculate the vehicle driving risk score to ensure reasonable speed planning in complex scenarios such as intersections. BRIEF DESCRIPTION OF THE DRAWINGS

[0097] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. The features and advantages of the present invention can be more clearly understood by referring to the drawings. The drawings are schematic and should not be understood as limiting the present invention in any way. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0098] Figure 1The present invention is a flowchart of an unmanned mining truck safety protection method based on double-layer risk estimation according to an embodiment of the present invention.

[0099] Figure 2 A schematic diagram of a map feature extraction network for an unmanned mining truck safety protection method based on double-layer risk estimation provided according to an embodiment of the present invention.

[0100] Figure 3 A schematic diagram of the FPN network structure of an unmanned mining truck safety protection method based on double-layer risk estimation provided according to an embodiment of the present invention.

[0101] Figure 4 A schematic diagram of collision detection of an unmanned mining truck safety protection method based on double-layer risk estimation provided according to an embodiment of the present invention.

[0102] Figure 5 A schematic diagram of a proportional-integral feedback controller of an unmanned mining truck safety protection method based on double-layer risk estimation provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0103] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0104] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0105] The following describes in detail the unmanned mining truck safety protection method based on double-layer risk estimation provided according to an embodiment of the present invention with reference to the accompanying drawings.

[0106] Explanation of symbols:

[0107] N i The i-th node;

[0108] v i The vector of the i-th node;

[0109] Vector v i Starting point;

[0110] Vector v i The end point;

[0111] P j The jth broken line;

[0112] a i The semantic label of the i-th node;

[0113] Nf i (l) The feature vector of the i-th node in layer l;

[0114] The feature vector representing the neighbor nodes of the previous node;

[0115] Multilayer Perceptron;

[0116] Pooling layer;

[0117] Indicates feature concatenation;

[0118] Pf j The characteristics of the jth broken line;

[0119] The k-th layer features of the m-th node;

[0120] Pf is the feature set of all polylines;

[0121] Pf n The characteristics of the nth polyline;

[0122] k sub-graph feature extraction forward propagation k layers;

[0123] m polyline Pfj j Contains m vectors;

[0124] n broken lines;

[0125] Pf j (l) The l-th layer feature of the j-th broken line;

[0126] A adjacency matrix of polyline features;

[0127] x The x-coordinate set of the historical trajectory;

[0128] x t The x-coordinate of the vehicle at time t;

[0129] y The y coordinate set of the historical trajectory;

[0130] y t The y coordinate of the vehicle at time t;

[0131] h The set of ego-vehicle headings in historical trajectories;

[0132] h t The heading of the vehicle at time t;

[0133] t is the current moment;

[0134] τ time step;

[0135] (x a ,y a ) The coordinates of the path point in the vehicle coordinate system;

[0136] (x1′, y1′) The coordinates of the corner point 1 of the vehicle envelope in the path point coordinate system;

[0137] (x1, y1) The coordinates of the corner point 1 of the vehicle envelope in the vehicle coordinate system;

[0138] The rotation angle from the path point coordinate system to the vehicle coordinate system;

[0139] (x o ,y o ) The position coordinates of the moving obstacle O;

[0140] v o The speed of the moving obstacle O;

[0141] θ o The angle between the velocity direction of the moving obstacle O and any point in the driving environment;

[0142] (x e ,y e ) The position coordinates of any point in the driving environment;

[0143] E v,o The kinetic energy field of the moving obstacle O at any point in the driving environment;

[0144] k1, k2, and k2 are constant terms;

[0145] r is the vector distance between the obstacle point and any point in the driving environment;

[0146] R o Factors affecting road conditions;

[0147] M o The virtual mass of the moving obstacle O;

[0148] (x Q ,y Q ) The coordinates of the static obstacle Q;

[0149] E R,Q The potential energy field of the static obstacle Q at the point in the vehicle envelope;

[0150] k4 and k5 are constant terms;

[0151] R Q Factors affecting road conditions;

[0152] M Q The virtual mass of the static obstacle Q;

[0153] T score Security risk score thresholds;

[0154] T dis Safety distance threshold;

[0155] vCurrent vehicle speed;

[0156] TTC collision time;

[0157] d min The minimum distance the vehicle needs to maintain from the obstacle ahead;

[0158] v target Target vehicle speed;

[0159] score security risk score;

[0160] dis safe distance;

[0161] k6 and k7 are constant terms;

[0162] t′ is time;

[0163] τ′ is the differential of time;

[0164] u(t′) proportional-integral controller in the time domain;

[0165] k P Proportional term coefficient;

[0166] k I Integral term coefficient;

[0167] e(t′) velocity error;

[0168] e(τ′) is the historical velocity error;

[0169] s represents the complex frequency domain unit;

[0170] A(s) frequency domain characteristics of acceleration;

[0171] Frequency domain characteristics of V(s) velocity;

[0172] A des (s) Frequency domain characteristics of target acceleration.

[0173] like Figure 1 As shown, the unmanned mining truck safety protection method based on double-layer risk estimation provided according to an embodiment of the present invention includes the following steps.

[0174] Step S1, predict the trajectory of the self-vehicle and obtain the predicted trajectory. The self-vehicle is an unmanned mining truck currently in operation. In this step, trajectory prediction is performed based on a deep learning method. According to the obtained historical trajectory of the self-vehicle, road map boundaries, vehicle motion status and vehicle control information, a multimodal trajectory prediction model based on a graph convolutional neural network (GNN) and a self-attention mechanism is constructed, and the mining area road information is represented by a vector map, and the mining area road represented by the vector map is encoded by a convolutional neural network. The historical trajectory of the self-vehicle is encoded based on the self-attention mechanism, and then the fusion network is used to fuse the features of the encoded map data and the features of the historical trajectory of the self-vehicle, and the multimodal predicted trajectory is output. Finally, the selection network is used to calculate the trajectory with the highest confidence as the predicted trajectory. Step S1 specifically includes the following steps.

[0175] Step S1.1, obtain the map information of the mining area, represent the elements in the map information with discrete points, and organize the coordinates of the discrete points into vector sets according to the geographical location and functional classification to obtain vectorized map information. Among them, all the elements in the map information are represented by discrete points, such as the road boundary is composed of a curve composed of a series of points, the reference driving route of the vehicle can be composed of a series of points, and the connection method of different roads in the intersection can also be abstracted by a reference trajectory composed of a series of points. According to the above-mentioned geographical location and functional classification, the coordinates of the above-mentioned points are organized into vector sets, and the vectorization of the map information is completed. Optionally, the map information of the mining area can be a bird's-eye view of the mining area.

[0176] Step S1.2, use the graph neural network to encode the vectorized map information to obtain the encoded map information. The vectorized map information consists of nodes and edges, where each node represents the starting point and end point of each vector in the above vector set. In order to distinguish different classifications (such as road boundary points, reference trajectory points, etc.), the classification label of the node is also saved in each node, so as to ensure that the semantic information can also be correctly encoded. For the connectivity between nodes, a hierarchical graph architecture is used to ensure the connection of their contextual information. For nodes with the same semantic information and the same group, they are connected to the same subgraph, and information exchange is ensured between different subgraphs through full connection. Specifically, for a node N i , assuming that its associated vector v i Belongs to the polyline P j , and the semantic label is a i , then the content of the node is:

[0177]

[0178] in and The vector v iThe starting point and end point of the vehicle can be expressed in the form of (x, y, z), where i represents the node number and j represents the line number. At the same time, in order to facilitate calculation, the UTM (Universal Transverse Mercator Grid System) coordinates recorded on the map are converted into a coordinate system centered on the current vehicle position. Figure 2 , the forward propagation of subgraph encoding is defined as:

[0179]

[0180] Among them, Nf i (l) refers to the feature vector of the i-th node in the l-th layer, Represents the feature vector of the neighboring nodes of the current node, It is a multi-layer perceptron (i.e., MLP network) that extracts features from the current node. is the pooling layer (i.e., pooling network), which is used to process the features of the neighboring nodes of the current node, and finally Concatenate the features of the current node with the features of the neighboring nodes. Get the features of each polyline in the map:

[0181]

[0182] Pf={Pf0,Pf1,…,Pf n} (4)

[0183] Among them, Pf j represents the characteristics of the j-th polyline, represents the feature vector of the mth node in the kth layer, Pf n represents the feature of the nth polyline, k represents the k-th layer of forward propagation of sub-graph feature extraction, and m represents the polyline Pf j Contains m vectors, n means there are n broken lines, and Pf represents the feature set of all broken lines, that is, map features. Using graph convolutional neural networks, further feature extraction is performed on the regional features after splicing the node features. The forward propagation method is:

[0184] Pf j (l+1) =GNN(Pf j (l) ,A) (5)

[0185] Among them, Pf j (l)represents the lth layer feature of the jth polyline, and A is the adjacency matrix of the polyline feature, which can introduce the connectivity information between roads on a larger scale. So far, by concatenating the features of each polyline at each layer, the encoding of the map information is completed, and the encoded map information is obtained.

[0186] Step S1.3, using a history trajectory encoding network including an FPN network (feature pyramid network) to encode the vehicle history trajectory to obtain the encoded vehicle history trajectory. Step S1.3 specifically includes the following steps.

[0187] Step S1.3.1, obtain the historical trajectory of the ego vehicle, vectorize it, and obtain the vectorized historical trajectory of the ego vehicle. The historical trajectory of the ego vehicle can be represented by a series of points with headings, which are vectorized:

[0188]

[0189] Among them, x represents the x-coordinate set of the historical trajectory, x t represents the x-coordinate of the vehicle at time t, y represents the y-coordinate set of the historical trajectory, and y t represents the y coordinate of the ego vehicle at time t, h represents the set of ego vehicle headings in the historical trajectory, and h t represents the heading of the vehicle at time t, t represents the current time, τ represents the time step, and t-τ represents the historical time.

[0190] Step S1.3.2, the vector of the quantized historical trajectory of the vehicle obtained above is used as the input of the historical trajectory encoding network, and CNN is used for training to obtain the features extracted by CNN, and then the features extracted by CNN are subjected to multi-scale feature extraction by FPN network to obtain the encoded historical trajectory of the vehicle. Among them, the use of FPN network for feature extraction can ensure that more multi-scale features of the historical trajectory are extracted, while minimizing the increase in computational complexity.

[0191] like Figure 3As shown in the figure, the FPN network consists of three parts, a bottom-up network and a top-down network. The two networks are connected horizontally for data transmission. For the bottom-up network, three residual blocks are used. Each residual block includes three 1D convolutional networks. After each convolution, the ReLU function (Rectified Linear Unit) is used for activation. The step size of the last convolution kernel in the last two residual blocks is set to 2 to ensure that the output of each residual block is half the output scale of the previous residual block, so as to extract smaller-scale features. For the top-down network, the high-level feature map result is upsampled by 2 times using the interpolation method to ensure that the feature tensor scale is the same as that of the bottom-up network output. Then, the upsampled mapping is merged with the corresponding bottom-up mapping by element-wise addition to obtain the result of the mapping merger. Then, the result of the mapping merger is processed by the convolution block, and the output is used as the input of the next layer of the top-down network. The result of the last layer of the top-down network is the final output of the FPN network. The convolutional block can include a 1D convolutional network, a linear network, and a ReLU activation function to reduce the aliasing effect of upsampling. The lateral connection can use a 1D convolutional network.

[0192] Step S1.4, the prediction network processes the features of the encoded map information and the features of the encoded historical trajectory of the ego vehicle to obtain the predicted trajectory. Since the ego vehicle is an unmanned mining truck, it is believed that in the absence of external intervention and safety procedures, the future operation state of the ego vehicle is only related to the task path, and external vehicles have no effect on the control output of the ego vehicle. The prediction network fuses the features of the encoded map information and the features of the encoded historical trajectory of the ego vehicle and outputs them to predict k possible future trajectories.

[0193] Step S1.5, score the k future trajectories output by the prediction network through the scoring network, and select the trajectory with the highest possibility as the predicted trajectory.

[0194] Step S2, obtain the information of the task path and perceived obstacles, use the task path and predicted trajectory to perform collision detection and safe distance calculation with the perceived obstacles, and obtain the safe distance. Use the task path and predicted trajectory of the vehicle to perform collision detection with the boundaries and obstacles of the surrounding environment. If a collision is detected, the distance from the nearest collision point to the vehicle is the safe distance. The task path is issued by the scheduling platform in the unmanned driving system and can be received using the TCP protocol. The information of the perceived obstacles is issued by the perception module in the unmanned driving system and received using the TCP protocol.

[0195] Step S2.1, such as Figure 4As shown in the figure, the predicted trajectory and task path are obtained, and the vehicle envelope of the vehicle is generated. The trajectory points generated by the predicted trajectory and the task path points read record the x, y coordinates and heading information of the path points. These path points are used as the rear axle center to generate the vehicle envelope. The specific process of generating the vehicle envelope includes the following steps.

[0196] Step S2.1.1, obtain the predicted trajectory and task path, traverse the path points of the measured trajectory and task path, use the vehicle speed as the basis for envelope expansion, and generate the vehicle envelope frame of the path point coordinate system.

[0197] Step S2.1.2, based on the relationship between the path point and the vehicle position, convert the coordinates of the vehicle envelope generated in step S2.2.1 from the path point coordinate system to the vehicle coordinate system. The specific coordinate conversion method is: take the coordinates of the path point in the vehicle coordinate system as (x a ,y a ), the coordinates of the corner point 1 of the vehicle envelope in the path point coordinate system are (x1′, y1′), and the rotation angle from the path point coordinate system to the vehicle coordinate system is The coordinates of the corner point 1 of the vehicle envelope after rotation in the vehicle coordinate system are (x1, y1), so the coordinate transformation formula is:

[0198]

[0199] Step S2.1.3, repeat the above step S2.1.2, calculate the coordinates of each corner point of the vehicle envelope in the vehicle coordinate system, and obtain the vehicle envelope converted from the path point coordinate system to the vehicle coordinate system.

[0200] Step S2.2, obtain the information of perceived obstacles, filter the obstacles, remove obstacles that are far away and obstacles without collision risks, obtain obstacles to be detected, and process the information of obstacles to be detected to obtain the obstacle envelope. In order to improve the efficiency of program operation, no collision detection is performed for obstacles that are obviously not risky or obstacles that are far away. Obstacles without collision risks may include, for example, filtering out obstacles behind the rear of the vehicle when moving forward, and filtering out obstacles in front of the front of the vehicle when reversing. Obstacles without collision risks may also include, for each trajectory point, first wrapping the vehicle at that point with a circle. If all corner points of the obstacle are outside the circle, it is considered that this obstacle has no risk of conflict with the trajectory point and is directly skipped.

[0201] Step S2.3, obtain the vehicle envelope and the obstacle envelope, perform collision detection, and obtain the collision detection result. Use the separating axis theorem for collision detection. The vehicle envelope and the obstacle envelope are both convex polygons. For these two convex polygons, if there is no collision between them, there will always be a straight line that can separate the two objects. This straight line is called the separating line. That is, the projections of the two polygons on the separating axis (Separating axis) perpendicular to this separating line (Separating line) do not overlap. This step S2.3 specifically includes the following steps.

[0202] Step S2.3.1, obtain the vehicle envelope box, take out an edge from the vehicle envelope box, and obtain its normal vector. The normal vector is used as a projection axis. Each edge of the obstacle envelope box and the vehicle envelope box is obtained in a loop, and they are projected onto the projection axis to obtain the projections of the vehicle envelope box and the obstacle envelope box.

[0203] Step S2.3.2, check whether the projections of the vehicle envelope frame and the obstacle envelope frame overlap. If not, determine that there is no collision between the vehicle envelope frame and the obstacle envelope frame. Take no collision as the collision detection result and execute step S2.4. If there is overlap, execute step 2.3.3.

[0204] Step S2.3.3, check the edges in the vehicle envelope box, select an edge from the edges that have not been compared and return to execute step S2.3.1; if all the edges of the vehicle envelope box have been checked, then check the edges in the obstacle envelope box, select an edge from the edges that have not been compared and return to execute step S2.3.1; if all the edges of the vehicle envelope box and the obstacle envelope box have been compared, and the above-mentioned no collision situation still does not exist, then it is determined that there is a collision between the vehicle envelope box and the obstacle envelope box, and the collision is taken as the collision detection result.

[0205] Step S2.4, according to the collision detection result, obtain and output the safe distance. If the collision detection result is a collision, that is, it is detected that the vehicle envelope has collided with an obstacle at a certain point on the task path or predicted trajectory, then the distance from the point where the collision occurred to the current position of the vehicle is used as the safe distance; if the collision detection result is no collision, the safe distance is set to a preset value, and the obtained safe distance and trajectory path are output to the downstream to indicate the risk level of the vehicle. As an example, the preset value of the above safety distance can be set to, for example, 100 meters.

[0206] Step S3, based on the information of perceived obstacles, calculate the safety risk field and obtain a safety risk score. The safety risk field is constructed according to the environment around the vehicle, and the safety risk score is output. In this implementation, the safety field model divides the factors affecting traffic safety into two categories: the risk generated by moving obstacles around the vehicle is called the kinetic field; the risk generated by static obstacles in the environment is called the potential field; unifying these two fields becomes a unified model of "driving safety field". This step S3 specifically includes the following steps.

[0207] Step S3.1, select a moving obstacle based on the information of the perceived obstacle, and calculate the kinetic field to obtain the kinetic field of the moving obstacle. The kinetic field is mainly determined by the moving obstacles in the vehicle driving environment. If a moving obstacle O, its position coordinates are (x o ,y o ), with a velocity of v o , calculate the moving obstacle to any point (x e ,y e ) , if the speed direction of the moving obstacle reaches this point (x e ,y e ) is the angle θ o , the kinetic energy field of the moving obstacle is:

[0208]

[0209] Among them, k1, k2, and k3 are constant terms. k1 evenly affects the intensity of the entire kinetic energy field. k2 represents the magnitude of the impact of the distance from the obstacle on the risk. k3 represents the magnitude of the impact of the speed of the obstacle on the risk field. r is the vector distance between the obstacle point and any point in the driving environment. Represents the direction of the field strength. The exp function can reflect the increase in risk as the speed of the obstacle increases exponentially, and ensure that the field strength is always positive regardless of the angle. o The magnitude of the representative field strength is related to the speed direction of the obstacle. The more it matches the speed direction of the obstacle, the higher the risk. o is the road condition influencing factor, M o is the virtual mass of the moving obstacle O, R o It is related to road conditions, slope, friction coefficient, etc. o It is related to the type, mass, size, etc. of the motion obstacle.

[0210] Step S3.2, select a static obstacle based on the information of the perceived obstacle, calculate the potential field, and obtain the potential field of the static obstacle. The potential field is mainly determined by the static obstacles in the environment. If a static obstacle Q, its coordinates are (x Q ,y Q), the static obstacle is located at any point (x e ,y e ) is:

[0211]

[0212] Compared with the kinetic energy field formula, the influence of obstacle speed is reduced. Q 、M Q are the road condition influencing factor and the virtual mass of the static obstacle Q, R Q It is related to road conditions, slope, friction coefficient, etc. Q It is related to the type, mass, size, etc. of the static obstacle. k4 and k5 are constant terms.

[0213] Step S3.3, the obtained kinetic energy field of the moving obstacle and the potential energy field of the static obstacle are combined to obtain a safety risk score as the safety risk field. The driving risk at the position of the vehicle is calculated using the above formulas (8) and (9) to obtain a safety risk score.

[0214] Step S4, based on the obtained safety distance and safety risk field, the control amount of the vehicle speed is calculated, and the control instruction is obtained and output to control the driving of the vehicle. First, the target vehicle speed is calculated according to the safety distance and the safety risk score, and then the longitudinal control amount is output using PID. This step S4 specifically includes the following steps.

[0215] Step S4.1, calculating the target speed of the vehicle according to the safety distance and the safety risk score. This step specifically includes the following steps.

[0216] Step S4.1.1, obtain the security risk score threshold T score and safety distance threshold T dis For both the safety distance and safety risk scores, an emergency stop threshold needs to be set. If the risk is less than the threshold, only the vehicle speed is limited. If the risk is higher than the threshold, it is considered very dangerous and needs to be stopped directly.

[0217] Among them, the security risk score threshold T score , can be set based on experience. Safety distance threshold T dis Can be set to be related to TTC (Time To Collision):

[0218] T dis =max(v·TTC,d min ) (9)

[0219] Where v is the current vehicle speed, TTC is the collision time, which can be set to 3s to 4s, and d minThe minimum distance that the vehicle needs to maintain from the obstacle in front.

[0220] Step S4.1.2, obtain the target vehicle speed:

[0221]

[0222] Wherein, score is the safety risk score obtained in step S3, v is the current vehicle speed, dis is the safety distance obtained in step S2, k6 and k7 are constant parameters. According to the calculation results of the safety distance and the safety risk score, the minimum value is taken as the target vehicle speed.

[0223] Step S4.2, based on the obtained target vehicle speed, the acceleration control amount is obtained through the controller. In this embodiment, the controller is a PID (proportional-integral-derivative control) controller. The PID controller includes two layers, one is a PI (proportional-integral) controller that calculates the expected acceleration from the target vehicle speed, and the other is a controller that converts the acceleration into throttle and brake control information. Specifically, the following steps are included.

[0224] Construct a proportional-integral controller.

[0225] The time domain formula of the proportional-integral controller is:

[0226]

[0227] Among them, k P is the proportional term coefficient, k I is the integral term coefficient, e(t′) is the speed error, that is, the target speed minus the actual speed, t′ is the time, e(τ′) is the historical speed error, and τ′ is the differential of time;

[0228] The transfer function of the controller is therefore:

[0229]

[0230] Where s represents the complex frequency domain unit;

[0231] Use the first-order system to simulate the acceleration system response process, which satisfies:

[0232]

[0233] Among them, A(s) represents the frequency domain characteristics of acceleration, V(s) represents the frequency domain characteristics of velocity, and A des (s) represents the frequency domain characteristics of target acceleration;

[0234] Then the feedback transfer function is:

[0235]

[0236] If Figure 5 , the overall transfer function is:

[0237]

[0238] The above transfer function is discretized to meet the needs of program calculation, and the target vehicle speed is input into the proportional-integral controller constructed by equation (12) or (16) to obtain the acceleration control value.

[0239] Step S4.3, using throttle calibration, the final throttle and brake openings are determined according to the obtained acceleration control amount, and are output as control instructions to control the driving of the vehicle.

[0240] All the above optional technical solutions can be arbitrarily combined to form optional embodiments of the present application, which will not be described one by one here.

[0241] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0242] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. The unmanned mining truck safety protection method based on double-layer risk estimation is characterized by: The following steps are involved: Step S1, predicting the vehicle trajectory to obtain a predicted trajectory; Step S2, obtaining information of the task path and perceived obstacles, performing collision detection and safety distance calculation on the task path and predicted trajectory and perceived obstacles to obtain a safety distance; Step S3, calculating the safety risk field based on the acquired information of the perceived obstacle to obtain a safety risk score; Step S4, calculating the control amount of the vehicle speed based on the obtained safety distance and safety risk field, obtaining and outputting a control instruction, and controlling the driving of the vehicle; Wherein, the step S1 specifically includes the following steps: Step S1.1, vectorizing the map information to obtain the map information of the mining area, representing the elements in the map information as discrete points, and composing the coordinates of the discrete points into vector sets according to the geographical location and functional classification, to obtain the vectorized map information; Step S1.2, using a graph neural network to encode the vectorized map information to obtain the encoded map information, wherein the vectorized map information is composed of nodes and edges, wherein each node represents the starting point and the end point of each vector in the vector set; Step S1.3, using a history trajectory encoding network including an FPN network to encode the vehicle history trajectory to obtain an encoded vehicle history trajectory; Step S1.4, processing the features of the encoded map information and the features of the encoded historical trajectory of the vehicle through the prediction network to obtain k future trajectories; Step S1.5, score the k future trajectories output by the prediction network and select the trajectory with the highest probability as the predicted trajectory.

2. The unmanned mining truck safety protection method based on double-layer risk estimation according to claim 1 is characterized in that: The step S1.3 specifically includes the following steps: Step S1.3.1, obtain the historical trajectory of the ego vehicle, vectorize it, and obtain the vectorized historical trajectory of the ego vehicle: x=[x t ,x t-τ ,x t-2τ ,…] T (6a) y=[y t ,y t-τ ,y t-2τ ,…] T (6b) h=[h t ,h t-τ ,h t-2τ ,…] T (6c) Among them, x represents the x-coordinate set of the historical trajectory, x t represents the x-coordinate of the vehicle at time t, y represents the y-coordinate set of the historical trajectory, and y t represents the y coordinate of the ego vehicle at time t, h represents the set of ego vehicle headings in the historical trajectory, and h t represents the heading of the vehicle at time t, t represents the current time, and τ represents the time step; In step S1.3.2, the vector of the quantized historical trajectory of the ego vehicle obtained above is used as the input of the historical trajectory encoding network, and CNN is used for training to obtain the features extracted by CNN. The features extracted by CNN are then subjected to multi-scale feature extraction using the FPN network to obtain the encoded historical trajectory of the ego vehicle.

3. The unmanned mining truck safety protection method based on double-layer risk estimation according to claim 2 is characterized in that: The step S2 specifically includes the following steps: Step S2.1, obtaining the predicted trajectory and task path, and generating the vehicle envelope of the ego vehicle; Step S2.2, obtaining information about perceived obstacles, filtering the obstacles, removing obstacles that are far away and obstacles that have no collision risk, obtaining obstacles to be detected, and processing the information about the obstacles to be detected to obtain an obstacle envelope; Step S2.3, obtaining the vehicle envelope and the obstacle envelope, performing collision detection, and obtaining the collision detection result; Step S2.4, based on the collision detection result, obtain the safe distance and output it.

4. The unmanned mining truck safety protection method based on double-layer risk estimation according to claim 3 is characterized in that: The step S2.1 specifically includes the following steps: Step S2.1.1, obtain the predicted trajectory and task path, traverse the path points of the measured trajectory and task path, use the vehicle speed as the basis for envelope expansion, and generate the vehicle envelope frame of the path point coordinate system; Step S2.1.2, based on the relationship between the path point and the vehicle position, the coordinates of the vehicle envelope are converted from the path point coordinate system to the vehicle coordinate system: Among them, (x a ,y a ) is the coordinate of the path point in the vehicle coordinate system, (x1′, y1′) is the coordinate of the corner point 1 of the vehicle envelope in the path point coordinate system, is the rotation angle from the path point coordinate system to the vehicle coordinate system, (x1, y1) is the coordinate of the corner point 1 of the vehicle envelope after rotation in the vehicle coordinate system; Step S2.1.3, repeat the above step S2.1.2, calculate the coordinates of each corner point of the vehicle envelope frame in the vehicle coordinate system, obtain the vehicle envelope frame converted from the path point coordinate system to the vehicle envelope frame in the vehicle coordinate system, and output the converted vehicle envelope frame in the vehicle coordinate system as the vehicle envelope frame.

5. The unmanned mining truck safety protection method based on double-layer risk estimation according to claim 4 is characterized in that: The step S2.3 specifically includes: Step S2.3.1, obtain the vehicle envelope, take out an edge from the vehicle envelope, and obtain its normal vector, the normal vector is used as a projection axis, and each edge of the obstacle envelope and the vehicle envelope is obtained in a loop, and they are projected onto the projection axis to obtain the projections of the vehicle envelope and the obstacle envelope; Step S2.3.2, check whether the projections of the vehicle envelope and the obstacle envelope overlap. If not, then determine that there is no collision between the vehicle envelope and the obstacle envelope, and take no collision as the collision detection result, and execute step S2.4; otherwise, there is overlap, and execute step 2.3.3; Step S2.3.3, check the edges in the vehicle envelope box, select an edge from the edges that have not been compared and return to execute step S2.3.1; if all the edges of the vehicle envelope box have been checked, then check the edges in the obstacle envelope box, select an edge from the edges that have not been compared and return to execute step S2.3.1; if all the edges of the vehicle envelope box and the obstacle envelope box have been compared and there is no collision, then it is determined that there is a collision between the vehicle envelope box and the obstacle envelope box, and the collision is taken as the collision detection result.

6. The unmanned mining truck safety protection method based on double-layer risk estimation according to claim 5 is characterized in that: The step S2.4 specifically includes: Obtain the collision detection result. If the collision detection result is a collision, that is, it is detected that the vehicle envelope has collided with an obstacle at a certain point on the task path or predicted trajectory, then the distance from the collision point to the current position of the vehicle is used as the safety distance; If the collision detection result is no collision, the safety distance is set to the preset value; The obtained safety distance and trajectory path are output to the downstream to indicate the risk level of the vehicle.

7. The unmanned mining truck safety protection method based on double-layer risk estimation according to claim 6 is characterized in that: The step S3 specifically includes: Step S3.1, select a moving obstacle based on the information of the perceived obstacle, and calculate the kinetic energy field to obtain the moving obstacle O to any point (x e ,y e ) at: Among them, v o is the speed of the moving obstacle O, θ o The speed of the moving obstacle O to any point in the driving environment (x e ,y e ), k1, k2, k3 are constant terms, r is the vector distance between the obstacle point and any point in the driving environment, Represents the direction of the field strength, R o is the road condition influencing factor, M o is the virtual mass of the moving obstacle O; Step S3.2, select a static obstacle based on the information of the perceived obstacle, calculate the potential energy field, and obtain the potential energy field of the static obstacle Q: Among them, R Q is the road condition influencing factor, M Q is the virtual mass of the static obstacle Q, k4 and k5 are constant terms respectively; Step S3.3, combining the obtained kinetic energy field of the moving obstacle and the potential energy field of the static obstacle to obtain a safety risk score as the safety risk field.

8. The unmanned mining truck safety protection method based on double-layer risk estimation according to claim 7 is characterized in that: The step S4 specifically includes: Step S4.1, calculating the target speed of the vehicle according to the safety distance and the safety risk score; Step S4.2, based on the obtained target vehicle speed, the acceleration control amount is obtained by processing through the controller; Step S4.3, using throttle calibration, the final throttle and brake openings are determined according to the obtained acceleration control amount, and are output as control instructions to control the driving of the vehicle.

9. The unmanned mining truck safety protection method based on double-layer risk estimation according to claim 8 is characterized in that: The step S4.1 specifically includes: Step S4.1.1, obtain the security risk score threshold T score and safety distance threshold T dis ; Step S4.1.2, obtain the target vehicle speed: Among them, score is the obtained safety risk score, v is the current vehicle speed, dis is the obtained safety distance, and k6 and k7 are constant parameters respectively.

10. The unmanned mining truck safety protection method based on double-layer risk estimation according to claim 9 is characterized in that: The step S4.2 specifically includes: Construct a proportional-integral controller, including the time domain of the proportional-integral controller: Among them, k P is the proportional term coefficient, k I is the integral term coefficient, e(t′) is the speed error, the target speed minus the actual speed, t′ is the time, τ′ is the differential of time, and e(τ′) is the historical speed error; The overall transfer function is: Where s represents the complex frequency domain unit; The target vehicle speed is input into the proportional-integral controller constructed by equation (12) or (16) to obtain the acceleration control variable.

Images