Heavy-load train group cooperative control method, device, equipment and medium

Abstract

The invention relates to the technical field of rail transit, in particular to a heavy-load train group cooperative control method, device and equipment and a medium. The safety of the heavy-load train cooperative control environment is judged according to the segmented data of the laser radar, the speed of the detected object and the distance to the solid-state laser radar, and when the solid-state laser radar detects that the cooperative control environment is safe, that is, no other object exists except the heavy-load train in front within the cooperative control distance range, the safety of the heavy-load train is ensured. The solid-state laser radar can obtain the actual distance and speed of the front cooperative control heavy-load train and is used for establishing a heavy-load train cooperative control adjusting system with the distance error and speed of the heavy-load train as input and the speed adjusting gain of the heavy-load train as output. The adaptive fuzzy PID controller with the minimum error is optimized based on the improved particle swarm optimization algorithm, the distance of cooperative control of the heavy haul train is shortened, the railway transportation capacity is improved, and the defects of a traditional control system in a dynamic environment are effectively overcome.

Description

Technical Field

[0001] The present disclosure relates to the field of rail transit technology, and in particular to a method, device, equipment and medium for coordinated control of a heavy-load train group. Background Art

[0002] With the rapid development of railway transportation, especially heavy-haul trains, coordinated control technology for train groups has become crucial for ensuring efficient and safe transportation. Traditional train control systems face challenges in addressing group coordinated control, including environmental perception accuracy, data processing efficiency, control algorithm stability, and adaptability to dynamic environments. This makes it difficult to maintain precise distance and speed regulation in complex environments. Currently, accurate distance measurement has become a key factor in virtual coupled mobility and vehicle-to-vehicle (V2V) systems.

[0003] Existing technologies do not fully consider the dynamic conditions, train composition, speed and distance changes required for virtual coupled trains, especially during maneuvering and V2V coupling operations in complex environments. Existing systems have difficulty in coping with precise vehicle spacing and dynamic adaptability. Summary of the Invention

[0004] In response to the above problems, the present disclosure provides a method, device, equipment and medium for collaborative control of a heavy-load train group, which are used to shorten the collaborative control distance of heavy-load trains and improve railway transportation capacity.

[0005] In a first aspect, a method for coordinated control of a heavy-load train group is provided, the method comprising:

[0006] The following vehicle in the heavy-load train group obtains real-time segmented data of the heavy-load train group's surrounding environment based on the lidar;

[0007] The real-time segmented data is preprocessed using a dynamic programming method to obtain preprocessed data;

[0008] The fuzzy clustering algorithm based on dynamic threshold is used to perform fuzzy processing on the pre-processed data, extract the objects of interest, and obtain the distances between the following vehicle and all objects of interest within the preset distance range;

[0009] When there is only a heavy-load train ahead within the preset distance range, an adaptive fuzzy PID controller that minimizes the error is used to adjust the distance and speed between the following vehicle and the heavy-load train ahead based on the distance between the following vehicle and the heavy-load train ahead.

[0010] Further, real-time segmented data includes:

[0011] The relative distance between the object of interest and the heavy-haul train equipped with the LiDAR, the speed of the object of interest relative to the heavy-haul train equipped with the LiDAR, and the segmented distance of the solid-state LiDAR;

[0012] Among them, the relative distance from the object of interest to the heavy-load train equipped with the laser radar and the speed of the object of interest relative to the heavy-load train equipped with the laser radar are directly measured;

[0013] The segment distance D of the solid-state laser radar is calculated by the following formula:

[0014]

[0015] Where, D represents the segment distance of the laser radar, d i and d i+1 They represent the distances of the two laser beams emitted by the solid-state laser radar, θ represents the angle between the two laser beams emitted by the solid-state laser radar, cos(θ) represents the cosine value of the angle θ, and i represents the number of segments of the laser radar.

[0016] Furthermore, the real-time segmented data is preprocessed using a dynamic programming method, including:

[0017] The dynamic programming method is used to preprocess the collected laser radar segmentation data and generate the corresponding binary attribution degree matrix as preprocessing data;

[0018] The dynamic programming formula is as follows:

[0019]

[0020] Where D=[D1,D2,...,D i ] represents the distance data between solid-state laser radar segments, i represents the number of laser radar segments, D i represents the distance data of the i-th segment of the lidar, diff[i] represents the difference between the i-th adjacent distance data points, diff[m] represents the difference between the m-th adjacent distance data points, j represents the starting index of the lidar distance data segment, dp[i] represents the optimal grouping cost of the first i segments, dp[j] represents the optimal grouping cost of the first j segments, cost(j+1,i) represents the cost of the first i segments from segment D j+1 to D i The cost and j * represents the optimal segmentation point with the minimum total cost, P represents the set of all optimal cutting points, j1 represents the optimal cutting point of the first segment group, and j S-1 represents the optimal cutting point of the S-1 segment group, S represents the optimal number of segment groups, represents the oth optimal cutting point, j o+1 represents the starting position of the o+1 segment; |P| represents the absolute value of P; argmin represents the variable value that minimizes the expression, that is, the value of the optimal segmentation point position; the binary attribution matrix formula is as follows:

[0021]

[0022] Among them, M p,q represents the qth data point in the pth segment, that is, the element in the data matrix, p represents the current pth segment, q represents the qth element in the current data point sequence, D q Represents an element in the data sequence, that is, the qth data point in the input data, B p,q represents the qth binary membership matrix element in the pth segment.

[0023] Furthermore, the adaptive fuzzy clustering algorithm based on dynamic threshold is formulated as follows:

[0024]

[0025] Among them, σ represents the σth iteration, l represents the lth column of the pre-grouping matrix, that is, the lth segment, r represents the rth row of the pre-grouping matrix, that is, the rth segment group, represents the center position of the rth segment group at the σth iteration, represents the degree of belonging of the lth segment in the rth segment group after the σth iteration of pre-grouping, D l Indicates the laser radar segment measurement distance of the current segment l, x lr represents the laser radar segment distance of the lth segment in the rth segment group of the pre-grouping matrix, i represents the number of laser radar segments, S represents the optimal number of segment groups, represents the belonging degree corresponding to the lth segment of the rth segment group of the σth iteration binary matrix, v represents the fuzzy index, Δc r represents the change value of the center of the rth segment group after t iterations, ò0 represents the initial threshold, σ Indicates the dynamic change of the convergence threshold of the σth iteration, error σ Indicates the error or change value of the σth iteration, error σ It is used to measure the state of the algorithm in the σth iteration, and α represents the adjustment factor.

[0026] Furthermore, a fuzzy clustering algorithm based on dynamic threshold is used to fuzzy process the preprocessed data and extract objects of interest, including:

[0027] The pre-processed data is fuzzy processed by a fuzzy clustering algorithm based on dynamic threshold, and a fuzzy attribution degree matrix is ​​obtained to represent the membership relationship between the object and the laser beam.

[0028] The number of objects within the laser radar detection field of view and the laser radar segment group corresponding to the object are determined according to the fuzzy attribution matrix, and then the distance of the object is obtained according to the laser radar ranging formula;

[0029] The cluster closest to the lidar sensor is considered the object of interest.

[0030] Furthermore, the distance and speed between the following vehicle and the heavy-load train ahead are adjusted, including:

[0031] The distance error value is calculated based on the distance between the following vehicle and the heavy-load train ahead and the expected cooperative distance. The speed of the heavy-load train is cooperatively controlled and adjusted according to the distance error value. The cooperative control adjustment is expressed as:

[0032]

[0033] Among them, D lateral,min Represents the three-dimensional distance of the object closest to the lidar, d ref Denotes the expected collaborative distance, L delay Indicates the running distance of the following train when the communication with the adjacent train is delayed, L break(r) Indicates the maximum braking distance of the following vehicle, L break(p) Indicates the maximum braking distance of the vehicle in front, L s represents the safety margin, max() represents the maximum value, t represents the time, e(t) represents the error between the actual distance of the train at time t and the target distance, d actual (t) represents the actual coordinated distance of the train at time t, Represents the rate of change of error e(t), T sample represents the sampling period, e(tT sample ) represents the error at the last sampling moment, v(t) represents the speed of the train at time t, v ref (t) represents the expected coordination speed, v ref (t) is obtained from the driving map, K represents the gain coefficient, which is used to control the response strength of speed to distance error, S k Indicates the idling stopping distance, S e represents the common braking distance of the train operation monitoring and recording device LKJ, v0 represents the initial braking speed, λ represents the braking calculation coefficient, ν h represents the train transfer braking rate, φ h Indicates the friction coefficient of the brake shoe, β c represents common braking parameters, ω0 represents the basic unit resistance of the train, t0 represents the driver's reaction time, v m Indicates the final speed of the heavy-load train after deceleration.

[0034] Furthermore, the adaptive fuzzy PID controller that minimizes the error is formulated as follows:

[0035]

[0036] Where u(t) represents the output signal of the controller at time t, e(t) represents the error between the actual distance of the train and the target distance at time t, and t represents the time. Indicates the rate of change of error e(t), K υ Represents the proportional gain parameter, K β Indicates the integral gain parameter, K d Denotes the differential gain parameter, K υ (t) represents the proportional gain at time t, K d (t) represents the differential gain at time t, K β (t) represents the integral gain at time t, K υ0 Indicates the initial proportional gain, K β0 Indicates the initial integral gain, K d0 Denotes the initial differential gain, ΔK υ Indicates the proportional gain adjustment, ΔK d Indicates the differential gain adjustment, ΔK β Indicates the integral gain adjustment, ΔK υ , ΔK d , ΔK β The three adjustment quantities are obtained through fuzzy reasoning.

[0037] Furthermore, it also includes:

[0038] An improved particle swarm optimization algorithm is used to dynamically adjust the initial parameters of the adaptive fuzzy PID controller with minimum error.

[0039] Among them, the particle update strategy of the improved particle swarm optimization algorithm is as follows:

[0040]

[0041] Among them, H n represents the position of the nth particle in the current search space, K υ Represents the proportional gain parameter, K β Indicates the integral gain parameter, K d represents the differential gain parameter, T represents the time step range of error calculation, Q(z) represents the error at the zth iteration, J represents the objective function to be minimized, and v n (z) represents the velocity of the nth particle at the zth iteration, w represents the inertia weight term, p n represents the current optimal position found by the nth particle, r1 and r2 represent random numbers between [0,1], c1 represents the first learning factor, c2 represents the second learning factor, g represents the global optimal position found among all particles, and x n (z) The position of particle n at the zth iteration, w max and w minRepresent the maximum and minimum values ​​of the inertia weight, z represents the number of iterations, and z max Maximum number of iterations, Represent the maximum values ​​of c1 and c2 respectively, Represent the minimum values ​​of c1 and c2 respectively.

[0042] In a second aspect, a heavy-load train group cooperative control device includes: a segmented data acquisition unit, a pre-processing unit, an object of interest extraction unit, and a cooperative control unit;

[0043] A segmented data acquisition unit is used for the following vehicle in the heavy-load train group to obtain real-time segmented data of the environment surrounding the heavy-load train group based on the laser radar;

[0044] A preprocessing unit, configured to preprocess the real-time segmented data using a dynamic programming method to obtain preprocessed data;

[0045] An object of interest extraction unit is used to perform fuzzy processing on the pre-processed data using a fuzzy clustering algorithm based on a dynamic threshold, extract the object of interest, and obtain the distance between the following vehicle and all objects of interest within a preset distance range;

[0046] The collaborative control unit is used to adjust the distance and speed between the following vehicle and the heavy-loaded train in front by using an adaptive fuzzy PID controller that minimizes the error based on the distance between the following vehicle and the heavy-loaded train in front when there is only a heavy-loaded train in front within a preset distance range.

[0047] Further, real-time segmented data includes:

[0048] The distance from the object of interest to the lidar, the speed of the object of interest relative to the heavy-load train carrying the lidar, and the segment distance for solid-state lidars;

[0049] The distance from the object of interest to the lidar and the speed of the object of interest relative to the heavy-load train equipped with the lidar are directly measured by the solid-state lidar.

[0050] The segment distance D of the solid-state laser radar is calculated as follows:

[0051]

[0052] Where, D represents the segment distance of the laser radar, d i and d i+1 They represent the distances of the two laser beams emitted by the solid-state laser radar, θ represents the angle between the two laser beams emitted by the solid-state laser radar, cos(θ) represents the cosine value of the angle θ, and i represents the number of segments of the laser radar.

[0053] Furthermore, the pre-processing unit is specifically used to:

[0054] The dynamic programming method is used to preprocess the collected laser radar segmentation data and generate the corresponding binary attribution degree matrix as preprocessing data;

[0055] The dynamic programming formula is as follows:

[0056]

[0057] Where D=[D1,D2,...,D i ] represents the distance data between solid-state laser radar segments, i represents the number of laser radar segments, D i represents the distance data of the i-th segment of the lidar, diff[i] represents the difference between the i-th adjacent distance data points, diff[m] represents the difference between the m-th adjacent distance data points, j represents the starting index of the lidar distance data segment, dp[i] represents the optimal grouping cost of the first i segments, dp[j] represents the optimal grouping cost of the first j segments, cost(j+1,i) represents the cost of the first i segments from segment D j+1 to D i The cost and j * represents the optimal segmentation point with the minimum total cost, P represents the set of all optimal cutting points, j1 represents the optimal cutting point of the first segment group, and j S-1 represents the optimal cutting point of the S-1 segment group, S represents the optimal number of segment groups, represents the oth optimal cutting point, j o+1 represents the starting position of the o+1 segment; |P| represents the absolute value of P; argmin represents the variable value that minimizes the expression, that is, the value of the optimal segmentation point position; the binary attribution matrix formula is as follows:

[0058]

[0059] Among them, M p,q represents the qth data point in the pth segment, that is, the element in the data matrix, p represents the current pth segment, q represents the qth element in the current data point sequence, D q Represents an element in the data sequence, that is, the qth data point in the input data, B p,q represents the qth binary membership matrix element in the pth segment.

[0060] Furthermore, the adaptive fuzzy clustering algorithm based on dynamic threshold is formulated as follows:

[0061]

[0062] Among them, σ represents the σth iteration, l represents the lth column of the pre-grouping matrix, that is, the lth segment, r represents the rth row of the pre-grouping matrix, that is, the rth segment group, represents the center position of the rth segment group at the σth iteration, represents the degree of belonging of the lth segment in the rth segment group after the σth iteration of pre-grouping, D l Indicates the laser radar segment measurement distance of the current segment l, x lr represents the laser radar segment distance of the lth segment in the rth segment group of the pre-grouping matrix, i represents the number of laser radar segments, S represents the optimal number of segment groups, represents the belonging degree corresponding to the lth segment of the rth segment group of the σth iteration binary matrix, v represents the fuzzy index, Δc r represents the change value of the center of the rth segment group after t iterations, ò0 represents the initial threshold, σ Indicates the dynamic change of the convergence threshold of the σth iteration, error σ Indicates the error or change value of the σth iteration, error σ It is used to measure the state of the algorithm in the σth iteration, and α represents the adjustment factor.

[0063] Furthermore, the object of interest extraction unit is specifically configured to:

[0064] The pre-processed data is fuzzy processed by a fuzzy clustering algorithm based on dynamic threshold, and a fuzzy attribution degree matrix is ​​obtained to represent the membership relationship between the object and the laser beam.

[0065] The number of objects within the laser radar detection field of view and the laser radar segment group corresponding to the object are determined according to the fuzzy attribution matrix, and then the distance of the object is obtained according to the laser radar ranging formula;

[0066] The cluster closest to the lidar sensor is considered the object of interest.

[0067] Furthermore, the collaborative control unit is specifically used to:

[0068] The distance error value is calculated based on the distance between the following vehicle and the heavy-load train ahead and the expected cooperative distance. The speed of the heavy-load train is cooperatively controlled and adjusted according to the distance error value. The cooperative control adjustment is expressed as:

[0069]

[0070] Among them, D lateral,min Represents the three-dimensional distance of the object closest to the lidar, d ref Denotes the expected collaborative distance, L delay Indicates the running distance of the following train when the communication with the adjacent train is delayed, L break(r)Indicates the maximum braking distance of the following vehicle, L break(p) Indicates the maximum braking distance of the vehicle in front, L s represents the safety margin, max() represents the maximum value, t represents the time, e(t) represents the error between the actual distance of the train at time t and the target distance, d actual (t) represents the actual coordinated distance of the train at time t, Represents the rate of change of error e(t), T sample represents the sampling period, e(tT sample ) represents the error at the last sampling moment, v(t) represents the speed of the train at time t, v ref (t) represents the expected coordination speed, v ref (t) is obtained from the driving map, K represents the gain coefficient, which is used to control the response strength of speed to distance error, S k Indicates the idling stopping distance, S e represents the common braking distance of the train operation monitoring and recording device LKJ, v0 represents the initial braking speed, λ represents the braking calculation coefficient, ν h represents the train transfer braking rate, φ h Indicates the friction coefficient of the brake shoe, β c represents common braking parameters, ω0 represents the basic unit resistance of the train, t0 represents the driver's reaction time, v m Indicates the final speed of the heavy-load train after deceleration.

[0071] Furthermore, the adaptive fuzzy PID controller that minimizes the error is formulated as follows:

[0072]

[0073] Where u(t) represents the output signal of the controller at time t, e(t) represents the error between the actual distance of the train and the target distance at time t, and t represents the time. Indicates the rate of change of error e(t), K υ Represents the proportional gain parameter, K β Indicates the integral gain parameter, K d Denotes the differential gain parameter, K υ (t) represents the proportional gain at time t, K d (t) represents the differential gain at time t, K β (t) represents the integral gain at time t, K υ0 Indicates the initial proportional gain, K β0 Indicates the initial integral gain, K d0 Denotes the initial differential gain, ΔK υ Indicates the proportional gain adjustment, ΔK d Indicates the differential gain adjustment, ΔKβ Indicates the integral gain adjustment, ΔK υ , ΔK d , ΔK β The three adjustment quantities are obtained through fuzzy reasoning.

[0074] Furthermore, an improved particle swarm optimization algorithm is used to dynamically adjust the initial parameters of the adaptive fuzzy PID controller with minimum error.

[0075] Among them, the particle update strategy of the improved particle swarm optimization algorithm is as follows:

[0076]

[0077] Among them, H n represents the position of the nth particle in the current search space, K υ Represents the proportional gain parameter, K β Indicates the integral gain parameter, K d represents the differential gain parameter, T represents the time step range of error calculation, Q(z) represents the error at the zth iteration, J represents the objective function to be minimized, and v n (z) represents the velocity of the nth particle at the zth iteration, w represents the inertia weight term, p n represents the current optimal position found by the nth particle, r1 and r2 represent random numbers between [0,1], c1 represents the first learning factor, c2 represents the second learning factor, g represents the global optimal position found among all particles, and x n (z) The position of particle n at the zth iteration, w max and w min Represent the maximum and minimum values ​​of the inertia weight, z represents the number of iterations, and z max Maximum number of iterations, Represent the maximum values ​​of c1 and c2 respectively, Represent the minimum values ​​of c1 and c2 respectively.

[0078] According to a third aspect, an electronic device includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus;

[0079] a memory storing a computer program;

[0080] The processor is used to implement the above-mentioned heavy-load train group coordinated control method when executing the computer program stored in the memory.

[0081] In a fourth aspect, a computer-readable storage medium stores a computer program, which, when executed by a processor, implements the above-mentioned method for coordinated control of a heavy-load train group.

[0082] The present disclosure includes at least the following beneficial effects:

[0083] This system uses real-time processing of lidar data, combined with dynamic programming and fuzzy control algorithms, to precisely adjust vehicle distance and speed. Using an adaptive fuzzy PID controller and an improved particle swarm optimization algorithm, it dynamically optimizes controller parameters, improving system stability and control accuracy. This effectively addresses the shortcomings of traditional control systems in dynamic environments and provides a new solution for the efficient and safe operation of heavy-load train groups.

[0084] The present invention determines the safety of the cooperative control environment of heavy-load trains based on the segmented data of the laser radar, the speed of the detected object, and the distance to the solid-state laser radar. When the solid-state laser radar detects that the cooperative control environment is safe, that is, there are no other objects within the cooperative control distance range except the heavy-load train in front, the solid-state laser radar will obtain the actual distance and speed of the cooperatively controlled heavy-load train in front, and use the distance error and speed of the heavy-load train as input and the speed adjustment gain of the heavy-load train as output to establish a heavy-load train cooperative control adjustment system. Based on the improved particle swarm optimization algorithm (IPSO), the adaptive fuzzy PID controller with minimum error is optimized, which shortens the distance of the cooperative control of heavy-load trains and improves the railway transportation capacity.

[0085] Other features and advantages of the present disclosure will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present disclosure. The purpose and other advantages of the present disclosure can be achieved and obtained through the structures indicated in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0087] Figure 1 This is a flow chart of the control method according to an embodiment of the present disclosure;

[0088] Figure 2 This is a schematic diagram of the structure of the control device according to an embodiment of the present disclosure;

[0089] Figure 3 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present disclosure;

[0090] Figure 4 This is a schematic diagram of the segmented distance of the laser radar according to an embodiment of the present disclosure;

[0091] Figure 5 This is a convergence curve diagram of the dynamic threshold method according to an embodiment of the present disclosure;

[0092] Figure 6 This is a radar test field of view diagram according to an embodiment of the present disclosure, wherein the object is in the middle of the field of view and close to the lidar;

[0093] Figure 7 This is a radar test field of view diagram according to an embodiment of the present disclosure, wherein the object is in the middle of the field of view and far away from the laser radar;

[0094] Figure 8 This is a radar test field of view diagram according to an embodiment of the present disclosure, in which only half of the object is within the laser radar field of view;

[0095] Figure 9 For the embodiment of the present disclosure Figure 6 Schematic diagram of the initial matrix corresponding to the radar test field of view;

[0096] Figure 10 For the embodiment of the present disclosure Figure 6 Schematic diagram of the fuzzy matrix corresponding to the radar test field of view;

[0097] Figure 11 For the embodiment of the present disclosure Figure 7 Schematic diagram of the initial matrix corresponding to the radar test field of view;

[0098] Figure 12 For the embodiment of the present disclosure Figure 7 Schematic diagram of the fuzzy matrix corresponding to the radar test field of view;

[0099] Figure 13 For the embodiment of the present disclosure Figure 8 Schematic diagram of the initial matrix corresponding to the radar test field of view;

[0100] Figure 14 For the embodiment of the present disclosure Figure 8 Schematic diagram of the fuzzy matrix corresponding to the radar test field of view;

[0101] Figure 15 This is a block diagram of the coordinated control of heavy-load trains according to an embodiment of the present disclosure;

[0102] Figure 16 A three-dimensional graph showing the relationship between the degree of membership, speed, and coordination distance of the heavy-loaded train ahead as measured by the rear train in the embodiment of the present disclosure;

[0103] Figure 17 A diagram showing the distance error and real-time distance error rate for collaboratively controlling a heavy-load train according to an embodiment of the present disclosure;

[0104] Figure 18 A diagram showing the speed error and real-time speed error rate of a coordinated heavy-load train controlled in accordance with an embodiment of the present disclosure;

[0105] Figure 19 A diagram of coordinated speed and coordinated distance for coordinated control of heavy-load trains according to an embodiment of the present disclosure;

[0106] Figure 20 This is a schematic diagram of the IPSO algorithm optimization process according to an embodiment of the present disclosure;

[0107] Figure 21 Schematic diagram comparing the iterative changes of the optimal individual fitness values ​​of the improved IPSO algorithm and the traditional PSO algorithm in the embodiment of the present disclosure. DETAILED DESCRIPTION

[0108] To make the objectives, technical solutions, and advantages of the embodiments of the present disclosure more clear, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present disclosure without making any creative efforts shall fall within the scope of protection of the present disclosure.

[0109] like Figure 1 As shown, a heavy-load train group coordinated control method includes:

[0110] S101, a following vehicle in a heavy-load train group obtains real-time segmented data of the heavy-load train group's surrounding environment based on a laser radar;

[0111] S102, preprocessing the real-time segmented data using a dynamic programming method to obtain preprocessed data;

[0112] S103, using a fuzzy clustering algorithm based on a dynamic threshold to perform fuzzy processing on the pre-processed data, extracting objects of interest, and obtaining the distances between the following vehicle and all objects of interest within a preset distance range;

[0113] S104, when there is only a heavy-load train ahead within a preset distance range, the following vehicle uses an adaptive fuzzy PID controller that minimizes the error to adjust the distance and speed between the following vehicle and the heavy-load train ahead based on the distance between the following vehicle and the heavy-load train ahead.

[0114] The specific implementation is as follows:

[0115] Based on real-time segmented data acquired by solid-state LiDAR, high-precision perception of the surrounding environment of heavy-load trains is performed to ensure environmental safety for collaborative control.

[0116] The distance and speed of the cooperative control of heavy-haul trains are adjusted according to the distance error of the cooperative control of heavy-haul trains obtained by the solid-state laser radar.

[0117] The real-time segmented data acquired from the solid-state laser radar is preprocessed using a dynamic programming method. In order to more accurately identify the front and back relationships of objects, an adaptive fuzzy clustering algorithm based on a dynamic threshold is further used to perform fuzzy processing on the preprocessed segmented data, thereby extracting the target object of interest and calculating the distance from the solid-state laser radar to the object of interest. An adaptive fuzzy PID controller that minimizes the error is used to accurately adjust the coordinated control distance and speed of the heavy-load train; the parameters of the adaptive fuzzy PID controller that minimizes the error can be dynamically adjusted; and an improved particle swarm optimization algorithm is used to dynamically adjust the initial parameters of the adaptive fuzzy PID controller that minimizes the error.

[0118] The segmented data includes: segmented distance of solid-state LiDAR, distance from the object of interest to LiDAR, and speed of heavy-load train. The distance from the object of interest to LiDAR and speed of heavy-load train are directly measured by solid-state LiDAR. The segmented distance of LiDAR is expressed as:

[0119]

[0120] Where, D represents the segment distance of the laser radar, d i and d i+1 They represent the distances of the two laser beams emitted by the solid-state laser radar, θ represents the angle between the two laser beams emitted by the solid-state laser radar, cos(θ) represents the cosine value of the angle θ, and i represents the number of segments of the laser radar.

[0121] The dynamic programming method is used to preprocess the collected laser radar segment data and generate the corresponding binary attribution matrix. The dynamic programming method is expressed as:

[0122]

[0123] Where D=[D1,D2,...,D i ] represents the distance data between solid-state laser radar segments, i represents the number of laser radar segments i, D i represents the distance data of the i-th segment of the laser radar, diff[i] represents the difference between the i-th adjacent distance data points, j represents the starting index of the laser radar distance data segment, dp[i] represents the optimal grouping cost of the first i segments, dp[j] represents the optimal grouping cost of the first j segments, cost(j+1,i) represents the cost of the optimal grouping cost from segment D j+1 to D i The cost and j *represents the optimal segmentation point with the minimum total cost, P represents the set of all optimal cutting points, j S-1 represents the optimal cutting point of the S-1 segment group, S represents the optimal number of segment groups, represents the kth optimal cutting point, j k+1 Indicates the starting position of the k+1 segment.

[0124] The formula for generating a binary attribution matrix is:

[0125]

[0126] Among them, M p,q represents the qth data point (element in the data matrix) in the pth segment, p represents the current pth segment, q represents the current qth element in the data point sequence, D q Represents the element D in the data sequence = [D1, D2, ..., D i ], that is, the qth data point in the input data, B p,q represents the qth binary matrix element in the pth segment.

[0127] The adaptive fuzzy clustering algorithm based on dynamic threshold is used to perform fuzzy processing on the pre-processed segmented data. The adaptive fuzzy clustering algorithm based on dynamic threshold is expressed as:

[0128]

[0129] Where σ represents the σ-th iteration, l represents the l-th column (l-th segment) of the pre-grouping matrix, r represents the r-th row (j-th segment group) of the pre-grouping matrix, represents the center position of the rth segment group at the σth iteration, represents the degree of belonging of the lth segment in the rth segment group after the σth iteration of pre-grouping, D i Indicates the laser radar segment measurement distance of the current i-th segment, x lr represents the laser radar segment distance of the lth segment in the rth segment group of the pre-grouping matrix, i represents the number of laser radar segments, S represents the optimal number of segment groups, It represents the belonging degree corresponding to the lth segment of the rth segment group of the σth iteration binary matrix, v is the fuzzy index usually greater than 1, Δc r represents the change value of the center of the rth segment group after t iterations, ò0 represents the initial threshold, σ Indicates the dynamic change of the convergence threshold of the σth iteration, error σ The error or change value of the σth iteration is used to measure the state of the algorithm in the σth iteration, and α represents the adjustment factor.

[0130] The cluster closest to the lidar sensor is taken as the object of interest, mainly because it provides a higher guarantee in terms of safety, as the closest object can be considered the most potentially dangerous. The fuzzy attribution matrix is ​​defined as:

[0131]

[0132] Among them, u (σ) Represents the fuzzy attribution matrix, κ represents the κth column of the fuzzy attribution matrix, which actually means there are κ objects within the field of view of the lidar detection, and r represents the rth row of the fuzzy attribution matrix, which is the rth segment group.

[0133] The number of objects within the LiDAR detection field of view and the LiDAR segment group corresponding to the object are determined based on the fuzzy attribution matrix. The distance to the object is then obtained based on the LiDAR ranging formula. The LiDAR ranging steps are as follows:

[0134] For the κth object, in order to ensure data validity while avoiding the introduction of excessive irrelevant or noisy data, all laser beams with fuzzy membership greater than 0.6 and meeting the angle conditions are selected. The formula for the effective beam set is as follows:

[0135]

[0136] Among them, M′ κ represents the effective beam set of the κth object, represents the membership value of the κth object and the rth laser beam, r represents the rth segment group, φ r represents the elevation angle of the rth segment group, φ min Minimum elevation angle of the segment group, φ min Take -15°, φ max Indicates the maximum elevation angle of the segment group, φ max Take 15°.

[0137] According to the effective beam set of the object, the laser radar coordinate conversion is then performed, and the formula is as follows:

[0138]

[0139] Among them, L represents the laser radar coordinate system, P L,r represents the coordinate value of the rth laser beam in the laser radar coordinate system, d r The distance of the rth laser segment, θ r represents the horizontal angle of the rth segment group, φ r represents the elevation angle of the rth segment group, B represents the vehicle coordinate system, The original coordinates of the rth laser segment in the vehicle coordinate system, R represents the rotation matrix, which can be obtained by the installation angle of the laser radar relative to the vehicle, and T represents the translation vector, that is, the offset of the coordinate origin, which can be obtained through calibration.

[0140] After the conversion from the radar coordinate system to the vehicle coordinate system, the obtained coordinates represent the position of the segment group in the vehicle coordinate system. These beam data need to be further processed to more accurately fuse and locate the target. The formula is as follows:

[0141]

[0142] Among them, L represents the laser radar coordinate system, B represents the vehicle coordinate system, and r represents the rth segment group. The weight of the rth beam, φ r represents the elevation angle of the rth segment group, φ opt represents the optimal angle, the optimal angle is 0°, λ represents the attenuation factor, δ represents the smoothing constant, represents the fusion position of the κth object, Represents the original position of the rth segment laser, M′ κ represents the valid segment set of the κth object, v B,r represents the velocity of the rth laser segment in the vehicle coordinate system, v r,doppler represents the Doppler velocity of the rth segment laser, represents the calibration position of the κth object after velocity and acceleration compensation, T sample Indicates the radar sampling period.

[0143] The position obtained is in vector form. To convert it into the actual plane distance, the formula is as follows:

[0144]

[0145] Where B represents the vehicle coordinate system, represents the calibration position of the κth object after velocity and acceleration compensation, It represents the lateral position of the object in the vehicle coordinate system with Doppler velocity compensation added. It represents the longitudinal position of the object in the vehicle coordinate system with Doppler velocity compensation added. Indicates the vertical position of the object in the vehicle coordinate system with Doppler velocity compensation added. Indicates the lateral position of the object in the vehicle coordinate system without adding Doppler velocity compensation, Indicates the longitudinal position of the object in the vehicle coordinate system without Doppler velocity compensation. It represents the vertical position of the object in the vehicle coordinate system without adding Doppler velocity compensation, T samplerepresents the sampling period, r represents the rth segment group, The weight of the r-th segment group, M′ κ represents the valid segment set of the κth object, v B,r,x represents the lateral velocity of the rth laser segment in the vehicle coordinate system, v B,r,y represents the longitudinal velocity of the rth laser segment in the vehicle coordinate system, v B,r,z Represents the vertical velocity of the rth laser segment in the vehicle coordinate system.

[0146] The cluster closest to the lidar sensor is considered the object of interest, and the formula is as follows:

[0147]

[0148] Among them, D lateral,κ The three-dimensional distance of the laser beam to the κth object, It represents the lateral position of the object in the vehicle coordinate system with Doppler velocity compensation added. It represents the longitudinal position of the object in the vehicle coordinate system with Doppler velocity compensation added. It represents the vertical position of the object in the vehicle coordinate system with Doppler velocity compensation added, D lateral,min represents the 3D distance of the object closest to the lidar, r represents the rth segment group, M′ κ represents the effective beam set of the κth object, and min represents the minimum value.

[0149] The distance error is calculated based on the distance from the train to the solid-state lidar and the expected cooperative distance. The distance error is calculated by subtracting the expected distance from the distance from the train to the solid-state lidar. The speed of the heavy-loaded train is then adjusted based on the distance error. The cooperative control adjustment is expressed as:

[0150]

[0151] Among them, D lateral,min Represents the three-dimensional distance of the object closest to the lidar, d ref Denotes the expected collaborative distance, L delay Indicates the running distance of the following train when the communication with the adjacent train is delayed, L break(r) Indicates the maximum braking distance of the following vehicle, L break(p) Indicates the maximum braking distance of the vehicle in front, L s represents the safety margin, max() represents the maximum value, t represents the time, e(t) represents the error between the actual distance of the train at time t and the target distance, d actual (t) represents the actual coordinated distance of the train at time t, Represents the rate of change of error e(t), T sample represents the sampling period, e(tTsample ) represents the error at the last sampling moment, v(t) represents the speed of the train at time t, v ref (t) represents the expected coordination speed, v ref (t) is obtained from the driving map, K represents the gain coefficient, which is used to control the response strength of speed to distance error, S k Indicates the idling stopping distance, S e represents the common braking distance of the train operation monitoring and recording device LKJ, v0 represents the initial braking speed, λ represents the braking calculation coefficient, ν h represents the train transfer braking rate, φ h Indicates the friction coefficient of the brake shoe, β c represents common braking parameters, ω0 represents the basic unit resistance of the train, t0 represents the driver's reaction time, v m Indicates the final speed of the heavy-load train after deceleration.

[0152] The goal of adjusting the gain coefficient K through real-time feedback is to keep the error at a minimum while avoiding over-adjustment and oscillation. The adaptive fuzzy PID controller that minimizes the error continuously adjusts the gain according to the error and the error change rate. The adaptive fuzzy PID controller that minimizes the error is expressed as:

[0153]

[0154] Where u(t) represents the output signal of the controller at time t, e(t) represents the error between the actual distance of the train and the target distance at time t, and t represents the time. Indicates the rate of change of error e(t), K υ Represents the proportional gain parameter, K β Indicates the integral gain parameter, K d Denotes the differential gain parameter, K υ (t) represents the proportional gain at time t, K d (t) represents the differential gain at time t, K β (t) represents the integral gain at time t, K υ0 Indicates the initial proportional gain, K β0 Indicates the initial integral gain, K d0 Denotes the initial differential gain, ΔK υ Indicates the proportional gain adjustment, ΔK d Indicates the differential gain adjustment, ΔK β Indicates the integral gain adjustment, ΔK υ , ΔK d , ΔK β The three adjustment quantities are obtained through fuzzy reasoning.

[0155] The improved particle swarm optimization algorithm is used to dynamically adjust the initial parameters of the adaptive fuzzy PID controller with minimum error, including:

[0156] The improved particle swarm optimization algorithm is used to dynamically adjust the initial parameters of the adaptive fuzzy PID controller with minimum error. The particle update strategy of the improved particle swarm optimization algorithm is expressed as:

[0157]

[0158] Among them, H n represents the position of the nth particle in the current search space, K υ Represents the proportional gain parameter, K β Indicates the integral gain parameter, K d represents the differential gain parameter, T represents the time step range of error calculation, Q(z) represents the error at the zth iteration, J represents the objective function to be minimized, and v n (z) represents the velocity of the nth particle at the zth iteration, w represents the inertia weight term, p n represents the current optimal position found by the nth particle, r1 and r2 represent random numbers between [0,1], c1 represents the first learning factor, c2 represents the second learning factor, g represents the global optimal position found among all particles, and x n (z) The position of particle n at the zth iteration, w max and w min Represent the maximum and minimum values ​​of the inertia weight, z represents the number of iterations, and z max Maximum number of iterations, Represent the maximum values ​​of c1 and c2 respectively, Represent the minimum values ​​of c1 and c2 respectively.

[0159] In the detection of the collaborative control environment of heavy-loaded trains, the safety of the collaborative control environment of heavy-loaded trains is judged based on the segmented data of the lidar, the speed of the detected object, and the distance to the solid-state lidar. When the collaborative control environment of heavy-loaded trains is safe, the data obtained by the solid-state lidar is the actual collaborative control distance and speed of the collaboratively controlled heavy-loaded train in front.

[0160] In the coordinated control and regulation of heavy-load trains, it is used to establish a heavy-load train coordinated control and regulation system with the distance error and speed of the heavy-load train as input and the speed regulation gain of the heavy-load train as output.

[0161] In the initial parameter adjustment part, the improved particle swarm optimization algorithm is used to optimize the adaptive fuzzy PID controller with minimum error, and the heavy-load train cooperative control system is controlled based on the optimized controller. The PSO optimizes three parameters in total.

[0162] like Figure 2 As shown, a heavy-load train group cooperative control device includes: a segmented data acquisition unit 201, a pre-processing unit 202, an object of interest extraction unit 203 and a cooperative control unit 204;

[0163] The segmented data acquisition unit 201 is used for the following vehicle in the heavy-load train group to acquire real-time segmented data of the environment surrounding the heavy-load train group based on the laser radar;

[0164] A preprocessing unit 202 is used to preprocess the real-time segmented data using a dynamic programming method to obtain preprocessed data;

[0165] The object of interest extraction unit 203 is configured to perform fuzzy processing on the pre-processed data using a fuzzy clustering algorithm based on a dynamic threshold, extract the object of interest, and obtain the distance between the following vehicle and all objects of interest within a preset distance range;

[0166] The collaborative control unit 204 is used to adjust the distance and speed between the following vehicle and the heavy-loaded train in front by using an adaptive fuzzy PID controller that minimizes the error based on the distance between the following vehicle and the heavy-loaded train in front when there is only a heavy-loaded train in front within a preset distance range.

[0167] like Figure 3 As shown, the present disclosure provides an electronic device, including a processor 301, a communication interface 302, a memory 303 and a communication bus 304, wherein the processor 301, the communication interface 302 and the memory 303 communicate with each other through the communication bus 304;

[0168] Memory 303, storing computer programs;

[0169] The processor 301 is configured to implement the above method when executing the computer program stored in the memory 303 .

[0170] The present disclosure provides a computer-readable storage medium storing a computer program, which implements the above method when executed by a processor.

[0171] The computer-readable storage medium may be included in the device / apparatus described in the above embodiments, or may exist independently without being incorporated into the device / apparatus. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of the present disclosure.

[0172] According to an embodiment of the present disclosure, a computer-readable storage medium may be a non-volatile computer-readable storage medium, such as, but not limited to, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0173] In order to enable those skilled in the art to better understand the present disclosure, the principles of the present disclosure are described as follows with reference to the accompanying drawings:

[0174] First, solid-state lidar segmentation technology is introduced in the coordinated control of heavy-haul trains. The sampling time of the solid-state lidar is 10 ms, and the specific parameters are shown in Table 1. The real-time segmented data acquired by the lidar is preprocessed using dynamic programming. A fuzzy clustering algorithm based on dynamic thresholds is then used to fuzzify the preprocessed segmented data and generate a corresponding fuzzy attribution matrix. This fuzzy attribution matrix is ​​then used to identify the safety of the coordinated control environment.

[0175] Table 1

[0176] parameter scope Number of segments / segment 8 Field of view 20° Geometric dimensions 70mm×35.9mm×71.2mm Use wavelength / nm 905 Distance measurement accuracy / cm 5 Working temperature / ℃ -40～85 Maximum detection distance / m 1315

[0177] By analyzing the number of solid-state lidar segments collected in reality (segment distance of solid-state lidar, distance from heavy-load train to solid-state lidar, distance from obstacle to solid-state lidar, speed of heavy-load train), the feasibility of dynamic dynamic programming method and fuzzy clustering algorithm based on dynamic threshold is verified to achieve accurate judgment of the collaborative environment of heavy-load train.

[0178] In the case of coordinated control of heavy-load trains, the segmented distances (such as Figure 4 The segmented distance diagram obtained by the solid-state LiDAR is real-time, which requires the algorithm to process the data quickly and accurately. The following steps should be followed to dynamically program the segmented distance obtained by the solid-state LiDAR:

[0179] Step 1: Calculate the adjacent difference of each segment distance obtained by the solid-state laser radar. The adjacent difference is expressed as:

[0180] diff[i]=|D i+1 -D i | (15)

[0181] Among them, diff[i] represents the distance difference between adjacent segments of the solid-state laser radar, D i+1 and Di Respectively represent the segment distances of the i+1th segment and the ith segment

[0182] Step 2: Initialize the dynamic programming array and initialize dp[0] = 0, which means that when there is no segmentation, the grouping cost is 0.

[0183] Step 3: Define segment costs: Here cost(j+1,i) is the cost from segment D i+1 to D i The cumulative difference of .

[0184] Step 4: Dynamic programming recursion, the recursive formula is as follows:

[0185]

[0186] For each i, traverse all possible segment starting points j and calculate the segment cost cot(j+1,i) from j+1 to i. Add the optimal cost of the current segment to the cumulative optimal cost dp[j] of the previous segment to obtain the group cost. Find the minimum cost among all possible j and assign it to dp[i].

[0187] Step 5: Calculate the optimal segmentation point. The formula is as follows:

[0188]

[0189] Find the cutting point j that makes dp[j] reach the minimum value * , and through j * Backtrack to determine the final split point.

[0190] Step 6: Calculate the optimal segment: P = {j1, j2......j S-1}, P is used to store all optimal cutting points, indicating the optimal segmentation position of the data.

[0191] Step 7: Calculate the optimal number of segments. The formula is as follows:

[0192] S=|P|+1 (18)

[0193] Calculate the final number of segments, that is, the optimal S segment

[0194] Step 8: Backtrack and calculate the optimal split point. The formula is as follows:

[0195]

[0196] Backtrack to find each optimal cutting point, determine the first segment, the second segment, until the S segment

[0197] Step 9: Output the optimal segmentation. After the dynamic programming is completed, the value of the array dp[i] is the optimal grouping cost for the entire lidar segmentation data.

[0198] Step 10: Construct the data matrix. The formula is as follows:

[0199]

[0200] Fill each data segment with the corresponding value, and fill the rest with 0. The first row of the matrix M structure: the first segment data, and the rest are filled with 0. The second row: the second segment data, and the rest are filled with 0. And so on, until the Sth segment.

[0201] Step 11: Construct a binary matrix. The formula is as follows:

[0202]

[0203] Used to mark which segment a data point belongs to. The structure of matrix B is as follows: first row: the position of the first segment is filled with 1, and the rest are filled with 0. Second row: the position of the second segment is filled with 1, and the rest are filled with 0. And so on, until the Sth segment.

[0204] Combining the above steps, the segmented distance data obtained by the solid-state laser radar in real time can be processed quickly and accurately, thereby improving the accuracy of fuzzy processing.

[0205] Fuzzy clustering algorithms based on dynamic thresholds include:

[0206] Fuzzy clustering is used to fuzzy process the segmented distances processed by dynamic programming. Through continuous iteration, an accurate and stable fuzzy attribution matrix is ​​obtained. In the iterative process, in order to avoid too few iterations and falling into infinite iterations, a dynamic threshold is used for iteration: First, the initial threshold is determined:

[0207] ò0=Δc r0 (twenty two)

[0208] Among them, ò0 represents the initial threshold, Δc r0 Indicates the error value at the initial moment.

[0209] The dynamic threshold formula is:

[0210]

[0211] Among them, ò0 represents the initial threshold, which represents the convergence threshold at the beginning of the algorithm, σ is the dynamically changing convergence threshold, error σ Indicates the error or change value of the current iteration, which is used to measure the state of the algorithm in the σth iteration, α σrepresents the dynamic adjustment coefficient, α0 represents the initial adjustment coefficient (usually 1 or other positive values), error max Indicates the maximum value of the error, usually the initial error value, error σ The error value of the current iteration.

[0212] According to the dynamic threshold formula:

[0213] when error σ When the error is large, α σ Larger, so that the threshold σ The adjustment range is larger, when error σ When the error is small, α σ Reduce, so that the threshold σ Adjust the amplitude finely, when error σ The error is less than the dynamic threshold σ , the algorithm is considered to have converged and the loop is stopped. To prevent max , when σ>σ max Forces the loop to stop.

[0214] According to the dynamic threshold formula:

[0215] when error σ When the error is large, α σ Larger, so that the threshold σ The adjustment range is larger, when error σ When the error is small, α σ Reduce, so that the threshold σ Adjust the amplitude finely, when error σ The error is less than the dynamic threshold σ , the algorithm is considered to have converged and the loop is stopped. To prevent max , when σ>σ max Forces the loop to stop.

[0216] like Figure 5 As shown in Figure 3, the dynamic threshold method converges faster.

[0217] like Figure 6-Figure 8 The lidar field of view diagrams corresponding to the three test conditions; Figure 6 The object is in the middle of the field of view and close to the lidar. Figure 7 The object is in the middle of the field of view and far away from the lidar. Figure 8 The object is only half within the LiDAR field of view and is close to the LiDAR. To better simulate real-world conditions, three scenarios were simulated to verify the algorithm's recognition accuracy.

[0218] Figure 9-10 The initial matrix and fuzzification matrix correspond to Figure 6 In the case of , the membership matrix consists of three clusters, Figure 9 is the initial matrix, Figure 10 is the membership matrix after fuzzification. Because the objects are close to the lidar and all objects are within the field of view, the membership values ​​do not change significantly relative to the initial matrix, which means that the intervals between clusters are large enough so that the membership values ​​in the boundary measurement are not fuzzified.

[0219] Figure 11-12 The initial matrix and fuzzification matrix correspond to Figure 7 situation, Figure 11 is the initial matrix, Figure 12 The blurred attribution matrix shows a slightly different situation in the second test. In this case, there is still an object in the center, but it is closer to the background wall.

[0220] Figure 13-14 The initial matrix and fuzzification matrix correspond to Figure 8 situation, Figure 13 is the initial matrix, Figure 14 This is the assignment matrix after fuzzification. When the object is partially introduced into the LIDAR field of view, a common phenomenon can be observed: when the object only covers a portion of the field of view, the measured distance of segment 5 often falls between the distances of the first and last clusters. In this case, the clustering algorithm typically assigns this segment to a specific cluster. Due to the characteristics of fuzzy clustering, the assignment of this intermediate value can be flexibly adjusted between two adjacent clusters, making the measurement results more accurate to the actual situation.

[0221] Figure 15 The block diagram of the coordinated control of heavy-load trains according to the embodiment of the present disclosure is shown as follows:

[0222] Step 103: Adopt the adaptive fuzzy PID controller with minimized error to accurately adjust the coordinated control distance and speed of the heavy-load train. When the coordinated distance does not reach the expected coordinated distance, the adaptive fuzzy PID controller with minimized error will adjust the speed control gain to make the heavy-load train change its speed.

[0223] The experience and knowledge accumulated by drivers through long-term practice are modeled using a fuzzy rule base to establish an offline query fuzzy matrix table; during online operation, fuzzy reasoning is used to design an adaptive fuzzy PID controller that estimates and minimizes the error. By processing the results of fuzzy logic rules, looking up tables, and calculating them, the influence of uncertain factors on the coordinated control of heavy-load trains during train operation is reduced. The input and output membership functions of the fuzzy controller are all overlapping and symmetrical generalized bell-shaped membership functions. The design of the adaptive fuzzy PID controller that minimizes the error is based on the coordinated distance error e(t) of the heavy-load train and the rate of change of the coordinated distance error. is the input, ΔK p , ΔK i , ΔK d The change in the parameter is the output, the reasoning adopts the maximum and minimum synthesis rule, and the output adopts the center of gravity defuzzification method, as follows:

[0224]

[0225] Among them, μ(Δk' j ),Δk' j is the membership degree of the corresponding fuzzy subset membership function and the fuzzy reasoning output, R j is a fuzzy relationship, Δk is an adaptive adjustment increment, s j Represents the membership degree in fuzzy rules.

[0226] The adaptive fuzzy PID controller that minimizes the error uses the fuzzy control principle to adjust the PID parameters online in real time, so that the controlled object has good adaptability and control performance. To establish a suitable fuzzy rule base, its fuzzy control rules should follow the following rules:

[0227] In order to better perform fuzzy control, set the threshold. When the corresponding threshold is reached, make corresponding adjustments. The threshold setting includes:

[0228] Error range definition: Error range e range is the collaborative distance d ref The error range is ±2% and is expressed as:

[0229] e range =±0.02·d ref (25)

[0230] where e range Indicates the error range, d ref Indicates the set coordination distance.

[0231] The error change rate is defined as:

[0232]

[0233] in, represents the error change rate, e range Indicates the error range, T sample represents the sampling period, d ref Indicates the set coordination distance.

[0234] The error range and error change rate range are combined with the input design of fuzzy control and divided into three fuzzy sets: "small" (S), "medium" (M), and "large" (L). The specific division is as follows:

[0235] The error threshold is divided into 3 levels:

[0236] Small (S): |e|≤0.5·e range , indicating that the error is close to the target value and the controller does not need to be significantly adjusted.

[0237] Medium (M): 0.5·e range <|e|≤0.8·e range It means that the error deviates from the target value in the medium range and the controller needs to be adjusted appropriately.

[0238] Large (L): |e|>0.8·e range , indicating that the error deviates far from the target value and the controller needs to be adjusted quickly.

[0239] The error change rate range is divided into 3 levels:

[0240] Small (S): This means that the error rate of change is slow and the controller only needs minor adjustments.

[0241] Medium: It indicates that the error change rate is moderate and the controller needs to be adjusted appropriately.

[0242] Large (L): Indicates that the error changes dramatically and the controller needs to respond quickly

[0243] Based on the error e and the error change rate, fuzzy rules are designed to adjust the PID parameters and design ΔK p , ΔK i , ΔK d The fuzzy rule tables are shown in Table 2, Table 3 and Table 4:

[0244] Table 2

[0245]

[0246] Table 3

[0247]

[0248]

[0249] Table 4

[0250]

[0251] The fuzzy rule design has dynamic adaptability and robustness, and can be used in real time according to the error e and error change rate. Dynamically adjust PID parameters (K p ,K i ,K d), achieving a balance between rapid response and stability. It simplifies complex system control through rule tables, eliminating the need for precise modeling. It offers excellent scalability and interpretability, and excels in handling dynamic and uncertain scenarios (such as coordinated control of heavy-load trains).

[0252] like Figure 16 As shown in the figure, two heavy-load trains are running in coordination at a speed of 50 km / h. The rear train measures the membership and speed of the front heavy-load train, and the three-dimensional diagram of the coordination distance relationship. It can be known that the coordination distance is 461m and the coordination speed is 50 km / h.

[0253] like Figure 17-18 As shown, the adaptive fuzzy PID controller with minimized error performs well in the coordinated control of heavy-load trains, with an average speed error rate of 1.18% and an average spacing error rate of 1.17%.

[0254] like Figure 19 As shown in the figure, the relationship between the collaborative distance and collaborative speed of heavy-load trains based on lidar collaborative control can greatly reduce the collaborative distance and improve the transportation capacity of the railway.

[0255] Step 104: Using an improved particle swarm optimization algorithm to dynamically adjust the initial parameters of the adaptive fuzzy PID controller with minimum error.

[0256] This embodiment designs an adaptive fuzzy PID controller that minimizes errors and uses an improved PSO to dynamically adjust the initial parameters of the controller. The improved PSO dynamically adjusts the initial parameters of the controller, including:

[0257] like Figure 20 As shown, the controller flow chart based on the improved IPSO algorithm is as follows:

[0258] The process of setting parameters of the improved IPSO algorithm is as follows:

[0259] ① Initialization: Set the particle swarm size to SwarmSize=30, the inertia weight w max =0.9, w min =0.2, learning factor c 1min =0.1, c 2min =0.2, c 1max =1,c 2max =2, the maximum value of the search space V max =1, minimum value V min = -1, the maximum number of iterations is MaxIter = 20, and the position and velocity of the particle are initialized at the same time.

[0260] ② In order to obtain the optimal solution, the absolute value of error time integral performance index is used as the fitness function of the IPSO algorithm. The criteria are as follows:

[0261]

[0262] Where F is the fitness value, t is the time, and e(t) is the system error.

[0263] ③ Generate particle swarm, particle swarm K p ,K i ,K d Assign values ​​and obtain the fitness value of each particle according to formula (14), and compare and select the best individual and global ones.

[0264] ④ Update the position and velocity of the particles. If the termination condition is met, output the optimal solution. Otherwise, perform particle update operation and generate a new particle swarm for the next cycle operation.

[0265] The variable universe fuzzy PID control parameter K is obtained by optimizing the particle swarm algorithm before and after improvement. p ,K i ,K d ,Depend on Figure 21 The optimization comparison results of the fitness values ​​show that the improved particle swarm algorithm can reach the optimal fitness value after 9 iterations, which shows that the improved particle swarm algorithm proposed in this disclosure has better optimization performance and the obtained optimization results meet the optimization target requirements to a greater extent.

[0266] Although the present disclosure has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure.

Claims

1. A method for coordinated control of a heavy-load train group, characterized in that: The method comprises: The following vehicle in the heavy-load train group obtains real-time segmented data of the heavy-load train group's surrounding environment based on the lidar; The real-time segmented data is preprocessed using a dynamic programming method to obtain preprocessed data; The fuzzy clustering algorithm based on dynamic threshold is used to perform fuzzy processing on the pre-processed data, extract the objects of interest, and obtain the distances between the following vehicle and all objects of interest within the preset distance range; When there is only a heavy-load train ahead within the preset distance range, an adaptive fuzzy PID controller that minimizes the error is used to adjust the distance and speed between the following vehicle and the heavy-load train ahead based on the distance between the following vehicle and the heavy-load train ahead.

2. A heavy-load train group coordinated control method according to claim 1, characterized in that: Real-time segmentation data, including: The relative distance between the object of interest and the heavy-haul train equipped with the LiDAR, the speed of the object of interest relative to the heavy-haul train equipped with the LiDAR, and the segmented distance of the solid-state LiDAR; Among them, the relative distance from the object of interest to the heavy-load train equipped with the laser radar and the speed of the object of interest relative to the heavy-load train equipped with the laser radar are directly measured; The segment distance D of the solid-state laser radar is calculated by the following formula: Where, D represents the segment distance of the laser radar, d i and d i+1 They represent the distances of the two laser beams emitted by the solid-state laser radar, θ represents the angle between the two laser beams emitted by the solid-state laser radar, cos(θ) represents the cosine value of the angle θ, and i represents the number of segments of the laser radar.

3. A heavy-load train group coordinated control method according to claim 1, characterized in that: The real-time segmented data is pre-processed using dynamic programming, including: The dynamic programming method is used to preprocess the collected laser radar segmentation data and generate the corresponding binary attribution degree matrix as preprocessing data; The dynamic programming formula is as follows: Where D=[D1,D2,...,D i ] represents the distance data between solid-state laser radar segments, i represents the number of laser radar segments, D i represents the distance data of the i-th segment of the lidar, diff[i] represents the difference between the i-th adjacent distance data points, diff[m] represents the difference between the m-th adjacent distance data points, j represents the starting index of the lidar distance data segment, dp[i] represents the optimal grouping cost of the first i segments, dp[j] represents the optimal grouping cost of the first j segments, cost(j+1,i) represents the cost of the first i segments from segment D j+1 to D i The cost and j * represents the optimal segmentation point with the minimum total cost, P represents the set of all optimal cutting points, j1 represents the optimal cutting point of the first segment group, and j S-1 represents the optimal cutting point of the S-1 segment group, S represents the optimal number of segment groups, represents the oth optimal cutting point, j o+1 represents the starting position of the o+1 segment; |P| represents the absolute value of P; argmin represents the variable value that minimizes the expression, that is, the value of the optimal segmentation point position; the binary attribution matrix formula is as follows: Among them, M p,q represents the qth data point in the pth segment, that is, the element in the data matrix, p represents the current pth segment, q represents the qth element in the current data point sequence, D q Represents an element in the data sequence, that is, the qth data point in the input data, B p,q represents the qth binary membership matrix element in the pth segment.

4. A heavy-load train group coordinated control method according to claim 1, characterized in that: The adaptive fuzzy clustering algorithm based on dynamic threshold is as follows: Among them, σ represents the σth iteration, l represents the lth column of the pre-grouping matrix, that is, the lth segment, r represents the rth row of the pre-grouping matrix, that is, the rth segment group, represents the center position of the rth segment group at the σth iteration, represents the degree of belonging of the lth segment in the rth segment group after the σth iteration of pre-grouping, D l Indicates the laser radar segment measurement distance of the current segment l, x lr represents the laser radar segment distance of the lth segment in the rth segment group of the pre-grouping matrix, i represents the number of laser radar segments, S represents the optimal number of segment groups, represents the belonging degree corresponding to the lth segment of the rth segment group of the σth iteration binary matrix, v represents the fuzzy index, Δc r represents the change value of the center of the rth segment group after t iterations, ò0 represents the initial threshold, σ Indicates the dynamic change of the convergence threshold of the σth iteration, error σ Indicates the error or change value of the σth iteration, error σ It is used to measure the state of the algorithm in the σth iteration, and α represents the adjustment factor.

5. The method for coordinated control of a heavy-load train group according to claim 1, characterized in that: The fuzzy clustering algorithm based on dynamic threshold is used to fuzzy process the preprocessed data and extract the objects of interest, including: The pre-processed data is fuzzy processed by a fuzzy clustering algorithm based on dynamic threshold, and a fuzzy attribution degree matrix is ​​obtained to represent the membership relationship between the object and the laser beam. The number of objects within the laser radar detection field of view and the laser radar segment group corresponding to the object are determined according to the fuzzy attribution matrix, and then the distance of the object is obtained according to the laser radar ranging formula; The cluster closest to the lidar sensor is considered the object of interest.

6. A heavy-load train group coordinated control method according to claim 1, characterized in that: Adjust the distance and speed between the following vehicle and the heavy-load train ahead, including: The distance error value is calculated based on the distance between the following vehicle and the heavy-load train ahead and the expected cooperative distance. The speed of the heavy-load train is cooperatively controlled and adjusted according to the distance error value. The cooperative control adjustment is expressed as: Among them, D lateral,min Represents the three-dimensional distance of the object closest to the lidar, d ref Denotes the expected collaborative distance, L delay Indicates the running distance of the following train when the communication with the adjacent train is delayed, L break(r) Indicates the maximum braking distance of the following vehicle, L break(p) Indicates the maximum braking distance of the vehicle in front, L s represents the safety margin, max() represents the maximum value, t represents the time, e(t) represents the error between the actual distance of the train at time t and the target distance, d actual (t) represents the actual coordinated distance of the train at time t, Represents the rate of change of error e(t), T sample represents the sampling period, e(tT sample ) represents the error at the last sampling moment, v(t) represents the speed of the train at time t, v ref (t) represents the expected coordination speed, v ref (t) is obtained from the driving map, K represents the gain coefficient, which is used to control the response strength of speed to distance error, S k Indicates the idling stopping distance, S e represents the common braking distance of the train operation monitoring and recording device LKJ, v0 represents the initial braking speed, λ represents the braking calculation coefficient, ν h represents the train transfer braking rate, φ h Indicates the friction coefficient of the brake shoe, β c represents common braking parameters, ω0 represents the basic unit resistance of the train, t0 represents the driver's reaction time, v m Indicates the final speed of the heavy-load train after deceleration.

7. A heavy-load train group coordinated control method according to claim 6, characterized in that: The adaptive fuzzy PID controller that minimizes the error is formulated as follows: Where u(t) represents the output signal of the controller at time t, e(t) represents the error between the actual distance of the train and the target distance at time t, and t represents the time. Indicates the rate of change of error e(t), K υ Represents the proportional gain parameter, K β Indicates the integral gain parameter, K d Denotes the differential gain parameter, K υ (t) represents the proportional gain at time t, K d (t) represents the differential gain at time t, K β (t) represents the integral gain at time t, K υ0 Indicates the initial proportional gain, K β0 Indicates the initial integral gain, K d0 Denotes the initial differential gain, ΔK υ Indicates the proportional gain adjustment, ΔK d Indicates the differential gain adjustment, ΔK β Indicates the integral gain adjustment, ΔK υ , ΔK d , ΔK β The three adjustment quantities are obtained through fuzzy reasoning.

8. The method for coordinated control of a heavy-load train group according to claim 1, characterized in that: Also includes: An improved particle swarm optimization algorithm is used to dynamically adjust the initial parameters of the adaptive fuzzy PID controller with minimum error. Among them, the particle update strategy of the improved particle swarm optimization algorithm is as follows: Among them, H n represents the position of the nth particle in the current search space, K υ Represents the proportional gain parameter, K β Indicates the integral gain parameter, K d represents the differential gain parameter, T represents the time step range of error calculation, Q(z) represents the error at the zth iteration, J represents the objective function to be minimized, and v n (z) represents the velocity of the nth particle at the zth iteration, w represents the inertia weight term, p n represents the current optimal position found by the nth particle, r1 and r2 represent random numbers between [0,1], c1 represents the first learning factor, c2 represents the second learning factor, g represents the global optimal position found among all particles, and x n (z) The position of particle n at the zth iteration, w max and w min Represent the maximum and minimum values ​​of the inertia weight, z represents the number of iterations, and z max Maximum number of iterations, Represent the maximum values ​​of c1 and c2 respectively, Represent the minimum values ​​of c1 and c2 respectively.

9. A heavy-load train group coordinated control device, characterized in that: include: Segmented data acquisition unit, pre-processing unit, object of interest extraction unit and collaborative control unit; A segmented data acquisition unit is used for the following vehicle in the heavy-load train group to obtain real-time segmented data of the environment surrounding the heavy-load train group based on the laser radar; A preprocessing unit, configured to preprocess the real-time segmented data using a dynamic programming method to obtain preprocessed data; An object of interest extraction unit is used to perform fuzzy processing on the pre-processed data using a fuzzy clustering algorithm based on a dynamic threshold, extract the object of interest, and obtain the distance between the following vehicle and all objects of interest within a preset distance range; The collaborative control unit is used to adjust the distance and speed between the following vehicle and the heavy-loaded train in front by using an adaptive fuzzy PID controller that minimizes the error based on the distance between the following vehicle and the heavy-loaded train in front when there is only a heavy-loaded train in front within a preset distance range.

10. The heavy-load train group cooperative control device according to claim 9, characterized in that: Real-time segmentation data, including: The distance from the object of interest to the lidar, the speed of the object of interest relative to the heavy-load train carrying the lidar, and the segment distance for solid-state lidars; The distance from the object of interest to the lidar and the speed of the object of interest relative to the heavy-load train equipped with the lidar are directly measured by the solid-state lidar. The segment distance D of the solid-state laser radar is calculated as follows: Where, D represents the segment distance of the laser radar, d i and d i+1 They represent the distances of the two laser beams emitted by the solid-state laser radar, θ represents the angle between the two laser beams emitted by the solid-state laser radar, cos(θ) represents the cosine value of the angle θ, and i represents the number of segments of the laser radar.

11. The heavy-load train group cooperative control device according to claim 9, characterized in that: The preprocessing unit is specifically used to: The dynamic programming method is used to preprocess the collected laser radar segmentation data and generate the corresponding binary attribution degree matrix as preprocessing data; The dynamic programming formula is as follows: Where D=[D1,D2,...,D i ] represents the distance data between solid-state laser radar segments, i represents the number of laser radar segments, D i represents the distance data of the i-th segment of the lidar, diff[i] represents the difference between the i-th adjacent distance data points, diff[m] represents the difference between the m-th adjacent distance data points, j represents the starting index of the lidar distance data segment, dp[i] represents the optimal grouping cost of the first i segments, dp[j] represents the optimal grouping cost of the first j segments, cost(j+1,i) represents the cost of the first i segments from segment D j+1 to D i The cost and j * represents the optimal segmentation point with the minimum total cost, P represents the set of all optimal cutting points, j1 represents the optimal cutting point of the first segment group, and j S-1 represents the optimal cutting point of the S-1 segment group, S represents the optimal number of segment groups, represents the oth optimal cutting point, j o+1 represents the starting position of the o+1 segment; |P| represents the absolute value of P; argmin represents the variable value that minimizes the expression, that is, the value of the optimal segmentation point position; the binary attribution matrix formula is as follows: Among them, M p,q represents the qth data point in the pth segment, that is, the element in the data matrix, p represents the current pth segment, q represents the qth element in the current data point sequence, D q Represents an element in the data sequence, that is, the qth data point in the input data, B p,q represents the qth binary membership matrix element in the pth segment.

12. The heavy-load train group cooperative control device according to claim 9, characterized in that: The adaptive fuzzy clustering algorithm based on dynamic threshold is as follows: Among them, σ represents the σth iteration, l represents the lth column of the pre-grouping matrix, that is, the lth segment, r represents the rth row of the pre-grouping matrix, that is, the rth segment group, represents the center position of the rth segment group at the σth iteration, represents the degree of belonging of the lth segment in the rth segment group after the σth iteration of pre-grouping, D l Indicates the laser radar segment measurement distance of the current segment l, x lr represents the laser radar segment distance of the lth segment in the rth segment group of the pre-grouping matrix, i represents the number of laser radar segments, S represents the optimal number of segment groups, represents the belonging degree corresponding to the lth segment of the rth segment group of the σth iteration binary matrix, v represents the fuzzy index, Δc r represents the change value of the center of the rth segment group after t iterations, ò0 represents the initial threshold, σ Indicates the dynamic change of the convergence threshold of the σth iteration, error σ Indicates the error or change value of the σth iteration, error σ It is used to measure the state of the algorithm in the σth iteration, and α represents the adjustment factor.

13. The heavy-load train group cooperative control device according to claim 9, characterized in that: The object of interest extraction unit is specifically used to: The pre-processed data is fuzzy processed by a fuzzy clustering algorithm based on dynamic threshold, and a fuzzy attribution degree matrix is ​​obtained to represent the membership relationship between the object and the laser beam. The number of objects within the laser radar detection field of view and the laser radar segment group corresponding to the object are determined according to the fuzzy attribution matrix, and then the distance of the object is obtained according to the laser radar ranging formula; The cluster closest to the lidar sensor is considered the object of interest.

14. The heavy-load train group cooperative control device according to claim 9, characterized in that: Collaborative control unit, specifically used for: The distance error value is calculated based on the distance between the following vehicle and the heavy-load train ahead and the expected cooperative distance. The speed of the heavy-load train is cooperatively controlled and adjusted according to the distance error value. The cooperative control adjustment is expressed as: Among them, D lateral,min Represents the three-dimensional distance of the object closest to the lidar, d ref Denotes the expected collaborative distance, L delay Indicates the running distance of the following train when the communication with the adjacent train is delayed, L break(r) Indicates the maximum braking distance of the following vehicle, L break(p) Indicates the maximum braking distance of the vehicle in front, L s represents the safety margin, max() represents the maximum value, t represents the time, e(t) represents the error between the actual distance of the train at time t and the target distance, d actual (t) represents the actual coordinated distance of the train at time t, Represents the rate of change of error e(t), T sample represents the sampling period, e(tT sample ) represents the error at the last sampling moment, v(t) represents the speed of the train at time t, v ref (t) represents the expected coordination speed, v ref (t) is obtained from the driving map, K represents the gain coefficient, which is used to control the response strength of speed to distance error, S k Indicates the idling stopping distance, S e represents the common braking distance of the train operation monitoring and recording device LKJ, v0 represents the initial braking speed, λ represents the braking calculation coefficient, ν h represents the train transfer braking rate, φ h Indicates the friction coefficient of the brake shoe, β c represents common braking parameters, ω0 represents the basic unit resistance of the train, t0 represents the driver's reaction time, v m Indicates the final speed of the heavy-load train after deceleration.

15. The heavy-load train group cooperative control device according to claim 14, characterized in that: The adaptive fuzzy PID controller that minimizes the error is formulated as follows: Where u(t) represents the output signal of the controller at time t, e(t) represents the error between the actual distance of the train and the target distance at time t, and t represents the time. Indicates the rate of change of error e(t), K υ Represents the proportional gain parameter, K β Indicates the integral gain parameter, K d Denotes the differential gain parameter, K υ (t) represents the proportional gain at time t, K d (t) represents the differential gain at time t, K β (t) represents the integral gain at time t, K υ0 Indicates the initial proportional gain, K β0 Indicates the initial integral gain, K d0 Denotes the initial differential gain, ΔK υ Indicates the proportional gain adjustment, ΔK d Indicates the differential gain adjustment, ΔK β Indicates the integral gain adjustment, ΔK υ , ΔK d , ΔK β The three adjustment quantities are obtained through fuzzy reasoning.

16. The heavy-load train group cooperative control device according to claim 9, characterized in that: An improved particle swarm optimization algorithm is used to dynamically adjust the initial parameters of the adaptive fuzzy PID controller with minimum error. Among them, the particle update strategy of the improved particle swarm optimization algorithm is as follows: Among them, H n represents the position of the nth particle in the current search space, K υ Represents the proportional gain parameter, K β Indicates the integral gain parameter, K d represents the differential gain parameter, T represents the time step range of error calculation, Q(z) represents the error at the zth iteration, J represents the objective function to be minimized, and v n (z) represents the velocity of the nth particle at the zth iteration, w represents the inertia weight term, p n represents the current optimal position found by the nth particle, r1 and r2 represent random numbers between [0,1], c1 represents the first learning factor, c2 represents the second learning factor, g represents the global optimal position found among all particles, and x n (z) The position of particle n at the zth iteration, w max and w min Represent the maximum and minimum values ​​of the inertia weight, z represents the number of iterations, and z max Maximum number of iterations, Represent the maximum values ​​of c1 and c2 respectively, Represent the minimum values ​​of c1 and c2 respectively.

17. An electronic device, characterized in that: The processor, the communication interface, the memory and the communication bus are connected to each other via the communication bus. a memory storing a computer program; The processor is used to implement a heavy-load train group collaborative control method according to any one of claims 1 to 8 when executing a computer program stored in a memory.

18. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, a heavy-load train group collaborative control method according to any one of claims 1 to 8 is implemented.

Images