Multi-target ship communication path planning method based on offshore waveguide channel map

Abstract

The invention provides a multi-target ship communication path planning method based on an offshore waveguide channel map, and relates to the technical field of wireless communication, and the method comprises the following steps: building a single-input single-output ship motion model and a communication model under the condition of offshore long-distance transmission; establishing a CGM according to the existing gas phase data, the evaporation waveguide model and the electromagnetic wave propagation model; proposing an optimization problem model; an NSGA-II-PSO-VS (Non-dominated Sorting Genetic The operation results of the NSGA-II algorithm and the NSGA-II-PSO-VS algorithm at the same time are compared; simulation experiments are carried out under different evaporation waveguide heights, and the reasonability and applicability of the algorithm are proved. For a data-driven multi-target path planning problem of obtaining user information according to a channel map and modeling motion and communication of a ship, an improved heuristic algorithm is provided, two traditional algorithms are improved and combined, and compared with the traditional algorithm, the improved heuristic algorithm shows good performance in multiple aspects and has a wide application prospect. And help is provided for a multi-target path planning problem based on data driving.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication technologies, and in particular, to a multi-target ship communication path planning method based on a maritime waveguide channel map. Background Art

[0002] Evaporation duct is an atmospheric phenomenon. Inside the duct layer, the atmospheric refractive index changes, resulting in an electromagnetic wave super-refraction effect, causing the electromagnetic wave to reflect back and forth between the duct layer and the sea surface. Most of the wave energy is concentrated, reducing the path loss and enhancing the received signal power. In ship-to-shore communication, evaporation duct communication has the advantage of breaking through the traditional line-of-sight limitation, being able to maintain connection with a coastal base station at a long distance, and helping to reduce costs. As one of the key technologies for environment-aware communication, the Channel Knowledge Map (CKM) is constructed by integrating a large amount of historical data of all terminals in the area to reflect the characteristics of the local wireless environment. Environmental prior knowledge can be obtained in advance according to the information of the terminal position or virtual position, avoiding repeated online environment perception and channel acquisition, thereby realizing fast and real-time prediction and inference of channel knowledge and improving communication and perception performance. A Channel Gain Map (CGM) can be constructed based on the Channel Knowledge Map (CKM), combining the evaporation duct model, electromagnetic propagation model with real-time meteorological data (such as weather radar and ocean buoys), establishing a maritime evaporation duct channel map, and conducting path planning for ships to minimize the transmission time and arrival time.

[0003] Existing path planning methods focus on Unmanned Aerial Vehicles (UAVs). They mainly include convex optimization methods, deep reinforcement learning methods, and heuristic algorithms. Convex optimization methods are the most common UAV path planning methods, such as Successive Convex Approximation (SCA), Sequential Convex Programming (SCP), Block Coordinate Descent (BCD), etc. By gradually transforming complex non-convex problems into a series of convex optimization sub-problems that are easier to solve and performing alternating optimization operations, the solution process is simplified. However, it may be difficult to transform some non-convex problems and non-convex constraints, and the optimization problem needs to meet certain structural properties (such as convexity or blockability). Deep reinforcement learning methods use deep learning networks for exploration and learning of trajectory planning strategies, such as Deep Q-Network (DQN), but have high requirements for training data in complex problems and high maintenance costs. Heuristic algorithms include Genetic Algorithm (GA), Differential Evolution Algorithm (DE), Ant Colony Optimization (ACO), etc., which can ignore the convexity and concavity of the optimization problem and are applicable to large-scale, non-linear, non-convex or discrete combinatorial optimization problems that are difficult to efficiently solve by traditional algorithms. However, in most cases, only sub-optimal solutions can be obtained. Summary of the Invention

[0004] The objective of the present invention is to solve the technical problem that the algorithms in the existing technology can only obtain sub-optimal solutions in most cases.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] A multi-target ship communication path planning method based on a marine waveguide channel map, comprising the following steps:

[0007] S1: In the case of long-distance transmission at sea, establish a single-input single-output ship motion model and communication model;

[0008] S2: Establish a CGM according to the existing gas phase data, evaporation duct model and electromagnetic wave propagation model;

[0009] S3: Propose an optimization problem model;

[0010] S4: Improve and combine the NSGA-II algorithm and the PSO algorithm to propose the NSGA-II-PSO-VS algorithm;

[0011] S5: Compare the operation results of the NSGA-II algorithm and the NSGA-II-PSO-VS algorithm at the same time; conduct simulation experiments under different evaporation duct heights to prove the rationality and applicability of the proposed algorithm.

[0012] Preferably, the specific steps of S4 are as follows:

[0013] S4.1: Take the direction angle of each time slot as a variable to establish a population,

[0014] S4.2: Calculate the fitness function value of each individual in the initial population, and sort the individuals according to non-dominated sorting and crowding degree calculation;

[0015] S4.3: Perform selection, crossover and mutation operations on the population, and generate a certain number of crossover and mutation offspring according to the crossover rate A c and the mutation rate A m and merge them with the original population to form a combined population Q,

[0016] S4.4: Perform non-dominated sorting on the combined population Q, sort the individuals of the same domination level according to the crowding distance, and fill the sorted combined population into the next generation population according to the number of the initial population;

[0017] S4.5: When the algorithm iterates to a certain number of times, enter the modified particle swarm optimization algorithm. First, use some of the previous parameters and set the specific parameters of the particle swarm optimization algorithm;

[0018] S4.6: Calculate the fitness function value of the population, establish an external repository, store the position information and fitness value in the repository, and initialize the repository, only retaining the individuals with the highest domination level;

[0019] S4.7: Update the velocity and position information of the individual;

[0020] S4.8: Repeat this process until the maximum number of iterations is reached.

[0021] Preferably, when establishing the initial population in S4.1, a certain number of randomly initialized individuals are first generated randomly, and then a certain number of individuals satisfying the constraints are generated according to the angle limit of each time slot, and the two are combined into an initial population containing Np individuals.

[0022] Preferably, in S4.2, the crowding distance Dis i of the i-th individual A i is:

[0023]

[0024] In the above formula, N obj is the number of objective functions, the population is sorted according to the value of the j-th objective function, f j (i + 1) and f j (i - 1) are respectively the objective function values of the previous individual and the next individual of the i-th individual, f j (A max ) and f j (A min ) are respectively the maximum value and the minimum value of the j-th objective function.

[0025] Preferably, in S4.3, the individual selection method adopts tournament selection. First, individuals with a higher dominance level are selected according to the dominance level. When the dominance levels are the same, individuals with a lower crowding degree are selected according to the crowding distance.

[0026] The crossover method is simulated binary crossover, and the next generation of individuals is obtained by crossing two parent individuals:

[0027]

[0028] In the above formula, and are two offspring individuals, and are two parent individuals, and β is expressed as

[0029]

[0030] In the above formula, η c is a user-defined crossover distribution parameter, and U is a random number between 0 and 1;

[0031] During the mutation process, polynomial mutation is adopted and the original method is modified:

[0032]

[0033] where A u is the upper limit of individual variation, and A l is the lower limit of individual variation. μ is expressed as:

[0034]

[0035] In the above formula, η m is the mutation distribution parameter. μ 2 is and the smaller value of the comparison between the two. In addition, Γ is expressed as:

[0036]

[0037] where V represents the dimensionality size of the individual. In the initial population, V = M, and subsequently, in the iteration of the population, V is made equal to the largest M in the population 2 , and the dimensions exceeding V are ignored during processes such as crossover and mutation. Therefore, the dimensionality of the individuals in the population will change continuously with the increase in the number of iterations. Due to the use of the dimensionality change mechanism, in the case of dimensionality reduction, the value gradually increases;

[0038] As the number of iterations increases, the mutation probability also increases, thereby increasing the possibility of jumping out of the local optimal solution. Before the mutation operation, considering that the speed angle of the ship will not mutate suddenly, a part of the parent generation is randomly selected for smoothing operation, that is

[0039] Preferably, the speed and position information in S4.7 are defined as follows:

[0040]

[0041] A α+1 = A α + v α

[0042] And a penalty function is set according to the constraint conditions. Calculate the new fitness function. If the constraint is violated, the penalty function is added to the fitness function; update the external repository with the new individual, add the new individual to the repository, and perform non-dominated sorting to delete the dominated individuals from the repository; then, sort the individuals in the updated repository according to the crowding distance, and select the part with the largest crowding distance as the historical optimal position candidate.

[0043] Preferably, the specific steps of S1 are as follows: Set the position of the base station as (0, 0, 0), establish a three-dimensional coordinate system with the base station as the origin, and set the antenna position of the base station as (0, 0, h r ), where h r is the height of the antenna. Consider the height of the ship as a moving point in three-dimensional space. Therefore, assume that the ship starts from the initial position (x 0 , y 0 , h 0 ) and heads to the key position (x d , y d , h d ). The position of the ship at any time t is:

[0044] X(t) = (x s (t), y s (t), h s (t)), 0 ≤ t ≤ T

[0045] In the above formula, T is the maximum sailing time. The sailing speed of the ship at any time t is denoted by v(t), and the speed direction angle is denoted by . Therefore, the speeds of the ship in the x-direction and y-direction are respectively expressed as:

[0046]

[0047] Assume that the ship starts from the initial position at an initial direction angle . Its turning ability is restricted as follows:

[0048]

[0049] In the above formula, δt is the time difference between the current moment and the previous moment, is the maximum turning angle of the ship between these two moments, and its magnitude depends on the magnitude of δt. At any time t, the distance between the ship and the base station is expressed as:

[0050]

[0051] Use h to denote the channel fading coefficient and x to denote the transmitting channel. Therefore, the signal y received at the base station is expressed as:

[0052] y = hx + n

[0053] where n is the noise vector. The channel fading coefficient h consists of the large-scale fading coefficient h L and the small-scale fading coefficient h S , and is expressed as:

[0054] h = h L h S

[0055] Ignore small-scale fading and only consider large-scale fading; assume the transmit power is P t , then the received power P r can be expressed as:

[0056] P r =|h L | 2 P t =L p G t G r P t

[0057] In the above formula, L p is the path loss, which is determined by the distance d(t) between the base station and the ship, G t is the transmit antenna gain, G r is the receive antenna gain. In free space, the path loss L p is defined as:

[0058] L p =L fs =32.45 + 20lg(d(t)) + 20lg(f)

[0059] In the above formula, the unit of the distance d(t) is kilometers, and f is the carrier frequency with the unit of MHz. Under the influence of evaporation duct, the path loss at sea shows different characteristics. Using the parabolic equation method, the propagation of electromagnetic waves in the evaporation duct environment is modeled, and its path loss L p can be expressed as

[0060]

[0061] In the above formula, u represents the field strength, r represents the radius of the earth, and λ represents the carrier wavelength. Since the transmission distance is usually very small compared to the radius of the earth, the above formula can be simplified to:

[0062] L ed = - 20lg|u| + 20lg(4π) + 10lg(d(t)) - 30lg(λ)

[0063] Therefore, the channel capacity is

[0064]

[0065] In the above formula, B is the channel bandwidth, n 0 is the noise power spectral density. Therefore, at any time t, the maximum achievable transmission rate R(t) can be expressed as:

[0066]

[0067] Preferably, in S2, the parabolic equation method is used to construct the CGM in the evaporation duct environment. Since the position of the ship at any time t is X(t), the large-scale channel gain at this position is defined as

[0068] The space is divided into many grids, and the width of each grid is Δd. The channel state remains unchanged within this grid. The selected Δd should have an appropriate size to ensure that the channel gain difference between adjacent grids is not too large. The grid height Δh in the vertical direction should be less than the grid width Δd in the horizontal direction;

[0069] Assume that the farthest transmission distance from the base station is N le Δd, where N le is an integer, and the obtained CGM is symmetric about the center of the base station. The height of the evaporation duct is h eva , then h eva = N ve Δh, where, N ve is also an integer. The coordinates of each network can be expressed as:

[0070] s(i, j, k) = (iΔd, jΔd, kΔh), i ∈ I N , j ∈ J N , k ∈ K N

[0071] In the above formula, I N = {-N le , -N le +1, …, N le -1, N le}, I N is the same as J N , K N = {0, 1, …, N ve -1, N ve}, and the channel gain at any position s(i, j, k) is equal to the received power P r in the formula of L p . Therefore, the large-scale channel gain between the base station and the ship can be obtained according to the grid where the ship's position is located at any time.

[0072] Preferably, in S3, the goal is to optimize the ship's trajectory to minimize the ship's transmission time and arrival time. The minimum transmission time is defined as T1, the minimum arrival time is defined as T2, and the ship's trajectory is expressed as K = {X(t), t ∈ [0, T 2}, and the optimization problem is expressed as

[0073] P1):

[0074] such that C 1 : T 1 ≤ T 2

[0075] C 2 : T 2 ≤ T

[0076] C 3 :

[0077] C 4 : X(0) = (x 0 , y 0 , h 0 )

[0078] C 5 : X(T 2 ) = (x d , y d , h d )

[0079] C 6 :

[0080] In the above formula, C 1 means that the ship needs to transmit all data before reaching the destination, and the transmission time is less than the arrival time;

[0081] C 2 means that the sailing time of the ship is less than the specified time;

[0082] C 3 means that the amount of data transmitted within the transmission time is greater than or equal to the amount of data to be transmitted;

[0083] C 4 means the starting point constraint;

[0084] C 5 means the ending point constraint;

[0085] C 6 means the turning constraint of the ship;

[0086] The constraint conditions C 3 and C 6 are difficult to express in the case of continuous time. Therefore, the entire sailing process is decomposed into multiple time slots, the length of each time slot is Δt, the number of time slots for data transmission is M 1 , the number of time slots required to reach the destination is M 2 , and the number of time slots for the maximum sailing time is The trajectory of the ship is K = {X s (0), Xs (1), …, X s (M 2 )}, the position formula of the ship at any time t and the formula of the limitation on the ship's steering ability are modified to:

[0087] X(i) = (x s (i), y s (i), h s (i)), 0 ≤ i ≤ M 2

[0088]

[0089] Regarding the ship's speed as uniform and ignoring the influence of sea wave fluctuations, therefore, the displacement within a single time slot is expressed as:

[0090] ||X(i + 1) - X(i)|| = vΔt, 0 ≤ i ≤ M 2 -1

[0091] The optimization problem (P1) is rewritten as:

[0092] P2):

[0093] s.t. C 1 : M 1 ≤ M 2

[0094] C 2 : M 2 ≤ M

[0095] C 3 :

[0096] C 4 : X(0) = (x 0 , d 0 , h 0 )

[0097] C 5 : X(M 2 ) = (x d , y d , h d )

[0098] C 6 :

[0099] Preferably, after completing S4, the running results of the NSGA-II algorithm and the NSGA-II-PSO-VS algorithm at the same time are compared; simulation experiments are carried out under different evaporation duct heights to prove the rationality and applicability of the proposed algorithm.

[0100] A multi - target propagation path planning method based on a marine evaporation duct channel map as described above models the movement and communication of ships for the data - driven multi - target path planning problem of obtaining user information from the channel map, proposes an improved heuristic algorithm, improves and combines two traditional intelligent algorithms, and shows good performance in many aspects compared with traditional algorithms, providing help for the data - driven multi - target path planning problem.

[0101] Compared with the existing technologies, the present invention has the following beneficial effects:

[0102] Generally speaking, the present invention adopts a heuristic algorithm, the NSGA - II - PSO - VS algorithm, which is improved based on the NSGA - II algorithm, and provides a solution for the data - driven ship trajectory planning. This algorithm shows better performance than the original algorithm in terms of convergence speed, number of solutions, and solution range. BRIEF DESCRIPTION OF THE DRAWINGS

[0103] Figure 1 It is a system block diagram of a multi - target ship communication path planning method based on a marine waveguide channel map in an embodiment of the present invention;

[0104] Figure 2 It is a channel gain map constructed in a multi - target ship communication path planning method based on a marine waveguide channel map in an embodiment of the present invention;

[0105] Figure 3 It is a comparison chart of the results of the NSGA - II algorithm and the proposed NSGA - II - PSO - VS algorithm in an embodiment of the present invention;

[0106] Figure 4 It is a trajectory map of the optimal points at different waveguide heights in an embodiment of the present invention. The selected optimal points are the points where the arrival time and the transmission time weights are the same, that is, the points closest to the origin. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0107] The following further describes the present invention in detail with reference to specific embodiments.

[0108] Please refer to Figure 1 , a multi - target ship communication path planning method based on a marine waveguide channel map, comprising the following steps:

[0109] S1: In the case of long - distance transmission at sea, establish a single - input single - output ship motion model and communication model;

[0110] Specifically, in one embodiment, the position of the base station is set to (0, 0, 0), and a three-dimensional coordinate system is established with the base station as the origin. The antenna position of the base station is set to (0, 0, h r ), where h r is the height of the antenna. Regarding the height of the ship as a moving point in three-dimensional space, therefore, assume that the ship departs from the initial position (x 0 , y 0 , h 0 ) and heads to the key position (x d , y d , h d ). The position of the ship at any time t is:

[0111] X(t) = (x s (t), y s (t), h s (t)), 0 ≤ t ≤ T (1)

[0112] In the above formula, T is the maximum sailing time. The sailing speed of the ship at any time t is denoted by v(t), and the speed direction angle is denoted by . Therefore, the speeds of the ship in the x-direction and y-direction are respectively expressed as:

[0113]

[0114] Due to the influence of factors such as the volume, draft, and speed of the ship, the maneuverability of the ship is restricted during turning. Assume that the ship departs from the initial position with an initial direction angle . Its turning ability is restricted as follows:

[0115]

[0116] In the above formula, δt is the time difference between the current moment and the previous moment, is the maximum turning angle of the ship between these two moments, and its magnitude depends on the magnitude of δt. At any time t, the distance between the ship and the base station is expressed as:

[0117]

[0118] Since the communication system is a single-input single-output system, using h to denote the channel fading coefficient and x to represent the transmitting channel, the signal y received at the base station is expressed as:

[0119] y = hx + n (6)

[0120] Among them, n is the noise vector, and the channel fading coefficient h consists of the large-scale fading coefficient h L and the small-scale fading coefficient h S , and is expressed as:

[0121] h = h L h S (7)

[0122] In the case of long - distance transmission over the vast ocean and a single - antenna system, compared with large - scale fading, the impact of small - scale fading is very small. Therefore, small - scale fading is ignored and only large - scale fading is considered. Assume the transmit power is P t , then the received power P r can be expressed as:

[0123] P r = |h L | 2 P t = L p G t G r P t (8)

[0124] In the above formula, L p is the path loss, which is determined by the distance d(t) between the base station and the ship. G t is the transmit - antenna gain, and G r is the receive - antenna gain. In free space, the path loss L p is defined as:

[0125] L p = L fs = 32.45 + 20lg(d(t)) + 20lg(f) (9)

[0126] In the above formula, the unit of the distance d(t) is kilometers, and f is the carrier frequency with the unit of MHz. Under the influence of the evaporation duct, the path loss at sea shows different characteristics. Using the parabolic - equation method, the propagation of electromagnetic waves in the evaporation - duct environment is modeled, and its path loss L p can be expressed as

[0127]

[0128] In the above formula, u represents the field strength, r represents the radius of the earth, and λ represents the carrier wavelength. Since the transmission distance is usually very small compared with the radius of the earth, the above formula can be simplified to:

[0129] L ed = - 20lg|u| + 20lg(4π + 10lg(d(t)) - 30lg(λ) (11)

[0130] Therefore, the channel capacity is

[0131]

[0132] In the above formula, B is the channel width, and n 0 is the noise power spectral density. Therefore, at any time t, the maximum achievable transmission rate R(t) can be expressed as:

[0133]

[0134] S2: Establish the CGM based on the existing gas-phase data, evaporation duct model, and electromagnetic wave propagation model; specifically, please refer to Figure 2 , in one embodiment, the parabolic equation method is used to construct the CGM in the evaporation duct environment. Since the position of the ship at any time t is X(t), the large-scale channel gain at this position is defined as

[0135] Due to the wireless continuity of space and the limitation of storage capacity, it is impossible to store the large-scale channel gains at all positions. Therefore, it is necessary to divide the space into many grids, each with a width of Δd, and the channel state remains unchanged within the grid. Since there are no obstacles in the long-distance transmission at sea in the evaporation duct environment, the large-scale channel gain in the horizontal direction does not change significantly over a short distance. Therefore, the selected Δd should be of an appropriate size to ensure that the difference in channel gain between adjacent grids is not too large.

[0136] Compared with the long-distance transmission in the horizontal direction, because the typical height range of the marine evaporation duct is between 10 and 40 meters, the change in the heights of the transmitting and receiving antennas will significantly affect the large-scale channel gain. Therefore, the grid height Δh in the vertical direction should be less than the grid width Δd in the horizontal direction;

[0137] Assume that the farthest transmission distance from the base station is N le Δd, where N le is an integer, and the obtained CGM is symmetric about the center of the base station. The height of the evaporation duct is h eva , then h eva = N ve Δh, where N ve is also an integer. The coordinates of each network can be expressed as:

[0138] s(i,j,k) = (iΔd,jΔd,kΔh), i ∈ I N , j ∈ J N , k ∈ K N (14)

[0139] In the above formula, I N = {-N le , -N le +1, …, N le -1, N le}, IN Same as J N and K N ={0, 1, …, N ve -1, N ve}, and the channel gain at any position s(i, j, k) is equal to L in Equation (8) p . Therefore, the large-scale channel gain between the base station and the ship can be obtained according to the grid where the ship's position is located at any time.

[0140] S3: Propose an optimization problem model;

[0141] Our goal is to optimize the ship's trajectory to minimize the ship's transmission time and arrival time. Define the minimum transmission time as T1, the minimum arrival time as T2, and the ship's trajectory is represented as K = {X(t), t ∈ [0, T 2}, and the optimization problem is expressed as

[0142] (P1):

[0143] s.t. C 1 : T 1 ≤ T 2

[0144] C 2 : T 2 ≤ T

[0145] C 3 :

[0146] C 4 : X(0) = (x 0 , y 0 , h 0 )

[0147] C 5 : X(T 2 ) = (x d , y d , h d )

[0148] C 6 :

[0149] In the above formula, C 1 means that the ship needs to transmit all data before reaching the destination, and the transmission time is less than the arrival time;

[0150] C 2 means that the ship's sailing time is less than the specified time;

[0151] C 3It means that the amount of data transmitted within the transmission time should be greater than or equal to the amount of data to be transmitted;

[0152] C 4 It means the starting point constraint;

[0153] C 5 It means the ending point constraint;

[0154] C 6 It means the steering constraint of the ship;

[0155] Constraint condition C 3 and C 6 It is difficult to express in the case of continuous time. Therefore, the entire navigation process is decomposed into multiple time slots. The length of each time slot is Δt, and the number of time slots for data transmission is M 1 , the number of time slots required to reach the destination is M 2 , the number of time slots for the maximum navigation time is The trajectory of the ship is K = {X s (0), X s (1), …, X s (M 2 )}, and formulas (1) and (4) are modified to:

[0156] X(i) = (x s (i), y s (i), h s (i)), 0 ≤ i ≤ M 2 (16)

[0157]

[0158] In addition, the trajectory of the ship is approximately a straight line within a single time slot, where the angular direction of the speed remains unchanged. To simplify the problem, the speed of the ship is regarded as uniform, and the influence of wave fluctuations is ignored. Therefore, the displacement within a single time slot is expressed as:

[0159] ||X(i + 1) - X(i)|| = vΔt, 0 ≤ i ≤ M 2 -1 (18)

[0160] The optimization problem (P1) is rewritten as:

[0161] P2):

[0162] s.t.C 1 :M 1 ≤ M 2

[0163] C 2 :M 2 ≤ M

[0164] C 3 :

[0165] C 4 : X(0) = (x 0 , d 0 , h 0 )

[0166] C 5 : X(M 2 ) = (x d , y d , h d )

[0167] C 6 :

[0168] S4: Improve and combine the NSGA-II algorithm and the PSO algorithm to propose the NSGA-II-PSO-VS algorithm;

[0169] In one embodiment, S4 includes the following steps:

[0170] S4.1: Use the direction angle of each time slot as a variable to establish a population. When establishing the initial population, first randomly generate a certain number of randomly initialized individuals, and then generate a certain number of individuals that meet the constraints according to the angle limit of each time slot, and combine the two into an initial population containing N p individuals.

[0171] S4.2: Calculate the fitness function value of each individual in the initial population, and sort the individuals according to non-dominated sorting and crowding degree calculation. The crowding distance Dis i of the i-th individual A i is:

[0172]

[0173] In the above formula, N obj is the number of objective functions, the population is sorted according to the value of the j-th objective function, f j (i + 1) and f j (i - 1) are respectively the objective function values of the previous individual and the next individual of the i-th individual, f j (A max ) and f j (A max ) are respectively the maximum and minimum values of the j-th objective function.

[0174] S4.3: Perform selection, crossover, and mutation operations on the population, and according to the crossover rate A c and the mutation rate A mGenerate a certain number of crossover and mutation offspring, and merge them with the original population to form a combined population Q.

[0175] The individual selection method adopts tournament selection. First, select individuals with a higher dominance level according to the dominance level. When the dominance levels are the same, select individuals with a lower crowding distance according to the crowding distance.

[0176] The crossover method is simulated binary crossover (SBX). Obtain the next-generation individuals by crossing two parent individuals:

[0177]

[0178] In the above formula, and are two offspring individuals, and are two parent individuals, and β can be expressed as

[0179]

[0180] In the above formula, η c is a user-defined crossover distribution parameter, and U is a random number between 0 and 1. During the mutation process, polynomial mutation is adopted and the original method is modified:

[0181]

[0182] where A u is the upper limit of the change of the individual, A l is the lower limit of the change of the individual, and μ is expressed as:

[0183]

[0184] In the above formula, η m is the mutation distribution parameter, μ 2 is and the smaller value of the two comparisons. In addition, Γ is expressed as:

[0185]

[0186] where V represents the dimension size of the individual. In the initial population, V = M. Subsequently, in the iteration of the population, make V equal to the largest M in the population 2 , and ignore the dimensions exceeding V during the processes of crossover, mutation, etc. Therefore, the dimension of the individuals in the population will change continuously with the increase of the number of iterations. Due to the use of the dimension change mechanism, in the case of dimension reduction, The value gradually increases. As the number of iterations increases, the mutation probability also increases, thus enhancing the likelihood of jumping out of the local optimal solution. Before the mutation operation, considering that the speed angle of the ship does not undergo sudden changes, a part of the parents is randomly selected for smoothing operation, that is

[0187] S4.4: Perform non - dominated sorting on the combined population Q, and sort the individuals in the same domination level according to the crowding distance. The sorted combined population is filled into the next - generation population according to the number of the initial population.

[0188] S4.5: When the algorithm iterates to a certain number of times, enter the modified particle swarm optimization algorithm. First, use some of the previous parameters and set the specific parameters of the particle swarm optimization algorithm.

[0189] S4.6: Calculate the fitness function value of the population and establish an external repository. Store the position information and fitness value in the repository, and initialize the repository, only retaining the individuals with the highest domination level.

[0190] S4.7: Update the velocity and position information of the individuals

[0191]

[0192] A α+1 = A α + v α (27)

[0193] And set the penalty function according to the constraint conditions. Calculate the new fitness function. If the constraint is violated, add the penalty function to the fitness function. Update the external repository with the new individuals, add the new individuals to the repository, and perform non - dominated sorting, deleting the dominated individuals from the repository. Then, sort the individuals in the updated repository according to the crowding distance and select the part with the largest crowding distance as the historical optimal position candidate.

[0194] S4.8: Repeat this process until the maximum number of iterations is reached.

[0195] S5: Compare the running results of the NSGA - II algorithm and the NSGA - II - PSO - VS algorithm at the same time; conduct simulation experiments under different evaporation duct heights to prove the rationality and applicability of the proposed algorithm.

[0196] Please refer to Figure 3 and Figure 4, in S5, the NSGA-II algorithm and the proposed NSGA-II-PSO-VS algorithm are respectively run for the same time with the same parameters. Four cases with evaporation duct heights of 20m, 25m, 30m, and 35m are selected, and the obtained results are compared.

[0197] Next, simulation experiments are carried out again under these four duct heights. Since a set of optimal solutions are obtained, the same weights are given to the transmission time and the arrival time, and the optimal points are selected to obtain the trajectory diagram.

[0198] In summary, a multi-objective propagation path planning method based on the offshore evaporation duct channel map disclosed in this application models the movement and communication of ships for the data-driven multi-objective path planning problem of obtaining user information according to the channel map, and proposes an improved heuristic algorithm, which improves and combines two traditional intelligent algorithms. Compared with the traditional algorithms, it shows good performance in many aspects and provides help for the data-driven multi-objective path planning problem.

Claims

1. A multi-target ship communication path planning method based on a marine waveguide channel map, characterized in that: The following steps are involved: S1: In the case of long-distance transmission at sea, a single-input single-output ship motion model and communication model are established; S2: Establish CGM based on the existing gas phase data, evaporation waveguide model and electromagnetic wave propagation model; S3: Propose an optimization problem model; S4: Improve and combine the NSGA-II algorithm and the PSO algorithm, and propose the NSGA-II-PSO-VS algorithm; S5: Compare the running results of NSGA-II algorithm and NSGA-II-PSO-VS algorithm under the same time; conduct simulation experiments at different evaporation waveguide heights to prove the rationality and applicability of the proposed algorithm.

2. The multi-target ship communication path planning method based on the marine waveguide channel map according to claim 1 is characterized in that: The specific steps of S4 are as follows: S4.1: Take the direction angle of each time slot as a variable and establish a population. S4.2: Calculate the fitness function value of each individual in the initial population, and sort the individuals according to the non-dominated sorting and crowding calculation; S4.3: Perform selection, crossover and mutation operations on the population, according to the crossover rate A c and mutation rate A m Generate a certain number of crossover and mutation offspring and merge them with the original population to form a combined population Q, S4.4: Perform non-dominated sorting on the combined population Q, sort the individuals of the same dominance level according to the crowding distance, and fill the sorted combined population into the next generation population according to the number of the initial population; S4.5: When the algorithm iterates a certain number of times, it enters the modified particle swarm optimization algorithm. First, some of the previous parameters are used and the specific parameters of the particle swarm optimization algorithm are set; S4.6: Calculate the fitness function value of the population, establish an external repository, store the position information and fitness value in the repository, and initialize the repository to retain only the individuals with the highest dominance level; S4.7: Update individual speed and position information; S4.8: Repeat this process until the maximum number of iterations is met.

3. The multi-target ship communication path planning method based on the marine waveguide channel map according to claim 2 is characterized in that: When establishing the initial population in S4.1, a certain number of randomly initialized individuals are first randomly generated, and then a certain number of individuals that meet the constraints are generated according to the angle limit of each time slot, and the two are combined into an initial population containing Np individuals.

4. The multi-target ship communication path planning method based on marine waveguide channel map according to claim 3 is characterized in that: S4.2 defines the i-th individual A i The crowding distance Dis i for: In the above formula, N obj is the number of objective functions, the population is sorted according to the value of the jth objective function, f j (i+1) and f j (i-1) are the objective function values ​​of the previous individual and the next individual, respectively, f j (A max ) and f j (A min ) are the maximum and minimum values ​​of the j-th objective function respectively.

5. The multi-target ship communication path planning method based on marine waveguide channel map according to claim 4 is characterized in that: The individual selection method in S4.3 adopts tournament selection. First, individuals with higher levels are selected according to the dominance level. When the dominance levels are the same, individuals with lower crowding are selected according to the crowding distance. The crossover method simulates binary crossover, and the next generation individuals are obtained by crossing two parent individuals: In the above formula, and For two offspring individuals, and are two parent individuals, and β is expressed as In the above formula, η c is a custom crossover distribution parameter, and U is a random number between 0 and 1; In the mutation process, polynomial mutation was adopted and the original method was modified: Among them A u is the upper limit of individual change, A l is the lower limit of individual variation, and η is expressed as: In the above formula, η m is the variation distribution parameter, μ2 is and The smaller value of the two, in addition, Γ is expressed as: Among them, V represents the dimension size of the individual. In the initial population, V = M. In the subsequent iteration of the population, V is made equal to the largest M2 in the population, and the dimensions exceeding V are ignored in the process of crossover and mutation. Therefore, the dimension of the individual in the population will continue to change with the increase of the number of iterations. Due to the use of the dimension change mechanism, when the dimension is reduced, The value of gradually increases; As the number of iterations increases, the probability of mutation also increases, thereby increasing the possibility of jumping out of the local optimal solution. Before the mutation operation, considering that the speed angle of the ship will not produce a sudden change, a part of the parent generation is randomly selected for smoothing operation, that is, 6. The multi-target ship communication path planning method based on marine waveguide channel map according to claim 5 is characterized in that: The speed and position information in S4.7 are defined as follows: And α+1 =A α +v α And set the penalty function according to the constraints. Calculate the new fitness function, and if the constraints are violated, add the penalty function to the fitness function; Update the external repository with new individuals, add the new individuals to the repository, and perform non-dominated sorting to remove dominated individuals from the repository; then, sort the individuals in the updated repository according to the crowding distance, and select the part with the largest crowding distance as the historical optimal position Candidate.

7. The multi-target ship communication path planning method based on marine waveguide channel map according to claim 6 is characterized in that: The specific steps of S1 are as follows: the position of the base station is set to (0,0,0), a three-dimensional coordinate system is established with the base station as the origin, and the antenna position of the base station is set to (0,0,h r ), where h r is the height of the antenna, and the height of the ship is regarded as the moving store in three-dimensional space. Therefore, suppose the ship starts from the initial position (x0, y0, h0) and moves to the key position (x d ,y d ,h d ), the position of the ship at any time t is: X(t)=(x s (t),y s (t),h s (t)),0≤t≤T In the above formula, T is the maximum sailing time, the sailing speed of the ship at any time t is represented by v(t), and the speed direction angle is represented by Therefore, the speed of the ship in the x direction and y direction are expressed as: Assume that the ship has an initial direction angle From this initial position, its ability to turn is limited as follows: In the above formula, δt is the time difference between the current moment and the previous moment. is the maximum turning angle of the ship between these two moments, and its size depends on the size of δt. At any time t, the distance between the ship and the base station is expressed as: Use h to identify the channel fading coefficient, and x to represent the transmission channel. Therefore, the signal y received at the base station is expressed as: y=hx+n Where n is the noise vector, and the channel fading coefficient h has a large-scale fading coefficient h L and the small-scale fading coefficient h S , composed of, expressed as: h=h L h S Ignore small-scale fading and only consider large-scale fading; assume that the transmission power is P t , then the received power P r It can be expressed as: P r =|h L | 2 P t =L p G t G r P t In the above formula, L p is the path loss, which is determined by the distance d(t) between the base station and the ship, G t is the transmitting antenna gain, G r is the receiving antenna gain, and in free space, the path loss L p Defined as: L p =L fs =32.45+20lg(d(t))+20lg(f) In the above formula, the distance d(t) is in kilometers, and f is the carrier frequency in MHz. Under the influence of the evaporation duct, the path loss at sea shows different characteristics. The propagation of electromagnetic waves in the evaporation duct environment is modeled using the parabolic equation method, and its path loss L is p It can be expressed as In the above formula, u represents the field strength, r represents the radius of the earth, and λ represents the carrier wavelength. Since the transmission distance is usually very small relative to the radius of the earth, the above formula can be simplified to: L ed =-20lg|u|+20lg(4π)+10lg(d(t))-30lg(λ) So the channel capacity is In the above formula, B is the channel width and n0 is the noise power spectrum density. Therefore, at any time t, the maximum achievable transmission rate R(t) can be expressed as:

8. The multi-target ship communication path planning method based on marine waveguide channel map according to claim 7 is characterized in that: In S2, the parabolic equation method is used to construct the CGM in the evaporative waveguide environment. Since the position of the ship at any time t is X(t), the large-scale channel gain at this position is defined as Divide the space into many grids, each with a width of Δd. The channel state remains unchanged within the grid. The selected Δd should have an appropriate size to ensure that the channel gain difference between adjacent grids is not too large. The vertical grid height Δh should be smaller than the horizontal grid width Δd. Assume that the maximum transmission distance from the base station is N le Δd, where N le is an integer, the obtained CGM is symmetrical about the center of the base station, and the height of the evaporation waveguide is h eva , then h eva =N ve Δh, where N ve is also an integer, and the coordinates of each network can be expressed as: s(i,j,k)=(iΔd,jΔd,kΔh),i∈I N ,j∈J N ,k∈K N In the above formula, I N ={-N le ,-N le +1,…,N le -1,N le },I N With J N Same, K N ={0,1,…,N ve -1,N ve }, the channel gain at any position s(i,j,k) is Equal to the received power P r L in the formula p ,Therefore, the large-scale channel gain between the base station and the ship can be obtained according to the grid where the ship’s position is at any time.

9. The multi-target ship communication path planning method based on marine waveguide channel map according to claim 8 is characterized in that: The goal of S3 is to optimize the trajectory of the ship to minimize the ship's transmission time and arrival time. The minimum transmission time is defined as T1, the minimum arrival time is defined as T2, the trajectory of the ship is represented by K = {X(t), t∈[0, T2]}, and the optimization problem is expressed as stC1:T1≤T2 C2:T2≤T C4: X(0) = (x0, y0, h0) C5:X(T2)=(x d ,y d ,h d ) In the above formula, C1 means that the ship must transmit all data before reaching the destination, and the transmission time is less than the arrival time; C2 means that the ship's sailing time must be less than the specified time; C3 means that the amount of data transmitted within the transmission time must be greater than or equal to the amount of data required to be transmitted; C4 represents the starting point constraint; C5 represents the end point constraint; C6 represents the ship's steering constraint; Constraints C3 and C6 are difficult to express in continuous time, so the entire navigation process is decomposed into multiple time slots. The length of each time slot is Δt. The number of time slots used for data transmission is M1, the number of time slots required to reach the destination is M2, and the number of time slots for the maximum navigation time is The trajectory of the ship is K = {X s (0),X s (1), …, X s (M2)}, the formula for the position of the ship at any time t and the formula for the limitation of the ship's steering ability are modified to: X(i)=(x s (i),y s (i),h s (i)),0≤i≤M2 The speed of the ship is considered to be uniform, and the influence of the wave fluctuation is ignored. Therefore, the displacement in a single time slot is expressed as: ||X(i+1)-X(i)||=vΔt, 0≤i≤M2-1 The optimization problem (P1) can be rewritten as: stC1:M1≤M2 C2:M2≤M C4: X(0) = (x0, y0, h0) C5:X(M2)=(x d ,y d ,h d ) 10. The multi-target ship communication path planning method based on marine waveguide channel map according to claim 9 is characterized in that: After completing S4, the running results of the NSGA-II algorithm and the NSGA-II-PSO-VS algorithm under the same time are compared; simulation experiments are carried out at different evaporation waveguide heights to prove the rationality and applicability of the proposed algorithm.

Images