An oris-assisted unmanned aerial vehicle relay fso communication flight path planning method

Abstract

The application discloses an ORIS-assisted unmanned aerial vehicle relay FSO communication path planning method, establishes an unmanned aerial vehicle path optimization model to obtain an ORIS optimal phase shift and an optimal path as an objective function when a communication system traversal capacity is maximized; the optimization model is decomposed into a first sub-model for solving an ORIS phase shift optimization problem of the unmanned aerial vehicle under a fixed path and a second sub-model for solving a path problem of the unmanned aerial vehicle under a fixed ORIS phase shift, the first sub-model is solved by using a simulated annealing algorithm, the second sub-model is solved by using a DDQN algorithm, a Q network meeting a condition is obtained through repeated iteration, and finally, a path of the unmanned aerial vehicle meeting the condition is obtained according to the Q network. The application fully utilizes the maneuverability of the unmanned aerial vehicle and the controllability of the ORIS, can provide high-quality communication for mobile users under the condition of maximizing the traversal capacity, optimizes the phase shift of the ORIS by using the simulated annealing algorithm to reduce the influence of the directivity error, and optimizes the path of the unmanned aerial vehicle by using the DDQN algorithm, so that the optimal path of the unmanned aerial vehicle is obtained.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicle (UAV) communication, specifically relating to an ORIS-assisted UAV relay FSO communication trajectory planning method. Background Technology

[0002] Free-space optical communication boasts advantages such as high capacity, high spectrum utilization, strong security, and strong resistance to electromagnetic interference, making it an effective means to alleviate spectrum congestion in next-generation wireless communication. However, due to the propagation characteristics of laser light sources, the applicability of free-space optical communication is often limited when the line-of-sight path between the transmitter and receiver is blocked.

[0003] In recent years, drones have gained widespread attention as relays to address non-line-of-sight (NLS) issues in free-space optical communication due to their ease of deployment, high mobility, and cost-effectiveness. Moreover, compared to fixed ground relays, drone relays can adaptively adjust their location and resources to achieve better communication conditions, providing a flexible network structure and wide coverage. However, free-space optical communication (FSO) is highly susceptible to weather conditions and prone to pointing errors, which can lead to interruptions and disconnections when directly applied to drones. Optical reconfigurable intelligent surface (ORIS) technology can effectively overcome these challenges, providing a reliable and efficient solution for drone relay communication.

[0004] However, fully leveraging the maneuverability of UAVs and the adjustability of ORIS to establish relay communication systems remains an open question. Therefore, researching UAV relay free-space optical communication trajectory planning methods assisted by optically reconfigurable smart surfaces to provide high-quality communication services for mobile users is crucial for the full application of UAV and ORIS technologies in FSO communication. Summary of the Invention

[0005] Purpose of the invention: This invention provides an ORIS-assisted method for planning the FSO communication path of unmanned aerial vehicles (UAVs) relays, which can provide high-quality communication for mobile users while maximizing traversal capacity.

[0006] Technical solution: The ORIS-assisted unmanned aerial vehicle (UAV) relay FSO communication trajectory planning method of the present invention includes the following steps:

[0007] (1) Establish a UAV trajectory optimization model, wherein the objective function is to obtain the optimal ORIS phase shift and the optimal trajectory when the communication system traversal capacity is maximized, and the constraints are ORIS phase shift constraint, UAV flight trajectory constraint and UAV energy consumption constraint.

[0008] (2) The UAV trajectory optimization model is decomposed into a first sub-model and a second sub-model. The first sub-model takes the optimal phase shift scheme under the fixed position of the UAV as the objective function, and the second sub-model takes the optimal trajectory under the fixed phase shift of ORIS as the objective function.

[0009] (3) Using the track points in the second sub-model as known track points, the first sub-model is solved using the simulated annealing algorithm to obtain the optimal ORIS phase shift scheme;

[0010] (4) Using the optimal ORIS phase shift scheme obtained from the first sub-model as the known scheme, solve the second sub-model according to the DDQN algorithm to obtain the track points of the UAV in the next time slot;

[0011] (5) Determine if the drone's energy is depleted. If not, return to step (3); if yes, proceed to step (6).

[0012] (6) Repeat steps (4) to (5) iteratively, and the online Q network is updated after each iteration until the number of iterations reaches the preset value. Based on the Q network trained after the iteration, the action with the highest value in each time slot is selected to form the UAV track.

[0013] Furthermore, the UAV trajectory optimization model described in step (1) is as follows:

[0014]

[0015] In the formula, P1 represents the UAV trajectory optimization model. This represents the traversal capacity of the time slot t system; This represents the optimal ORIS phase shift scheme, specifically the optimal reflection angle of ORIS in each time slot. The optimal flight path is specifically the set of optimal flight path points for each time slot of the UAV. This represents the rotation angle of ORIS along the x-axis in time slot t. This represents the rotation angle of ORIS along the y-axis in time slot t; Indicates the initial value of the waypoint. The three-dimensional coordinates of the initial flight path of the UAV; This represents the speed of the UAV in time slot t. and These are the minimum and maximum speeds of the drone; the drone's flight altitude. Limited to minimum height and maximum height between; The drone has already consumed energy. This indicates the maximum energy consumption of the drone; This represents the acceleration of the UAV in time slot t. and These represent the minimum and maximum accelerations of the drone, respectively.

[0016] Furthermore, the system traversal capacity in step (1) is:

[0017]

[0018] In the formula, for The probability density function, for:

[0019]

[0020] in, The channel coefficient is:

[0021]

[0022] In the formula, , and These are atmospheric attenuation, atmospheric turbulence, and directionality error, respectively.

[0023] Furthermore, the atmospheric attenuation coefficient is:

[0024]

[0025] in, and They represent visibility and... The atmospheric attenuation coefficients for sunny and foggy days, and These represent the lengths the beam travels through in clear sky and fog, respectively.

[0026] Furthermore, the atmospheric turbulence uses a Gamma-Gamma distribution model, which is:

[0027]

[0028] In the formula, It is a modified Bessel function of the second kind. It is the gamma equation. Let represent the number of small-turbulence and large-turbulence units, respectively, as follows:

[0029]

[0030]

[0031] Among them, the Rytov variance was calculated. , It is a refractive index structural parameter. For the total propagation length, λ is the wavelength of the optical signal.

[0032] Furthermore, the pointing error refers to the misalignment vector between the center point of the photodetector at the receiving end and the beam footprint. The error is caused by the laser power distribution on the photodetector. The center point of the beam footprint is calculated using a rotation matrix combined with specular reflection. The normalized optical power spatial distribution at the communication distance D is approximately:

[0033]

[0034] In the formula, , They are in The light intensity at point = 0, and the beam divergence angle are... The beam waist at communication distance D; if the communication link is long enough, the receiving aperture... The radius is much smaller than The receiving aperture area Assuming the light intensity is constant, the geometric loss factor caused by the directivity error is:

[0035] .

[0036] Furthermore, the first sub-model in step (2) specifically refers to:

[0037] .

[0038] Furthermore, the implementation process of step (3) is as follows:

[0039] Set the initial annealing temperature, the minimum annealing temperature, and initialize the current phase shift. In each iteration, a new phase shift is generated by perturbing the current phase shift. Calculate the traversal capacity under the current phase shift and the new phase shift respectively, and determine whether to accept the new solution using an acceptance probability, which is:

[0040]

[0041] in, , The current temperature is used; as the iteration progresses, the temperature gradually decreases through the cooling coefficient. When the temperature falls below the minimum temperature, the iteration terminates, and the maximum traversal capacity at the current track point is output.

[0042] Furthermore, the implementation process of step (4) is as follows:

[0043] Objective: To maximize the traversal capacity of the system under the optimal ORIS phase shift scheme obtained from the first sub-model;

[0044] action The action selected by the time slot agent is specifically a combination of the component accelerations of the UAV along the three-dimensional coordinate axes, that is, a combination of the x-axis component acceleration, y-axis component acceleration, and z-axis component acceleration taking values ​​from the set {-2, 0, 2} respectively.

[0045] state The time slot t contains the track points, the target user's location, the drone's speed, acceleration, and the drone's flight energy consumption.

[0046] reward function : , This represents the reward after the action is performed in time slot t. and All are coefficients of constant terms. This indicates the first part of the reward. This represents the ergodic capacity of the system under the optimal ORIS phase shift in the nth time slot. This indicates the second part of the reward, which is a preset fixed value used to penalize drones for exceeding preset limits in speed or position. The final reward at the end of a round. This indicates the number of time slots after the drone's energy is depleted;

[0047] Strategy: Employ an ε-greedy strategy, which uses an exploratory strategy with a probability of ε, i.e., randomly selecting actions, and an exploitation strategy with a probability of 1-ε, i.e., selecting the action with the highest value through an online Q-network.

[0048] Experience Pool: It is stored as a set of experiences in the experience pool. Indicates the selection in time slot t The state of the drone at time t+1 under the action. This marks the end of an agent's round.

[0049] The DDQN algorithm, following the above settings, selects a fixed amount of data from the experience pool to train the online Q-network, with the target network Q... ’ The parameters of the online Q-network are copied according to the set period. After training, the UAV trajectory under the optimal ORIS phase shift scheme obtained by the first sub-model is obtained.

[0050] Furthermore, the online Q-network described in step (6) is updated after each iteration as follows;

[0051] DDQN minimizes the Q-value obtained by the online Q-network and the Q-value of the target network. ’ The loss between the obtained target values ​​is used to update the online Q network, and the target value is... The calculation is as follows:

[0052]

[0053] In the formula, Indicates the discount factor. and These are the online Q network and the target network Q. ’ Network parameters, Represents the action corresponding to the maximum Q-value in an online Q-network; loss function The calculation is as follows:

[0054]

[0055] In the formula, This indicates the number of data points selected from the experience pool.

[0056] Beneficial effects: Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention introduces ORIS-assisted UAV relay and makes full use of the maneuverability of UAVs and the adjustability of ORIS, which can provide high-quality communication for mobile users while maximizing traversal capacity; the phase shift of ORIS is optimized by simulated annealing algorithm to reduce the impact of pointing error, and on this basis, the UAV trajectory is optimized by DDQN algorithm to obtain the optimal UAV trajectory. Attached Figure Description

[0057] Figure 1 This is a flowchart of the present invention;

[0058] Figure 2 This is a schematic diagram of the system model used in this invention;

[0059] Figure 3 The flowchart shows the ORIS phase shift optimization based on the simulated annealing algorithm.

[0060] Figure 4 A 3D UAV flight path map;

[0061] Figure 5 This is a comparison chart showing the system traversal capacity with and without trajectory planning in this invention;

[0062] Figure 6 This is a comparison chart of the traversal capacity of the present invention and traditional UAV relay systems;

[0063] Figure 7 A comparison chart of system ergodic capacity under different receiver radii and beam divergence angles. Detailed Implementation

[0064] The invention will now be further described with reference to the accompanying drawings.

[0065] like Figure 1 As shown, this invention provides an ORIS-assisted UAV relay FSO communication trajectory planning method, comprising the following steps:

[0066] Step 1: Establish a drone trajectory optimization model.

[0067] The target-moving ORIS-assisted UAV relay FSO communication system model used in this invention is as follows: Figure 2 As shown, the relay drone is assisted by ORIS. The moving target is a dynamic drone that provides services to ground users. The moving target flies at a fixed altitude and is located directly above the distribution center of the ground users. The ground users move according to a random movement model.

[0068] The UAV trajectory optimization model takes the optimal ORIS phase shift and optimal trajectory as the objective function to maximize the traversal capacity of the communication system, and uses ORIS phase shift constraints, UAV flight trajectory constraints, and UAV energy consumption constraints as constraints. Specifically:

[0069]

[0070] In the formula, P1 represents the UAV trajectory optimization model. This represents the traversal capacity of the time slot t system; This represents the optimal ORIS phase shift scheme, specifically the optimal reflection angle of ORIS in each time slot. The optimal flight path is specifically the set of optimal flight path points for each time slot of the UAV. This represents the rotation angle of ORIS along the x-axis in time slot t. This represents the rotation angle of ORIS along the y-axis in time slot t; Indicates the initial value of the waypoint. The three-dimensional coordinates of the initial flight path of the UAV; This represents the speed of the UAV in time slot t. and These are the minimum and maximum speeds of the drone; the drone's flight altitude. Limited to minimum height and maximum height between; The drone has already consumed energy. This indicates the maximum energy consumption of the drone; This represents the acceleration of the UAV in time slot t. and These represent the minimum and maximum accelerations of the drone, respectively.

[0071] The system traversal capacity is calculated as follows:

[0072]

[0073] In the formula, for The probability density function, It can be represented as:

[0074]

[0075] in, The channel coefficient can be expressed as:

[0076]

[0077] In the formula , and These are atmospheric attenuation, atmospheric turbulence, and directionality error, respectively.

[0078] The atmospheric attenuation coefficient can be expressed as:

[0079]

[0080] in, and They represent visibility and... The atmospheric attenuation coefficients for sunny and foggy days, and These represent the lengths the beam travels through in clear sky and fog, respectively.

[0081] Atmospheric turbulence, using the Gamma-Gamma distribution model, can be represented as:

[0082]

[0083] In the formula, It is a modified Bessel function of the second kind. It is the gamma equation. The numbers representing the number of small-turbulence and large-turbulence elements can be expressed as follows:

[0084]

[0085]

[0086] The Rytov variance can be calculated. , It is a refractive index structural parameter. For the total propagation length, λ is the wavelength of the optical signal.

[0087] There is a misalignment vector between the center point of the photodetector at the receiving end and the beam footprint. The pointing error is caused by the laser power distribution on the photodetector. The center point of the beam footprint can be calculated using a rotation matrix combined with specular reflection. The normalized optical power spatial distribution at the communication distance D is approximately:

[0088]

[0089] In the formula, , They are in The light intensity at point = 0, and the beam divergence angle are... The beam waist at communication distance D. If the communication link is long enough, the receiving aperture... The radius is much smaller than This allows for the reasonable allocation of the receiving aperture area. Assuming the light intensity is constant, the geometric loss factor caused by the pointing error can be expressed as:

[0090]

[0091] Step 2: Decompose the UAV trajectory optimization model into a first sub-model and a second sub-model.

[0092] The first sub-model takes obtaining the optimal phase shift scheme for a fixed UAV position as its objective function, specifically:

[0093]

[0094] The second sub-model takes obtaining the optimal trajectory under ORIS fixed phase shift as its objective function, specifically:

[0095]

[0096] Step 3: Using the track points in the second sub-model as known track points, solve the first sub-model using the simulated annealing algorithm to obtain the optimal ORIS phase shift scheme.

[0097] First, the waypoints obtained from the second sub-model (initial value is a preset starting point) are used as known waypoints. Simulated annealing is then used to optimize the phase shift of ORIS, yielding the maximum traversal capacity and optimal phase shift for the current waypoint. The flowchart is as follows: Figure 3 As shown. The simulated annealing algorithm avoids getting trapped in local optima and can effectively solve irregular optimization problems that are difficult to solve using conventional methods. To solve the first sub-model problem, firstly, set the initial annealing temperature, the minimum annealing temperature, and initialize the current phase shift. In each iteration, a new phase shift is generated by perturbing the current phase shift. Then, the traversal capacity under the current phase shift and the new phase shift are calculated respectively. A receiving probability is used to determine whether to accept the new solution. The receiving probability can be expressed as:

[0098]

[0099] in, , The current temperature is used. As the iteration progresses, the temperature gradually decreases through the cooling coefficient. When the temperature falls below the minimum temperature, the iteration terminates, and the maximum traversal capacity at the current track point is output.

[0100] Step 4: Using the optimal ORIS phase shift scheme obtained from the first sub-model as the known scheme, solve the second sub-model using the DDQN algorithm to obtain the UAV's waypoints for the next time slot. The DDQN algorithm uses... The probability of randomly selecting the drone's actions or using The probability is used to select the most valuable action through an online Q-network, thus obtaining the waypoint for the next time slot. The online Q-network is used by the UAV to select the most valuable action in each time slot. In addition, the target network Q... ’ The online Q-network is used to evaluate the value of the drone's state in the next time slot. It is updated after each iteration. The target network Q... ’ The parameters of the online Q network are copied according to the set period.

[0101] The DDQN algorithm learns using the following settings to obtain the UAV trajectory under the optimal ORIS phase-shift scheme obtained from the first sub-model:

[0102] Objective: To maximize the traversal capacity of the system under the optimal ORIS phase shift scheme obtained from the first sub-model.

[0103] action The action selected by the time slot agent is specifically a combination of the component accelerations of the UAV along the three-dimensional coordinate axes, that is, a combination of the x-axis component acceleration, y-axis component acceleration, and z-axis component acceleration taking values ​​from the set {-2, 0, 2} respectively.

[0104] state The time slot includes the track point, the target user's location, the drone's speed, acceleration, and the drone's flight energy consumption.

[0105] reward function : , This represents the reward after the action is performed in time slot t. and All are coefficients of constant terms. This indicates the first part of the reward. This represents the ergodic capacity of the system under the optimal ORIS phase shift in the nth time slot. This indicates the second part of the reward, which is a preset fixed value used to penalize drones for exceeding preset limits in speed or position. The final reward at the end of a round. This indicates the number of time slots after the drone's energy is depleted.

[0106] Strategy: Employ an ε-greedy strategy, which uses an exploratory strategy with a probability of ε, i.e., randomly selecting actions, and an exploitation strategy with a probability of 1-ε, i.e., selecting the action with the highest value through an online Q-network.

[0107] Experience Pool: It is stored as a set of experiences in the experience pool. Indicates the selection in time slot t The state of the drone at time t+1 under the action. This is a marker indicating the end of an agent's round.

[0108] The DDQN algorithm, following the above settings, selects a fixed amount of data from the experience pool to train the online Q-network, with the target network Q... ’ The parameters of the online Q-network are copied according to the set period. After training, the UAV trajectory under the optimal ORIS phase shift scheme obtained by the first sub-model can be obtained.

[0109] The network in the DDQN algorithm is updated in the following way:

[0110] The Q-value obtained by minimizing the online Q-network is compared with the target network Q. ’ The loss between the obtained target values ​​is used to update the online Q network, and the target value is... The calculation is as follows:

[0111]

[0112] In the formula, Indicates the discount factor. and These are the online Q network and the target network Q. ’ Network parameters, This represents the action corresponding to the maximum Q-value in an online Q-network. Loss function. The calculation is as follows:

[0113]

[0114] In the formula, This indicates the number of data points selected from the experience pool.

[0115] Step 5: Determine if the drone's energy is depleted.

[0116] If not, return to step 3; if yes, proceed to step 6.

[0117] Step 6: Repeat steps 4 to 5 iteratively until the number of iterations reaches the preset value. Based on the Q-network trained after the iteration, select the action with the highest value in each time slot to form the UAV track.

[0118] The invention was verified by simulation. Figure 4 This is a 3D drone flight path map, where the drone's starting point is located within a dense fog area. From Figure 4 The drone's flight path shows that it quickly flew out of the dense fog area that severely affected communication quality, and then adaptively adjusted its position to ensure a high-quality communication link with the moving target. Subsequently, the drone hovered in areas with strong signal quality to maintain optimal communication.

[0119] Figure 5 A comparison chart showing the system traversal capacity for each time slot after UAV trajectory planning and when the UAV is hovering. Figure 5 The traversal capacity after UAV trajectory planning increases rapidly from a low level and then remains at a high level, which confirms that... Figure 4 The UAV quickly flew out of the fog area with poor signal quality and then maintained a better communication status. In addition, when the ORIS-assisted UAV relay hovered in dense fog and clear sky areas respectively, located at [400,400,450] and [400,400,350], the system's traversal capacity was low, which also verifies that UAV trajectory planning improves the system's communication quality.

[0120] Figure 6 A comparison chart of traversal capacity for UAV relay under ORIS-assisted UAV relay and decoding / forwarding modes is presented. Figure 6 As can be seen, the traversal capacity of the system increases with the increase of the transmitting power. However, under the same conditions, the traversal capacity of ORIS-assisted UAV relay is greater than that of traditional UAV relay. This is because ORIS's phase shift optimization compensates for the pointing error at the receiving end, increases the optical signal energy on the photodetector, and thus increases the traversal capacity.

[0121] To further verify the impact of beam width and photodetector area on the ORIS-assisted UAV relay system Figure 7 This diagram shows a comparison of ergodic capacity under different detector radii and beam divergence angles. Figure 7 As can be seen, a smaller beam divergence angle concentrates the optical signal energy, resulting in a more concentrated energy reception on the photodetector and an increased system traversal capacity. Furthermore, a larger detector area can capture more optical signals, thereby enhancing the system's traversal capacity.

[0122] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An ORIS-assisted method for planning the FSO communication trajectory of an unmanned aerial vehicle (UAV) relay, characterized in that, Includes the following steps: (1) Establish a UAV trajectory optimization model, wherein the objective function is to obtain the optimal ORIS phase shift and the optimal trajectory when the communication system traversal capacity is maximized, and the constraints are ORIS phase shift constraint, UAV flight trajectory constraint and UAV energy consumption constraint. (2) The UAV trajectory optimization model is decomposed into a first sub-model and a second sub-model. The first sub-model takes the optimal phase shift scheme under the fixed position of the UAV as the objective function, and the second sub-model takes the optimal trajectory under the fixed phase shift of ORIS as the objective function. (3) Using the track points in the second sub-model as known track points, solve the first sub-model using the simulated annealing algorithm to obtain the optimal ORIS phase shift scheme; (4) Using the optimal ORIS phase shift scheme obtained from the first sub-model as the known scheme, solve the second sub-model according to the DDQN algorithm to obtain the track points of the UAV in the next time slot; (5) Determine if the drone's energy is depleted. If not, return to step (3); if yes, proceed to step (6). (6) Repeat steps (4) to (5) iteratively, and the online Q network is updated after each iteration; until the number of iterations reaches the preset value, the action with the highest value in each time slot is selected based on the Q network trained after the iteration to form the UAV track; The specific UAV trajectory optimization model described in step (1) is as follows: ; In the formula, P1 represents the UAV trajectory optimization model. This represents the traversal capacity of the time slot t system; This represents the optimal ORIS phase shift scheme, specifically the optimal reflection angle of ORIS in each time slot. The optimal flight path is specifically the set of optimal flight path points for each time slot of the UAV. This represents the rotation angle of ORIS along the x-axis in time slot t. This represents the rotation angle of ORIS along the y-axis in time slot t; This represents the initial value of the waypoint. The three-dimensional coordinates of the initial flight path of the UAV; This represents the speed of the UAV in time slot t. and These are the minimum and maximum speeds of the drone; the drone's flight altitude. Limited to minimum height and maximum height between; The drone has already consumed energy. This indicates the maximum energy consumption of the drone; This represents the acceleration of the UAV in time slot t. and These are the minimum and maximum accelerations of the drone, respectively. The system traversal capacity in step (1) is: ; In the formula, for The probability density function, for: ; in, The channel coefficient is: ; In the formula, , and These are atmospheric attenuation coefficient, atmospheric turbulence, and directionality error, respectively. The atmospheric attenuation coefficient is: ; in, and They represent visibility and... The atmospheric attenuation coefficients for sunny and foggy days, and These represent the lengths the light beam travels through in clear sky and fog, respectively. The atmospheric turbulence uses a Gamma-Gamma distribution model, which is: ; In the formula, It is a modified Bessel function of the second kind. It is the gamma equation. Let represent the number of small-turbulence and large-turbulence units, respectively, as follows: ; ; Among them, the Rytov variance was calculated. , It is a refractive index structural parameter. For the total propagation length, λ is the wavelength of the optical signal.

2. The ORIS-assisted unmanned aerial vehicle relay FSO communication trajectory planning method according to claim 1, characterized in that, The pointing error is the misalignment vector between the center point of the photodetector at the receiving end and the beam footprint. The error is caused by the laser power distribution on the photodetector. The center point of the beam footprint is calculated using a rotation matrix combined with specular reflection. The normalized optical power spatial distribution at the communication distance D is approximately: ; In the formula, Is Light intensity at point =0 The beam divergence angle is The beam waist at communication distance D; If the communication link is long enough, the receiving aperture... The radius is much smaller than The receiving aperture area Assuming the light intensity is constant, the geometric loss factor caused by the directivity error is: 。 3. The ORIS-assisted unmanned aerial vehicle relay FSO communication trajectory planning method according to claim 1, characterized in that, Step (2) The first sub-model is specifically as follows: 。 4. The ORIS-assisted unmanned aerial vehicle relay FSO communication trajectory planning method according to claim 1, characterized in that, The implementation process of step (3) is as follows: Set the initial annealing temperature, the minimum annealing temperature, and initialize the current phase shift. ; In each iteration, a new phase shift is generated by perturbing the current phase shift. Calculate the traversal capacity under the current phase shift and the new phase shift respectively, and determine whether to accept the new solution using an acceptance probability, which is: ; in, , The current temperature is used; as the iteration progresses, the temperature gradually decreases through the cooling coefficient. When the temperature falls below the minimum temperature, the iteration terminates, and the maximum traversal capacity at the current track point is output.

5. The ORIS-assisted unmanned aerial vehicle relay FSO communication trajectory planning method according to claim 1, characterized in that, The implementation process of step (4) is as follows: Objective: To maximize the traversal capacity of the system under the optimal ORIS phase shift scheme obtained from the first sub-model; action The action selected by the time slot agent is specifically a combination of the component accelerations of the UAV along the three-dimensional coordinate axes, that is, a combination of the x-axis component acceleration, y-axis component acceleration, and z-axis component acceleration taking values ​​from the set {-2, 0, 2} respectively. state The time slot t contains the track points, the target user's location, the drone's speed, acceleration, and the drone's flight energy consumption. reward function : , This represents the reward after the action is performed in time slot t. and All are coefficients of constant terms. This indicates the first part of the reward. This represents the ergodic capacity of the system under the optimal ORIS phase shift in time slot t. This indicates the second part of the reward, which is a preset fixed value used to penalize drones for exceeding preset limits in speed or position. The final reward at the end of a round. This indicates the number of time slots after the drone's energy is depleted; Strategy: Employ an ε-greedy strategy, which uses an exploratory strategy with a probability of ε, i.e., randomly selecting actions, and an exploitation strategy with a probability of 1-ε, i.e., selecting the action with the highest value through an online Q-network. Experience Pool: It is stored as a set of experiences in the experience pool. Indicates the selection in time slot t The state of the drone at time t+1 under the action. This marks the end of an agent's round. The DDQN algorithm, following the above settings, selects a fixed amount of data from the experience pool to train the online Q-network, with the target network Q... ’ The parameters of the online Q-network are copied according to the set period. After training, the UAV trajectory under the optimal ORIS phase shift scheme obtained by the first sub-model is obtained.

6. The ORIS-assisted unmanned aerial vehicle relay FSO communication trajectory planning method according to claim 1, characterized in that, The online Q-network described in step (6) is updated after each iteration, and the implementation process is as follows: DDQN minimizes the Q-value obtained by the online Q-network and the Q-value of the target network. ’ The loss between the obtained target values ​​is used to update the online Q network, and the target value is... The calculation is as follows: ; In the formula, For the reward function, Indicates the discount factor. and These are the online Q network and the target network Q. ’ Network parameters, Represents the action corresponding to the maximum Q-value in an online Q-network; loss function The calculation is as follows: ; In the formula, This indicates the number of data points selected from the experience pool.

Images