Path planning method and system based on wireless sensor network

Abstract

The invention discloses a path planning method and system based on a wireless sensor network, relates to the technical field of wireless communication, and is used for solving the technical problem that an optimal planning path is difficult to obtain based on the wireless sensor network due to the fact that the power consumption of a sensor node cannot be reduced in the prior art. According to the technical key points, the method comprises the following steps: collecting communication data between the flight equipment and a ground sensor, wherein the communication data comprises motion data, state data and alarm data of the flight equipment; the motion data comprises position and speed; the state data comprises electric quantity and working duration; the alarm data comprises low electric quantity alarm, obstacle alarm and fault alarm; constructing a channel optimization model based on the optimized air-ground channel throughput; and constructing a path planning model, and solving the path planning model by using a deep neural network algorithm to obtain a flight equipment full-coverage path. According to the invention, full-coverage optimal path planning based on combination of channel optimization of the wireless sensor network and an improved neural network path planning algorithm is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of wireless sensor control, and particularly to a path planning method and system based on a wireless sensor network. Background Art

[0002] In the path planning and communication system of a flying device, a wireless sensor network (WSN) is widely used to support data communication between the flying device and ground sensor nodes. The wireless sensor network has become a key means to obtain the flight data and environmental information of the flying device due to its advantages such as self-organization, low cost, and convenience. However, the wireless sensor network usually faces problems such as energy consumption, insufficient channel throughput, and unstable transmission. In particular, the high power consumption problem of sensor nodes will significantly affect the continuity of data transmission and the overall performance of the system, making it difficult to achieve the optimal path planning in a complex environment.

[0003] The existing technologies have obvious deficiencies in aspects such as sensor power consumption, channel technology, path planning, and communication-path coordination. In terms of sensor power consumption, ground sensor nodes are usually powered by batteries and are energy-limited, resulting in frequent high-power consumption phenomena of nodes and shortening the life cycle of the wireless sensor network. Due to the high power consumption problem, nodes may fail frequently, resulting in data transmission interruption or loss, making it difficult for the flying device to obtain continuous and stable communication data. Interference of complex environments on communication: In complex environments such as the presence of obstacles and occlusions, signals are easily interfered, resulting in a decrease in the quality of obtained communication data and affecting the accuracy of subsequent channel optimization and path planning. These problems lead to discontinuous data acquisition and poor communication stability, resulting in insufficient channel throughput. In terms of channel technology, the existing channel optimization technologies are difficult to adapt to complex and changeable channel environments and cannot adjust the air-ground communication link in real time. When the flying device is in a high-speed flight or long-distance mission scenario, the channel attenuation problem is aggravated, making it difficult for the channel throughput to remain at an ideal level. In terms of path planning and communication-path coordination, path planning is limited by communication interruption or data loss caused by high power consumption of sensor nodes. The path planning model is difficult to obtain complete environmental data and flight data, restricting the accuracy of the planned path. The coordination between communication and path planning is insufficient, resulting in weak dynamic adjustment ability. The dynamic adjustment ability is insufficient, making it difficult to generate high-quality planned paths in complex environments. When the environment of the flying device or ground sensor changes, such as node failure and channel attenuation, the system lacks dynamic adjustment ability, affecting the reliability of the setting of key positions and the drawing of the optimal path. Summary of the Invention

[0004] In view of this, the present invention proposes a path planning method and system based on a wireless sensor network to solve the technical problem that it is difficult to obtain the optimal planned path based on the wireless sensor network due to the inability to reduce the power consumption of sensor nodes in the prior art.

[0005] According to one aspect of the present invention, a path planning method based on a wireless sensor network is proposed, including the following steps:

[0006] Step 1: Collect communication data between the flying device and the ground sensor. The communication data includes motion data, status data, and alarm data of the flying device. Among them, the motion data includes position and speed; the status data includes battery power and working duration; the alarm data includes low battery alarm, obstacle alarm, and fault alarm;

[0007] Step 2: Build a channel optimization model based on optimizing the air-ground channel throughput;

[0008] Step 3: Build a path planning model based on the communication data and the channel optimization model, and use the deep neural network algorithm to solve the path planning model to obtain the full coverage path of the flying device.

[0009] Further, after obtaining the full coverage path of the flying device, it further includes:

[0010] Step 4: Set waypoints based on the full coverage path, dynamically adjust the distribution of waypoints using the motion data of the flying device, and output an optimized set of waypoints;

[0011] Step 5: Integrate the optimized set of waypoints and the full coverage path set into an optimal path, draw the optimal path through a Web interface, and display it on an integrated map.

[0012] Further, the channel optimization model in Step 2 is built as follows:

[0013]

[0014] In the formula, M represents the number of reflection units of the intelligent reflecting surface; m represents that the flying device is in a flying state, represents the signal-to-noise ratio of the air-ground communication between the intelligent reflecting surface and the flying device, and σ 2 is the power of additive white Gaussian noise, and P s represents the power of the source sensor node, and h SD represents the channel gain between the source sensor node and the intelligent reflecting surface, represents the channel gain from the intelligent reflecting surface at the height of point H to the sensor node; h SR represents the channel gain between the source sensor node and the intelligent reflecting surface, represents the signal-to-noise ratio in the flying state of the flying device; and there is:

[0015]

[0016] Among them, ρ represents the path loss, and dSD denotes the distance from the source sensor node to the target sensor node, h denotes the channel gain; d(φ SR - φ RD ) represents the distance between φ SR and φ RD ; φ SR represents the cosine of the angle of arrival of the signal from the source sensor node to the intelligent reflecting surface, φ RD represents the cosine of the angle of arrival of the signal from the intelligent reflecting surface to the target sensor node, R represents the intelligent reflecting surface, D represents the target sensor node, and S represents the source sensor node; represents the path loss of the distance between the source sensor node and the intelligent reflecting surface, and τ is the path loss exponent between the source sensor node and the intelligent reflecting surface; represents the path loss of the distance between the intelligent reflecting surface and the target sensor node, and k is the loss exponent between the intelligent reflecting surface and the target sensor node; θ i represents the optimal reflection phase of each element of the intelligent reflecting surface, λ represents the carrier wavelength, and arg(h) represents the phase angle of the channel gain h.

[0017] Furthermore, constructing a path planning model based on the communication data and the channel optimization model in step three includes:

[0018] Step 3-1. Flight area modeling; use a two-dimensional grid map to model the flight area, and divide the flight area into a dark area and a bright area. Among them, the dark area indicates that there are obstacles in the area and flight equipment is prohibited from entering, and the bright area indicates that there are no obstacles in the area and flight equipment is allowed to enter; set the length of the flight area as L, the width as W, and the side length of the grid as D, then the total number of grids is:

[0019]

[0020] Step 3-2. Establish an evaluation function and a reward function;

[0021] The evaluation function is:

[0022]

[0023] In the formula, E c represents the flight path coverage rate; E r represents the flight path repetition rate; E h represents the high-frequency repetition rate of the flight path; A r represents the accident penalty; b, c, d, and e are set proportionality coefficients, indicating the adjustment of the task objectives; among them, the flight path coverage rate is: S represents the total area of the bright area, S 1 represents the area of the covered area, S 2is the area of the dark area; the flight path repetition rate is: D 1 is the total path length, representing the total number of flight device step lengths, D 2 is the standard path length, representing the number of blank grids; the high-frequency repetition rate of the flight path is: S h represents the area of the region where the coverage times exceed a certain threshold, S r represents the total area of the covered area; the accident penalty A r is to constrain the dark area of the flight device and the behavior of the flight device;

[0024] The reward function is used to maximize the throughput, and the reward function is:

[0025] R(T) = λ 1 ·θ i -λ 2 ·Penalty(t)

[0026] In the formula, λ 1 and λ 2 are weight coefficients, and Penalty(t) is a penalty function for violating constraints, indicating that if the throughput at any moment is lower than the minimum requirement, a penalty is imposed;

[0027] Step 3: Taking the maximized channel optimization model as the objective function, using the evaluation function and the reward function as constraints, and using the communication data as input, construct a path planning model.

[0028] Furthermore, the use of the deep neural network algorithm to solve the path planning model in Step 3 to obtain the full-coverage path of the flight device includes: The deep neural network algorithm includes an input layer, a hidden layer, and an output layer; among them,

[0029] The input of the input layer is: {S 1 , S 2 , S 3 , S 4 , S 5 , Z 1 , Z 2 , Z 3 , Z 4 , Z 5 , θ i , H, V, E, W, F, A l , R(T), X, Y}; represents the obstacle information in 5 directions, namely directly to the right, right front, directly in front, left front, and directly to the left of the flight device output by identifying the intelligent reflecting surface through the two-dimensional grid map, d represents the range of the air-ground channel of the wireless sensor network, d iIndicates the distance to the obstacle in the i-th direction. When there is no obstacle within the range of the wireless sensor air-ground channel between the flying device and the ground receiver, then d i = d; Indicates the coverage information in 5 directions obtained from the two-dimensional grid map, t i Is the number of times the flying device stays within the grid where the i-th wireless sensor end is located, t max Is the maximum number of coverage times; H is the height, V is the speed, E is the power, W is the working duration, A l Is the alarm data; X, Y are the positions of the flying device;

[0030] The hidden layer includes 6 neurons, and the ReLU activation function is used: f(x) = max(0, x). When the input value is greater than 0, the value is output, otherwise 0 is output;

[0031] The output layer uses the logistic function as the activation function: Where e is the natural constant and p is the rate of change of the S-shaped function; the output quantity is expressed as: {y 1 , y 2 , y 3 , y 4}, y i Represents neurons of 4 consecutive values, and each neuron corresponds to the lift value of a motor, respectively controlling the movement of the flying device in the four directions of up, down, left, and right, y 1 Controls the upward lift, y 2 Controls the downward lift, y 3 Controls the leftward lift, y 4 Controls the rightward lift;

[0032] By the lift of the four motors output, the movement flight path of the flying device is adjusted in real time, and then the full-coverage path of the entire flight area is obtained.

[0033] According to another aspect of the present invention, a path planning system based on a wireless sensor network is proposed, including:

[0034] Data acquisition module: It is configured to collect communication data between the flying device and the ground sensor, and the communication data includes the movement data, status data, and alarm data of the flying device; wherein, the movement data includes the position and speed; the status data includes the power and working duration; the alarm data includes low power alarm, obstacle alarm, and fault alarm;

[0035] Channel optimization module: It is configured to build a channel optimization model based on optimizing the air-ground channel throughput;

[0036] Path planning module: It is configured to construct a path planning model based on the communication data and the channel optimization model, and use a deep neural network algorithm to solve the path planning model to obtain a full-coverage path for the flying device.

[0037] Furthermore, the system further includes:

[0038] Waypoint setting module: It is configured to set waypoints based on the full-coverage path, dynamically adjust the distribution of waypoints using the motion data of the flying device, and output an optimized set of waypoints;

[0039] Path drawing module: It is configured to integrate the optimized set of waypoints and the full-coverage path into an optimal path, draw the optimal path through a Web interface, and display it on an integrated map.

[0040] Furthermore, the channel optimization model in the channel optimization module is constructed as follows:

[0041]

[0042] In the formula, M represents the number of reflection units of the intelligent reflecting surface; m represents that the flying device is in a flying state, represents the air-ground communication signal-to-noise ratio between the intelligent reflecting surface and the flying device, and σ 2 is the additive white Gaussian noise power, and P s represents the power of the source sensor node, and h SD represents the channel gain between the source sensor node and the intelligent reflecting surface, represents the channel gain from the intelligent reflecting surface at the H point height to the sensor node; h SR represents the channel gain between the source sensor node and the intelligent reflecting surface, represents the signal-to-noise ratio in the flying state of the flying device; and there is:

[0043]

[0044] Among them, ρ represents the path loss, d SD represents the distance from the source sensor node to the target sensor node, and h represents the channel gain; d(φ SR -φ RD ) represents the distance between φ SR and φ RD , φ SR represents the cosine of the signal arrival angle from the source sensor node to the intelligent reflecting surface, and φ RD represents the cosine of the signal arrival angle from the intelligent reflecting surface to the target sensor node, R represents the intelligent reflecting surface, D represents the target sensor node, and S represents the source sensor node; The path loss representing the distance between the source sensor node and the intelligent reflecting surface, where τ is the path loss exponent between the source sensor node and the intelligent reflecting surface; The path loss representing the distance between the intelligent reflecting surface and the target sensor node, where k is the loss exponent between the intelligent reflecting surface and the target sensor node; θ i Represents the optimal reflection phase of each element of the intelligent reflecting surface, λ represents the carrier wavelength, and arg(h) represents the phase angle of the channel gain h.

[0045] Furthermore, the construction of the path planning model based on the communication data and the channel optimization model in the path planning module includes:

[0046] Step 3-1: Modeling the flight area; The flight area is modeled using a two-dimensional grid map, and the flight area is divided into a dark area and a bright area. Among them, the dark area indicates that there are obstacles in the area and flight equipment is prohibited from entering, and the bright area indicates that there are no obstacles in the area and flight equipment is allowed to enter; Set the length of the flight area as L, the width as W, and the side length of the grid as D, then the total number of grids is:

[0047]

[0048] Step 3-2: Establishing the evaluation function and the reward function;

[0049] The evaluation function is:

[0050]

[0051] In the formula, E c Represents the flight path coverage rate; E r Represents the flight path repetition rate; E h Represents the high-frequency repetition rate of the flight path; A r Represents the accident penalty; b, c, d, e are set proportionality coefficients, indicating the adjustment of the task objectives; Among them, the flight path coverage rate is: S represents the total area of the bright area, S 1 Represents the area of the covered area, S 2 Is the area of the dark area; The flight path repetition rate is: D 1 Is the total path length, representing the total number of steps of the flight equipment, D 2 Is the standard path length, representing the number of blank grids; The high-frequency repetition rate of the flight path is: S h Represents the area of the region where the coverage times exceed a certain threshold, S r Represents the total area of the covered area; The accident penalty A r Is to restrict the dark area of the flight equipment and the behavior of the flight equipment;

[0052] The reward function is used to maximize the throughput, and the reward function is:

[0053] R(T) = λ 1 ·θ i -λ 2 ·Penalty(t)

[0054] In the formula, λ 1 and λ 2 are weight coefficients, and Penalty(t) is a penalty function for violating the constraints, indicating that a penalty is imposed if the throughput at any time is lower than the minimum requirement;

[0055] Step 3: Taking the maximized channel optimization model as the objective function, using the evaluation function and the reward function as the constraint conditions, and using the communication data as the input, a path planning model is constructed.

[0056] Furthermore, the use of the deep neural network algorithm in the path planning module to solve the path planning model to obtain the full-coverage path of the flying device includes: The deep neural network algorithm includes an input layer, a hidden layer, and an output layer; among them,

[0057] The input of the input layer is: {S 1 , S 2 , S 3 , S 4 , S 5 , Z 1 , Z 2 , Z 3 , Z 4 , Z 5 , θ i , H, V, E, W, F, A l , R(T), X, Y}; represents the obstacle information in 5 directions directly to the right, right front, straight ahead, left front, and straight left of the flying device output by identifying the intelligent reflecting surface through the two-dimensional grid map, d represents the range of the air-ground channel of the wireless sensor network, and d i represents the distance of the obstacle in the i-th direction. When there is no obstacle within the range of the air-ground channel of the wireless sensor between the flying device and the ground receiver, then d i = d; represents the coverage information in 5 directions obtained from the two-dimensional grid map, t i is the number of times the flying device stays in the grid at the end of the i-th wireless sensor, and t max is the maximum coverage number; H is the height, V is the speed, E is the power, W is the working duration, and A l is the warning data; X, Y are the positions of the flying device;

[0058] The hidden layer consists of 6 neurons, and the ReLU activation function is: f(x) = max(0, x). When the input value is greater than 0, the value is output; otherwise, 0 is output.

[0059] The output layer uses the logistic function as the activation function: where e is the natural constant and p is the rate of change of the sigmoid function; the output is expressed as: {y 1 , y 2 , y 3 , y 4}, where y i represents neurons with 4 consecutive values, and each neuron corresponds to the lift value of a motor, respectively controlling the movement of the flying device in the four directions of up, down, left, and right. y 1 controls the upward lift, y 2 controls the downward lift, y 3 controls the leftward lift, y 4 controls the rightward lift;

[0060] By the lift of the four motors output, the movement flight path of the flying device is adjusted in real time, and then the full-coverage path of the entire flight area is obtained.

[0061] The beneficial technical effects of the present invention are:

[0062] 1) The communication quality is improved through the channel optimization model;

[0063] In the present invention, by constructing a channel optimization model, based on the communication data of the flying device, the throughput of the air-ground channel is optimized, the channel throughput and the channel gain matrix are maximized, and the communication quality between the flying device and the ground sensor is effectively improved. During the communication process between the flying device and the ground sensor, the channel quality is often affected by environmental factors, the state of the flying device, obstacles, etc.; by optimizing the signal-to-noise ratio of the air-ground communication between the intelligent reflecting surface and the flying device, it is ensured that the flying device can maintain a high communication quality throughout the flight area; in order to ensure the smooth completion of the flight mission, the channel optimization model first collects the motion data, state data, alarm data, etc. of the flying device, and constructs the channel throughput and the channel gain matrix based on these data, thereby optimizing the communication path; by optimizing the throughput of the air-ground channel, not only can the communication efficiency between the flying device and the ground sensor be improved, but also stable communication in a complex flight environment can be ensured.

[0064] 2) Intelligent path planning is achieved through the deep neural network algorithm;

[0065] The present invention constructs a path planning model, which uses a deep neural network algorithm to input information such as communication data, motion data, and status data of a flying device to optimize the flight path. The deep neural network can process complex environmental information and dynamically adjust the path of the flying device, optimizing the communication quality and safety during the flight while ensuring the completion of the task; during the path planning process, the flying device not only needs to avoid obstacles, but also optimize the path according to the signal strength in the flight area and the power status of the flying device, etc., so as to ensure that the flying device can adapt to different challenges in a dynamic environment. Through the deep neural network, the flying device can adjust the flight path according to real-time environmental data; information such as obstacles in the flight area, the battery status of the flying device, and channel throughput are used as input data for the neural network, and the network learns the rules of path planning through multiple layers of neurons; whenever the flying device receives new information, the deep neural network will immediately adjust the path to prevent the flying device from entering an obstacle area or a place with poor signal. The path planning method of the present invention not only improves the flight efficiency, but also effectively reduces the risks during the flight. Especially in a complex environment, the flying device can autonomously identify and avoid potential dangers.

[0066] 3) The full coverage path realizes the full coverage of the flight area;

[0067] The planning of the full coverage path of the present invention enables the flying device to efficiently and comprehensively complete the task, and ensures the communication quality and flight safety. By combining channel optimization and a deep neural network algorithm, the obtained full coverage path of the flying device enables the flying device to cover each effective area within the flight area, avoiding blind spots and ensuring that no area that needs to be covered is missed in the flight task; in the flight task, it is crucial to ensure the coverage of each part of the flight area. Especially in a complex environment, the flying device often needs to avoid obstacles, ensure communication quality, and comply with task restrictions; by optimizing the channel and path, the flying device can efficiently perform the full coverage flight task.

[0068] The communication throughput and signal-to-noise ratio data provided by the channel optimization model provide effective guidance for path planning, helping the flying device to select the path with the strongest signal and avoid entering areas with weak or interrupted signals. Combining with the deep neural network algorithm, the flying device can adjust the flight path in real time to ensure that each effective area within the flight area can be covered. At the same time, the network will also continuously optimize the path through reinforcement learning methods, enabling the flying device to efficiently complete the task throughout the flight process. In the planning of the full coverage path, the status data of the flying device, such as power and alarm data, are also considered. According to the battery life and working duration of the flying device, the path is dynamically adjusted to avoid failure to complete the task due to insufficient power or equipment failure during the mission.

[0069] 4) The generation and display of the optimal path are realized through dynamic optimization of waypoints;

[0070] Based on the full-coverage path, the present invention can adjust the waypoints of the flying device in real time, generate the optimal path, and visualize it on an integrated map through a Web interface. This can effectively cope with environmental changes during the flight mission, such as the dynamic appearance of obstacles, the reduction of power, or the fluctuation of communication quality, ensuring efficient coverage of the target area and avoiding invalid or repetitive paths. Based on the optimized waypoints, the deep neural network dynamically adjusts the path according to the real-time position, status data of the flying device, and environmental changes; when the flying device approaches an obstacle or a poor-signal area, the waypoints will be automatically adjusted to ensure bypassing these areas and maintaining the best communication quality; the optimized waypoint set is displayed in real time through the Web interface, facilitating users to monitor and control. Description of the Drawings

[0071] The present invention can be better understood by referring to the following description in conjunction with the accompanying drawings, which are included in this specification together with the following detailed description and form a part of this specification, and are used to further illustrate the preferred embodiments of the present invention and explain the principles and advantages of the present invention.

[0072] Figure 1 It is a flowchart of a path planning method based on a wireless sensor network according to an embodiment of the present invention;

[0073] Figure 2 It is a schematic diagram of the air-ground channel communication between a flying device and a ground sensor based on an intelligent reflecting surface in a wireless sensor network according to an embodiment of the present invention;

[0074] Figure 3 It is a schematic diagram of measuring the threat situation of obstacles directly to the right, right front, straight ahead, left front, and straight left of the flying device according to an embodiment of the present invention;

[0075] Figure 4 It is a schematic diagram of the neural network algorithm adopted by the path planning model according to an embodiment of the present invention;

[0076] Figure 5 It is a change curve graph of optimizing the evaluation function using the neural network planning algorithm according to an embodiment of the present invention; the vertical axis "Cost" is the reciprocal of the path score, and the horizontal axis "Steps" represents the number of iterations. Detailed Embodiments

[0077] To enable those skilled in the art to better understand the solution of the present invention, the exemplary embodiments or examples of the present invention will be described below in conjunction with the accompanying drawings. Obviously, the described embodiments or examples are only a part of the embodiments or examples of the present invention, rather than all of them. All other embodiments or examples obtained by those of ordinary skill in the art based on the embodiments or examples in the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0078] An embodiment of the present invention provides a path planning method based on a wireless sensor network. As Figure 1 shown, the method includes the following steps:

[0079] Step 1: Collect communication data between a flying device and a ground sensor. The communication data includes motion data, status data, and alarm data of the flying device. Among them, the motion data includes position and speed; the status data includes power and working duration; the alarm data includes low-power alarm, obstacle alarm, and fault alarm.

[0080] Step 2: Build a channel optimization model based on optimizing the air-ground channel throughput.

[0081] Step 3: Build a path planning model based on the communication data and the channel optimization model, and use a deep neural network algorithm to solve the path planning model to obtain a full-coverage path of the flying device.

[0082] Step 4: Set waypoints based on the full-coverage path, dynamically adjust the distribution of waypoints using the motion data of the flying device, and output an optimized set of waypoints.

[0083] Step 5: Integrate the optimized set of waypoints and the full-coverage path into an optimal path, draw the optimal path through a Web interface, and display it on an integrated map.

[0084] The method starts from Step 1. In Step 1, collect communication data between a flying device and a ground sensor. The communication data includes motion data, status data, and alarm data of the flying device. Among them, the motion data includes position and speed; the status data includes power and working duration; the alarm data includes low-power alarm, obstacle alarm, and fault alarm.

[0085] According to an embodiment of the present invention, real-time monitoring and data feedback can obtain motion data and status data to monitor the flight status and battery condition of a flying device in real time. This real-time feedback enables the monitoring system of the flying device to adjust the flight path in a timely manner, ensuring the smooth completion of the mission and reducing the risk of flight interruption or crash caused by low battery or abnormal flight status. Precise positioning and flight navigation can utilize the latitude and longitude information contained in the communication data of the flying device, which helps improve the positioning accuracy of the flying device, thereby enhancing the accuracy of navigation and path planning. This is particularly important for flying devices operating in complex environments such as densely populated urban areas with high-rise buildings or mountainous environments, enabling them to complete tasks more effectively and reducing the risk of collisions caused by navigation errors. The introduction of alarm data in the multi-level alarm mechanism, including low battery alarm, obstacle alarm, and fault alarm; helps the flying device have self-protection capabilities during flight. The low battery alarm can remind the flying device to return or land in a timely manner, the obstacle alarm enables the flying device to avoid flight obstacles, and the fault alarm helps take immediate measures when the device malfunctions, reducing the impact of the fault on the mission and extending the service life of the device. The intelligent reflecting surface can enhance communication stability, allowing the intelligent reflecting surface on the surface of the flying device to move with the wireless flying vehicle based on its mobility, optimizing the reflection and coverage of wireless signals. This technology enhances the communication stability and signal reception ability of the flying device in different flight postures, enabling the flying device to maintain a good connection with ground sensors even at long distances or in complex environments, thereby improving the mission completion rate and reliability. The low-power design improves the endurance ability by optimizing the data transmission efficiency of the sensor and communication system and reducing the possible multiple reflections and energy consumption losses during communication through the intelligent reflecting surface. While maintaining high communication stability, the flying device can significantly reduce energy consumption and increase the endurance time. This is particularly crucial for application scenarios such as long-duration missions or long-range cruises.

[0086] Then, step two is executed. In step two, a channel optimization model is constructed based on optimizing the air-ground channel throughput.

[0087] The channel optimization model described in step two includes a source sensor node, an intelligent reflecting surface, a communication link, a target sensor node, and a ground sensor node; where the communication link is: the line-of-sight link between the source sensor node and the intelligent reflecting surface, and the line-of-sight link between the intelligent reflecting surface and the target sensor node.

[0088] The construction of the channel optimization model described in step two is as follows:

[0089]

[0090] In the formula, M represents the number of reflection units of the intelligent reflecting surface; m represents the flying state of the flying device, Denote the signal-to-noise ratio of the intelligent reflecting surface and the air-ground communication of the flying device, σ 2 is the additive white Gaussian noise power, P s Denote the source sensor node power, h SD Denote the channel gain between the source sensor node and the intelligent reflecting surface, Denote the channel gain from the intelligent reflecting surface representing the H-point height to the sensor node; h SR Denote the channel gain between the source sensor node and the intelligent reflecting surface, Denote the signal-to-noise ratio in the flight state of the flying device; and there is:

[0091]

[0092] Among them, ρ denotes the path loss, d SD Denote the distance from the source sensor node to the target sensor node, h denotes the channel gain; d(φ SR -φ RD ) denotes φ SR and φ RD The distance between them, φ SR Denote the cosine of the angle of arrival of the signal from the source sensor node to the intelligent reflecting surface, φ RD Denote the cosine of the angle of arrival of the signal from the intelligent reflecting surface to the target sensor node, R denotes the intelligent reflecting surface, D denotes the target sensor node, S denotes the source sensor node; Denote the path loss of the distance between the source sensor node and the intelligent reflecting surface, τ is the path loss exponent between the source sensor node and the intelligent reflecting surface; Denote the path loss of the distance between the intelligent reflecting surface and the target sensor node, k is the loss exponent between the intelligent reflecting surface and the target sensor node; θ i Denote the optimal reflection phase of each element of the intelligent reflecting surface, λ denotes the carrier wavelength, arg(h) denotes the phase angle of the channel gain h.

[0093] According to the embodiments of the present invention, the throughput can measure the communication quality of the flying device and the impact on the communication data volume. Among them, when the flying device is far from the ground sensor, the throughput approaches the static mode; when the flying device shortens the distance from the ground sensor through mechanical flight, the throughput is in the dynamic mode, that is, the flying device is in the flight mode; as Figure 2 shown, the mobility of the flying device in the flight mode helps to achieve a better air-to-ground channel, thereby further improving the system throughput. The optimization goal of this embodiment is to maximize the throughput of the flying device by setting the optimal phase shift matrix, so as to obtain the maximum data volume of the communication data of the flying device.

[0094] Then, execute Step 3. In Step 3, based on the communication data and the channel optimization model, construct a path planning model, and use the deep neural network algorithm to solve the path planning model to obtain the full-coverage path of the flying device.

[0095] Traditional path planning methods usually adopt techniques such as the straight reciprocating method, the spiral coverage method, or the rectangular sliding method. However, these traditional solutions are often difficult to fully meet the requirements in actual operation, and there are limitations in terms of execution efficiency and area coverage, especially in scenarios with complex obstacle distributions. The path planning model in this embodiment combines wireless sensor network technology to enable the flying device to perform the full-coverage path planning task in an environment with dense obstacles.

[0096] The path planning model includes flight area construction, establishment of evaluation functions and reward functions, and algorithm design.

[0097] Step 3-1: Flight area modeling;

[0098] Flight area modeling; Use a two-dimensional grid map to model the flight area, and divide the flight area into dark areas and bright areas. Among them, the dark area indicates that there are obstacles in the area and the flying device is prohibited from entering, and the bright area indicates that there are no obstacles in the area and the flying device is allowed to enter; specifically: First, use a two-dimensional grid map to identify and describe the obstacle information in the terrain environment, such as buildings, trees, shrubs, etc., divide the obstacles according to the grid, and mark the grid where the obstacle is located as a prohibited area, which is displayed as a dark area, and the grid without obstacles is marked as an allowed area, which is displayed as a bright area; the bright area is the flight area of the flying device. Then, preset an extension value to perform edge extension processing on the boundaries of the grid and the boundaries of the obstacles to ensure the safe flight of the flying device. Specifically, when the dark area occupies more than 20% of the grid, the grid cell where the grid is located is marked as a prohibited state, otherwise it is marked as an allowed state. Figure 3 Shows the obstacle threat situations measured directly to the right, right front, directly in front, left front, and directly to the left of the flying device.

[0099] Then, set the theoretical minimum value of the grid as the spray width D of the flying device, so as to improve the accuracy of the full-coverage path planning. Use a two-dimensional grid map to model the flight area, and divide the flight area into dark areas and bright areas. Among them, the dark area indicates that there are obstacles in the area and the flying device is prohibited from entering, and the bright area indicates that there are no obstacles in the area and the flying device is allowed to enter; set the length of the flight area as L, the width as W, and the side length of the grid as D, then the total number of grids n in this environment is as follows:

[0100]

[0101] Step 3-2: Establish evaluation functions and reward functions;

[0102] As an optimization mechanism, the evaluation function introduces a path optimization mechanism. It comprehensively considers the coverage rate, path repetition rate, high-frequency repetition rate, and accident penalty of the flying device, and can obtain the evaluation function F of the flying device path, which is used to generate multiple sets of neural network parameters. By comparing the advantages and disadvantages of the generated paths through the evaluation function, the global optimal solution can be found. The evaluation function is as follows:

[0103]

[0104] In the formula, E c represents the flight path coverage rate; E r represents the flight path repetition rate; E h represents the flight path high-frequency repetition rate; A r represents the accident penalty; b, c, d, and e are set proportionality coefficients, indicating the adjustment of the task objectives.

[0105] The coverage rate, path repetition rate, high-frequency repetition rate, and accident penalty in the evaluation function are as follows:

[0106] Coverage rate: Taking the full coverage of the flight area as an indicator, the coverage rate evaluates the indicator as shown in the following formula:

[0107]

[0108] In the formula, S represents the total area of the entire operation area, S1 represents the covered area, and S2 represents the area covered by obstacles.

[0109] Path repetition rate: Multiple coverages of the same area by the flying device are regarded as over-coverage. The path repetition rate is set as shown in the following formula:

[0110]

[0111] In the formula, D 1 represents the total path length and the total number of steps of the flying device; D 2 is the standard path length, which is represented by the number of blank grids in the grid map.

[0112] High-frequency repetition rate: The high-frequency repetition rate is one of the important indicators to measure the performance of the path planning algorithm. By analyzing the high-frequency repetition rate, the ability of the algorithm to break through the local optimal solution can be effectively evaluated. Therefore, taking the high-frequency repetition rate as an indicator in the evaluation function helps the flying device to formulate a more superior path planning scheme, thereby improving the flight efficiency and reducing the risk. The high-frequency repetition rate is as follows:

[0113]

[0114] In the formula, S hRepresents the area of the region where the coverage times exceed a certain threshold, S r Represents the total covered area.

[0115] Accident penalty: Accident penalty A r To restrict the dark area of the flying device and restrict the behavior of the flying device. When the flying device enters the area where the obstacle is located or crosses the map boundary, it is regarded as a mission failure. Mission failure will have a series of negative impacts on the overall flight operation, such as the need to redesign the route, increase the time cost, reduce the operation efficiency, etc. To improve the execution efficiency and quality of the flight mission, a penalty mechanism is introduced into the evaluation function, so that the flight mission is more efficient and reliable. At the same time, for the flying device that successfully completes the mission, a certain reward is given to encourage it to continue to maintain a good execution state and efficiency.

[0116] The reward function is used to maximize the throughput, and the reward function is:

[0117] R(T) = λ 1 ·θ i ′ - λ 2 ·Penalty(t)

[0118] In the formula, λ 1 and λ 2 are weight coefficients, and Penalty(t) is the penalty function for violating the constraints, indicating that if the throughput at any moment is lower than the minimum requirement, a penalty is imposed;

[0119] Step 3: Taking the maximized channel optimization model as the objective function, using the evaluation function and the reward function as constraints, and using the communication data as the input, construct a path planning model.

[0120] As Figure 4 shown, design an optimized deep neural network algorithm to solve the path planning model, calculate the comprehensive coverage path of the flying device, and plan its optimal flight route. The designed path planning model inputs the surrounding grid environment information, uses the neural network to generate the next control instruction, and continuously improves the control performance of the flying device through continuous learning. The algorithm consists of multiple neural units. Each neural unit takes the output of the previous layer as the input and generates the output through a specific activation function. By stacking multiple neural units, an efficient and powerful deep neural network algorithm is constructed. Taking the full-coverage path that maximizes the throughput as the goal, using the evaluation function and the reward function as the path planning rules, using the communication data of the flying device as the input, and combining the position of the flying device as (x, y), design a deep neural network algorithm.

[0121] According to an embodiment of the present invention, using the deep neural network algorithm to solve the path planning model to obtain the full-coverage path of the flying device includes: the deep neural network algorithm includes an input layer, a hidden layer, and an output layer; wherein,

[0122] The input of the input layer is: {S 1 , S 2 , S 3 , S 4 , S 5 , Z 1 , Z 2 , Z 3 , Z 4 , Z 5 , θ i , H, V, E, W, F, A l , R(T), X, Y}; represents the obstacle information in 5 directions, namely directly to the right, right front, directly in front, left front, and directly to the left of the flying device output by identifying the intelligent reflecting surface through the two-dimensional grid map. d represents the range of the air-ground channel of the wireless sensor network. d i represents the distance of the obstacle in the i-th direction. When there is no obstacle within the range of the air-ground channel of the wireless sensor between the flying device and the ground receiver, then d i = d; represents the coverage information in 5 directions obtained from the two-dimensional grid map. t i is the number of times the flying device stays in the grid where the i-th wireless sensor end is located. t max is the maximum coverage number; H is the height, V is the speed, E is the power, W is the working duration, A l is the warning data; X, Y are the positions of the flying device;

[0123] The hidden layer includes 6 neurons, and the ReLU activation function is used: f(x) = max(0, x). When the input value is greater than 0, the value is output, otherwise 0 is output;

[0124] The output layer uses the logistic function as the activation function: where e is the natural constant and p is the change rate of the S-shaped function; the output quantity is expressed as: {y 1 , y 2 , y 3 , y 4}, y i represents a neuron of 4 consecutive values, and each neuron corresponds to the lift value of a motor, respectively controlling the movement of the flying device in the up, down, left, and right four directions. y 1 controls the upward lift, y 2 controls the downward lift, y 3Control the lift force to the left, y 4 Control the lift force to the right;

[0125] Adjust the motion flight path of the flying device in real time through the lift forces of the four output motors, so as to obtain a full-coverage path for the entire flight area.

[0126] Introduce the reinforcement learning method: The flying device learns to identify the flight paths within the entire flight area based on the obstacle information, motion data, state data, and communication throughput, and according to the path planning rules; by adjusting the lift forces of the four motors, the flight path is adjusted in real time, so as to obtain a full-coverage path for the entire flight area.

[0127] Then execute Step 4. In Step 4, set waypoints based on the full-coverage path, and dynamically adjust the distribution of waypoints using the motion data of the flying device, and output an optimized set of waypoints.

[0128] Finally, execute Step 5. In Step 5, integrate the optimized set of waypoints and the full-coverage path into an optimal path, and draw the optimal path through a Web interface and display it on an integrated map.

[0129] According to the embodiments of the present invention, the integrated map is one of the core components of the Web interface, and is responsible for loading the waypoint data of the flying device and the pre-planned full-coverage planning path onto the map. First, load the waypoints. The integrated map will load the waypoint information of the flying device's flight mission and display the specific positions of these points on the map. Then integrate the set of waypoints. In addition to displaying the positions of individual waypoints, the integrated map will also load all the optimized sets of waypoints at the same time, and display these waypoints in the order required by the mission, ensuring that the flying device can clearly execute the mission according to the path order. Then load the full-coverage planning path. The full-coverage planning path generated according to the mission requirements is loaded onto the map to show the path executed by the flying device when covering the entire target area; finally, load the map background and geographical information. The integrated map will load the base map with the geographical location of the working area of the flying device, which is convenient for users to identify the mission area.

[0130] The drawing module is a dynamic module in the Web interface responsible for generating and adjusting paths; it includes: Dynamically adjusting waypoints, the drawing module allows users to adjust the specific positions or parameters of waypoints on the map, and these changes are fed back to the system in real time, triggering the replanning and optimization of the path. Generating and drawing the optimal planned path, after loading the waypoints and the initial full-coverage path, the drawing module will automatically draw the optimal planned path on the integrated map and display this path. This process requires no manual intervention, making the entire path generation process efficient and automated, ensuring that the flying device completes tasks in an efficient and safe manner. Waypoint integration and path optimization, when the user adjusts the waypoint positions, the system will calculate a new set of waypoints in real time, re-optimize the path, and automatically update the full-coverage path to ensure that the path is closely combined with the adjusted set of waypoints, avoiding any redundant or invalid paths.

[0131] The display interface is the part where the user interacts with the system, used to display the optimal planned path and its related information. Its main functions include displaying the optimal path, path information display, and dynamic update and feedback. Displaying the optimal path is to show the generated optimal planned path on the integrated map, clearly marking the waypoints and flight routes that the flying device needs to pass through; Path information display marks the waypoint positions, task information such as hovering, flight altitude, position, etc., to help users understand the path details; Dynamic update and feedback is to update the optimized path in real time after the waypoints and the path are adjusted, ensuring that users can track the changes and make corresponding decisions.

[0132] Collaborative work between modules. The integrated map loads the waypoints and the initial full-coverage path, providing basic geographical information support for the task. After receiving the waypoints, the drawing module automatically generates the optimal path and draws it on the map, and further optimizes the path in combination with the user's dynamic adjustments. The display interface displays the optimal path and waypoint information in real time at the front end, helping users quickly grasp the path planning situation of the flying device.

[0133] In the overall work process, the user sets waypoints or imports the initial path on the interface. The system loads this information into the integrated map as the input for path optimization. The drawing module takes over this data, generates the optimal path through calculation, and automatically draws it on the map.

[0134] The display interface displays the generated optimal path in real time and dynamically updates the path display after the user adjusts or optimizes the path. This Web interface realizes the automation and visualization of the optimal planned path, and is an efficient route generation and presentation solution in the task management of flying devices, reflecting the improvement of the efficiency and operability of flying device task management in aspects such as path optimization and dynamic display.

[0135] According to another embodiment of the present invention, a path planning system based on a wireless sensor network is proposed, and this system includes:

[0136] Data acquisition module: It is configured to acquire communication data between the flying device and the ground sensor, and the communication data includes motion data, status data, and alarm data of the flying device; wherein, the motion data includes position and speed; the status data includes battery power and working duration; the alarm data includes low battery alarm, obstacle alarm, and fault alarm;

[0137] Channel optimization module: It is configured to build a channel optimization model based on optimizing the air-ground channel throughput;

[0138] Path planning module: It is configured to build a path planning model based on the communication data and the channel optimization model, and use the deep neural network algorithm to solve the path planning model to obtain the full coverage path of the flying device;

[0139] Waypoint setting module: It is configured to set waypoints based on the full coverage path, dynamically adjust the distribution of waypoints using the motion data of the flying device, and output an optimized set of waypoints;

[0140] Path drawing module: It is configured to integrate the optimized set of waypoints and the full coverage path into an optimal path, draw the optimal path through a Web interface and display it on an integrated map.

[0141] In this embodiment, preferably, the system further includes:

[0142] Waypoint setting module: It is configured to set waypoints based on the full coverage path, dynamically adjust the distribution of waypoints using the motion data of the flying device, and output an optimized set of waypoints;

[0143] Path drawing module: It is configured to integrate the optimized set of waypoints and the full coverage path into an optimal path, draw the optimal path through a Web interface and display it on an integrated map.

[0144] In this embodiment, preferably, the channel optimization model in the channel optimization module is constructed as follows:

[0145]

[0146] In the formula, M represents the number of reflection units of the intelligent reflecting surface; m represents that the flying device is in a flying state, represents the air-ground communication signal-to-noise ratio between the intelligent reflecting surface and the flying device, σ 2 is the additive white Gaussian noise power, P s represents the source sensor node power, h SD represents the channel gain between the source sensor node and the intelligent reflecting surface, represents the channel gain from the intelligent reflecting surface at the H point height to the sensor node; h SRdenotes the channel gain between the source sensor node and the intelligent reflecting surface, denotes the signal-to-noise ratio in the flight state of the flying device; and there is:

[0147]

[0148] where ρ represents the path loss, d SD denotes the distance from the source sensor node to the target sensor node, h represents the channel gain; d(φ SR -φ RD ) represents the distance between φ SR and φ RD , φ SR denotes the cosine of the angle of arrival of the signal from the source sensor node to the intelligent reflecting surface, φ RD denotes the cosine of the angle of arrival of the signal from the intelligent reflecting surface to the target sensor node, R represents the intelligent reflecting surface, D represents the target sensor node, and S represents the source sensor node; represents the path loss of the distance between the source sensor node and the intelligent reflecting surface, and τ is the path loss exponent between the source sensor node and the intelligent reflecting surface; represents the path loss of the distance between the intelligent reflecting surface and the target sensor node, and k is the loss exponent between the intelligent reflecting surface and the target sensor node; θ i denotes the optimal reflection phase of each element of the intelligent reflecting surface, λ represents the carrier wavelength, and arg(h) represents the phase angle of the channel gain h.

[0149] In this embodiment, preferably, the construction of the path planning model based on the communication data and the channel optimization model in the path planning module includes:

[0150] Step 3-1. Flight area modeling; use a two-dimensional grid map to model the flight area, and divide the flight area into a dark area and a bright area. Among them, the dark area indicates that there are obstacles in the area and the flying device is prohibited from entering, and the bright area indicates that there are no obstacles in the area and the flying device is allowed to enter; set the length of the flight area as L, the width as W, and the side length of the grid as D, then the total number of grids is:

[0151]

[0152] Step 3-2. Establish an evaluation function and a reward function;

[0153] The evaluation function is:

[0154]

[0155] In the formula: E c represents the flight path coverage rate; E r represents the flight path repetition rate; E hIndicates the high-frequency repetition rate of the flight path; A r Indicates accident penalty; b, c, d, e are set proportionality coefficients, indicating the adjustment of the task objective; among them, the flight path coverage rate is: S represents the total area of the bright area, S 1 Represents the area of the covered area, S 2 Is the area of the dark area; the flight path repetition rate is: D 1 Is the total path length, indicating the total number of steps of the flight device, D 2 Is the standard path length, indicating the number of blank grids; the high-frequency repetition rate of the flight path is: S h Represents the area of the region where the coverage times exceed a certain threshold, S r Represents the total area of the covered area; accident penalty A r Is to restrict the dark area of the flight device and the behavior of the flight device;

[0156] The reward function is used to maximize the throughput, and the reward function is:

[0157] R(T) = λ 1 ·θ i -λ 2 ·Penalty(t)

[0158] In the formula, λ 1 And λ 2 Are weight coefficients, Penalty(t) is the penalty function for violating the constraints, indicating that if the throughput at any time is lower than the minimum requirement, a penalty is imposed;

[0159] Step Three: Taking the maximized channel optimization model as the objective function, using the evaluation function and the reward function as constraints, and using the communication data as the input, construct a path planning model.

[0160] In this embodiment, preferably, in the path planning module, using the deep neural network algorithm to solve the path planning model to obtain the full-coverage path of the flight device includes: The deep neural network algorithm includes an input layer, a hidden layer, and an output layer; among them,

[0161] The input of the input layer is: {S 1 , S 2 , S 3 , S 4 , S 5 , Z 1 , Z 2 , Z 3 , Z 4 , Z 5 , θ i , H, V, E, W, F, Al , R(T), X, Y}; Indicates the obstacle information in five directions: directly to the right, right front, directly in front, left front, and directly to the left of the flying device output by identifying the intelligent reflecting surface through the two-dimensional grid map. d represents the range of the air-ground channel of the wireless sensor network, d i Indicates the distance of the obstacle in the i-th direction. When there is no obstacle within the range of the air-ground channel of the wireless sensor between the flying device and the ground receiver, then d i = d; Indicates the coverage rate information in five directions obtained from the two-dimensional grid map, t i Is the number of times the flying device stays within the grid where the i-th wireless sensor ends, t max Is the maximum number of coverage times; H is the height, V is the speed, E is the power, W is the working duration, A l Is the warning data; X, Y are the positions of the flying device;

[0162] The hidden layer includes 6 neurons, and the ReLU activation function is: f(x) = max(0, x). When the input value is greater than 0, it outputs that value, otherwise it outputs 0;

[0163] The output layer uses the logistic function as the activation function: Where e is the natural constant and p is the rate of change of the sigmoid function; the output is expressed as: {y 1 , y 2 , y 3 , y 4}}, y i Represents neurons of 4 consecutive values, and each neuron corresponds to the lift value of a motor, respectively controlling the movement of the flying device in the up, down, left, and right four directions. y 1 Controls the upward lift, y 2 Controls the downward lift, y 3 Controls the leftward lift, y 4 Controls the rightward lift;

[0164] By adjusting the lift of the four motors output in real time, the movement flight path of the flying device is adjusted, and then the full-coverage path of the entire flight area is obtained.

[0165] It should be noted that the functions of the path planning system based on the wireless sensor network described in this embodiment can be illustrated by the aforementioned path planning method based on the wireless sensor network. For the parts not detailed in this embodiment, refer to the above method embodiments.

[0166] Further verify the technical effects of the present invention through experiments.

[0167] Experimental parameter settings: Initial battery level of the flying device is 80%; Initial position longitude is 30.5, latitude is 114.5; Flying speed is 10 m / s; Flying altitude is 100 m; Flying duration is 60 min; Size of the flying area is 500 m x 500 m; Grid size for area division is 10 m, and the total number of grids is 2,500; Area environment changes: Randomly distributed obstacles, area signal strength changes; Power of the channel source sensor node is 10 W; Intelligent reflecting surface gain is 15 dB; Noise power is -100 dBm; Input layer of the deep neural network is the real-time position, status data, obstacle information, and communication throughput of the flying device; Hidden layer has 6 neurons, and the activation function is ReLU; Output layer is the lift adjustment values in four directions (up, down, left, right); Reinforcement learning parameters are learning rate of 0.01 and discount factor of 0.9.

[0168] Collect communication data between the flying device and ground sensors, including motion data, status data, longitude and latitude data, and alarm data; Build a channel optimization model based on the collected data to optimize the air-ground channel throughput, and provide effective signal information for path planning by maximizing the channel throughput and channel gain matrix; Build a path planning model based on the optimized channel throughput, use the deep neural network algorithm for path optimization to obtain a full-coverage path; In the model, the flying area is divided into bright areas and dark areas. The dark areas are obstacle areas where the flying device is prohibited from entering; The bright areas are flyable areas where the flying device is allowed to enter; Dynamically adjust the distribution of waypoints based on the motion data of the flying device to ensure that the flying device avoids obstacles and covers the entire flying area; Real-time display the optimized flight path through the Web interface and conduct visual analysis.

[0169] Under the optimized channel model, the communication throughput of the flying device has increased by approximately 30%. The optimized path selects the channel with the strongest signal, avoiding areas with low signal intensity, effectively reducing signal loss and data latency caused by poor communication quality. In areas with many obstacles or weak signals, the system can automatically avoid bad paths, maintain stable communication quality, and ensure real-time data transmission and the smooth progress of tasks. The coverage rate of the path optimized by the deep neural network algorithm in the flight area has reached 98.5%. Compared with traditional path planning methods, the optimized path can effectively avoid obstacles and cover more effective areas, reducing the occurrence of repeated paths and significantly improving the utilization efficiency of the path. The flying device avoids mid-air task stops caused by insufficient power or signal interruption by dynamically adjusting waypoints, and the task is successfully completed throughout the process. During the experiment, the power consumption of the flying device remains within an acceptable range, with an average power consumption of approximately 25%, and the power consumption is relatively uniform, avoiding the situation of incomplete tasks due to insufficient power. The optimized path design can intelligently adjust the flight strategy according to power and flight duration, avoiding excessive power consumption that affects task completion. When obstacles appear or the communication quality changes in the flight area, the system can promptly sense and adjust the path. The flying device can learn and adapt to different environmental changes, further optimize the flight path, and improve the efficiency and reliability of task execution.

[0170] Verified through experiments, the communication quality has been significantly improved. The channel optimization algorithm ensures that the flying device always maintains good communication quality during flight, with a 30% increase in throughput and more stable data transmission; the path planning efficiency has been improved. Combining deep neural networks and reinforcement learning methods, the coverage rate of the flight path has reached 98.5%, avoiding blind spots and repeated paths, and improving the efficiency of flight tasks; the entire task is guaranteed. The power of the flying device is maintained within a reasonable range, avoiding task interruptions caused by insufficient power, and ensuring the smooth completion of the flight task.

[0171] As Figure 5 shown, the experimental results demonstrate the effectiveness of the path planning method based on wireless sensor networks proposed in the present invention in flight tasks. Especially in dynamic environments, the combination of channel optimization and deep neural networks can achieve efficient and reliable flight path planning, ensuring the smooth completion of tasks.

[0172] It should be noted that although several units, modules, or sub-modules are mentioned in the above detailed description, this division is merely exemplary and not mandatory. In fact, according to the embodiments of the present invention, the features and functions of two or more of the above-described modules can be embodied in one module. Conversely, the features and functions of one module described above can be further divided and embodied by multiple modules.

[0173] Moreover, although the operations of the method of the present invention are depicted in a specific order in the drawings, this is not a requirement or implication that these operations must be performed in that specific order, or that all of the operations shown must be performed to achieve the desired result. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step for execution, and / or one step may be broken down into multiple steps for execution.

[0174] Although the spirit and principles of the present invention have been described with reference to several specific embodiments, it should be understood that the present invention is not limited to the specific embodiments disclosed, and the division of each aspect does not mean that the features in these aspects cannot be combined for benefit. Such a division is only for convenience of expression. The present invention is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.

Claims

1. A path planning method based on a wireless sensor network, characterized in that: The following steps are involved: Step 1: Collect communication data between the flight equipment and the ground sensor, wherein the communication data includes motion data, status data and alarm data of the flight equipment; wherein the motion data includes position and speed; the status data includes power level and working time; and the alarm data includes low power alarm, obstacle alarm and fault alarm; Step 2: construct a channel optimization model based on optimizing the air-ground channel throughput; Step three: construct a path planning model based on the communication data and the channel optimization model, and use a deep neural network algorithm to solve the path planning model to obtain a full coverage path for the flight equipment.

2. The path planning method based on wireless sensor network according to claim 1, characterized in that: After obtaining the full coverage path of the flight equipment, it also includes: Step 4: setting waypoints based on the full coverage path, dynamically adjusting the distribution of waypoints using the motion data of the flight device, and outputting an optimized waypoint set; Step 5: Integrate the optimized waypoint set and the full coverage path into an optimal path, draw the optimal path through a Web interface and display it on an integrated map.

3. A path planning method based on a wireless sensor network according to claim 1 or 2, characterized in that: The channel optimization model described in step 2 is constructed as follows: Where M represents the number of reflective units of the intelligent reflective surface; m represents the flight status of the flight device. represents the air-to-ground communication signal-to-noise ratio between the intelligent reflective surface and the flight equipment, σ 2 is the additive white Gaussian noise power, P s represents the source sensor node power, h SD represents the channel gain between the source sensor node and the smart reflective surface, h represents the channel gain from the smart reflector at the height of point H to the sensor node; SR represents the channel gain between the source sensor node and the smart reflective surface, Indicates the signal-to-noise ratio of the flight equipment in flight status; and: Where ρ represents the path loss, d SD represents the distance from the source sensor node to the target sensor node, h represents the channel gain; d(φ SR -φ RD ) represents φ SR With φ RD The distance between SR represents the cosine of the signal arrival angle from the source sensor node to the smart reflector, φ RD represents the cosine of the signal arrival angle from the smart reflective surface to the target sensor node, R represents the smart reflective surface, D represents the target sensor node, and S represents the source sensor node; represents the path loss between the source sensor node and the smart reflective surface, τ is the path loss index between the source sensor node and the smart reflective surface; represents the path loss between the smart reflective surface and the target sensor node, k is the loss index between the smart reflective surface and the target sensor node; θ i Indicates the optimal reflection phase of each element of the smart reflective surface, λ represents the carrier wavelength, and arg(h) represents the phase angle of the channel gain h.

4. A path planning method based on a wireless sensor network according to claim 3, characterized in that: The step 3 of constructing a path planning model based on the communication data and the channel optimization model includes: Step 31: Model the flight area. Use a two-dimensional grid map to model the flight area, and divide the flight area into dark areas and bright areas. The dark area indicates that there are obstacles in the area and the flight equipment is prohibited from entering. The bright area indicates that there are no obstacles in the area and the flight equipment is allowed to enter. Set the length of the flight area to L, the width to W, and the side length of the grid to D. Then the total number of grids is: Step 32: Establish evaluation function and reward function; The evaluation function is: In the formula, E c represents the flight path coverage; E r represents the flight path repetition rate; E h Indicates the high frequency repetition rate of the flight path; A r represents the accident penalty; b, c, d, and e are the set proportional coefficients, indicating the adjustment of the mission target; among them, the flight path coverage rate is: S represents the total area of ​​the bright area, S1 represents the area of ​​the covered area, and S2 represents the area of ​​the dark area; the flight path repetition rate is: D1 is the total path length, which indicates the total number of flight equipment steps; D2 is the standard path length, which indicates the number of blank grids; the high-frequency repetition rate of the flight path is: S h Indicates the area of ​​the region where the number of coverages exceeds a certain threshold, S r Indicates the total area of ​​the covered area; accident penalty A r To constrain dark areas of flying equipment and constrain flying equipment behavior; The reward function is used to maximize the throughput, and the reward function is: R(T)=λ1·θ i -λ2·Penalty(t) Where λ1 and λ2 are weight coefficients, and Penalty(t) is the penalty function for violating the constraint, which means that if the throughput at any time is lower than the minimum requirement, a penalty is imposed; Step 33: Taking the maximization channel optimization model as the objective function, the evaluation function and the reward function as constraints, and the communication data as input, a path planning model is constructed.

5. A path planning method based on a wireless sensor network according to claim 4, characterized in that: The step 3 of using the deep neural network algorithm to solve the path planning model to obtain the full coverage path of the flight equipment includes: the deep neural network algorithm includes an input layer, a hidden layer, and an output layer; wherein, The input layer has the following inputs: {S1,S2,S3,S4,S5,Z1,Z2,Z3,Z4,Z5,θ i ,H,V,E,W,F,A l ,R(T),X,Y}; i∈{1,2,3,4,5}, represents obstacle information in the five directions of the right, right front, front, left front and left of the flight equipment output by the intelligent reflective surface through the two-dimensional grating image recognition, d represents the range of the air-ground channel of the wireless sensor network, d i represents the obstacle distance in the i-th direction. When there is no obstacle within the range of the wireless sensor air-ground channel between the flight equipment and the ground receiver, d i =d; Represents the coverage information in five directions obtained from the two-dimensional grid map, t i is the number of times the flying device stays in the grid where the end of the i-th wireless sensor is located, t max is the maximum number of coverage times; H is the height, V is the speed, E is the power, W is the working time, A l is the warning data; X, Y is the position of the flight equipment; The hidden layer includes 6 neurons, and the ReLU activation function is: f(x) = max(0, x). When the input value is greater than 0, the value is output, otherwise 0 is output; The output layer uses the logistic function as the activation function: Where e is a natural constant, p is the rate of change of the S-shaped function; the output is expressed as: {y1, y2, y3, y4}, y i Represents 4 continuous value neurons, each neuron corresponds to the lift value of a motor, which controls the movement of the flight device in the four directions of up, down, left, and right. y1 controls the upward lift, y2 controls the downward lift, y3 controls the left lift, and y4 controls the right lift; By outputting the lift of the four motors, the flight path of the flying device is adjusted in real time to obtain a full coverage path for the entire flight area.

6. A path planning system based on a wireless sensor network, characterized in that: The following steps are involved: Data acquisition module: configured to acquire communication data between the flight equipment and the ground sensors, wherein the communication data includes motion data, status data and warning data of the flight equipment; wherein the motion data includes position and speed; the status data includes power level and working time; and the warning data includes low power warning, obstacle warning and fault warning; A channel optimization module: configured to construct a channel optimization model based on optimizing air-ground channel throughput; Path planning module: It is configured to build a path planning model based on the communication data and the channel optimization model, and use a deep neural network algorithm to solve the path planning model to obtain a full coverage path for the flight equipment.

7. The path planning system based on wireless sensor network according to claim 6, characterized in that: Also includes: A waypoint setting module: configured to set waypoints based on the full coverage path, dynamically adjust the distribution of waypoints using the motion data of the flight device, and output an optimized waypoint set; Path drawing module: It is configured to integrate the optimized waypoint set and the full coverage path into an optimal path, draw the optimal path through a Web interface and display it on an integrated map.

8. A path planning system based on a wireless sensor network according to claim 7, characterized in that: The channel optimization model in the channel optimization module is constructed as follows: Where M represents the number of reflective units of the intelligent reflective surface; m represents the flight status of the flight device. represents the air-to-ground communication signal-to-noise ratio between the intelligent reflective surface and the flight equipment, σ 2 is the additive white Gaussian noise power, P s represents the source sensor node power, h SD represents the channel gain between the source sensor node and the smart reflective surface, h represents the channel gain from the smart reflector at the height of point H to the sensor node; SR represents the channel gain between the source sensor node and the smart reflective surface, Indicates the signal-to-noise ratio of the flight equipment in flight status; and: Where ρ represents the path loss, d SD represents the distance from the source sensor node to the target sensor node, h represents the channel gain; d(φ SR -φ RD ) represents φ SR With φ RD The distance between SR represents the cosine of the signal arrival angle from the source sensor node to the smart reflector, φ RD represents the cosine of the signal arrival angle from the smart reflective surface to the target sensor node, R represents the smart reflective surface, D represents the target sensor node, and S represents the source sensor node; represents the path loss between the source sensor node and the smart reflective surface, τ is the path loss index between the source sensor node and the smart reflective surface; represents the path loss between the smart reflective surface and the target sensor node, k is the loss index between the smart reflective surface and the target sensor node; θ i Indicates the optimal reflection phase of each element of the smart reflective surface, λ represents the carrier wavelength, and arg(h) represents the phase angle of the channel gain h.

9. A path planning system based on a wireless sensor network according to claim 8, characterized in that: The construction of a path planning model based on the communication data and the channel optimization model in the path planning module includes: Step 31: Model the flight area. Use a two-dimensional grid map to model the flight area, and divide the flight area into dark areas and bright areas. The dark area indicates that there are obstacles in the area and the flight equipment is prohibited from entering. The bright area indicates that there are no obstacles in the area and the flight equipment is allowed to enter. Set the length of the flight area to L, the width to W, and the side length of the grid to D. Then the total number of grids is: Step 32: Establish evaluation function and reward function; The evaluation function is: In the formula, E c represents the flight path coverage; E r represents the flight path repetition rate; E h Indicates the high frequency repetition rate of the flight path; A r represents the accident penalty; b, c, d, and e are the set proportional coefficients, indicating the adjustment of the mission target; among them, the flight path coverage rate is: S represents the total area of ​​the bright area, S1 represents the area of ​​the covered area, and S2 represents the area of ​​the dark area; the flight path repetition rate is: D1 is the total path length, which indicates the total number of flight equipment steps; D2 is the standard path length, which indicates the number of blank grids; the high-frequency repetition rate of the flight path is: S h Indicates the area of ​​the region where the number of coverages exceeds a certain threshold, S r Indicates the total area of ​​the covered area; accident penalty A r To constrain dark areas of flying equipment and constrain flying equipment behavior; The reward function is used to maximize the throughput, and the reward function is: R(T)=λ1·θ i -λ2·Penalty(t) Where λ1 and λ2 are weight coefficients, and Penalty(t) is the penalty function for violating the constraint, which means that if the throughput at any time is lower than the minimum requirement, a penalty is imposed; Step 33: Taking the maximization channel optimization model as the objective function, the evaluation function and the reward function as constraints, and the communication data as input, a path planning model is constructed.

10. A path planning system based on a wireless sensor network according to claim 9, characterized in that: The path planning module uses the deep neural network algorithm to solve the path planning model to obtain the full coverage path of the flight equipment, which includes: the deep neural network algorithm includes an input layer, a hidden layer, and an output layer; wherein, The input layer has the following inputs: {S1,S2,S3,S4,S5,Z1,Z2,Z3,Z4,Z5,θ i ,H,V,E,W,F,A l ,R(T),X,Y}; i∈{1,2,3,4,5}, represents obstacle information in the five directions of the right, right front, front, left front and left of the flight equipment output by the intelligent reflective surface through the two-dimensional grating image recognition, d represents the range of the air-ground channel of the wireless sensor network, d i represents the obstacle distance in the i-th direction. When there is no obstacle within the range of the wireless sensor air-ground channel between the flight equipment and the ground receiver, d i =d; Represents the coverage information in five directions obtained from the two-dimensional grid map, t i is the number of times the flying device stays in the grid where the end of the i-th wireless sensor is located, t max is the maximum number of coverage times; H is the height, V is the speed, E is the power, W is the working time, A l is the warning data; X, Y is the position of the flight equipment; The hidden layer includes 6 neurons, and the ReLU activation function is: f(x) = max(0, x). When the input value is greater than 0, the value is output, otherwise 0 is output; The output layer uses the logistic function as the activation function: Where e is a natural constant, p is the rate of change of the S-shaped function; the output is expressed as: {y1, y2, y3, y4}, y i Represents 4 continuous value neurons, each neuron corresponds to the lift value of a motor, which controls the movement of the flight device in the four directions of up, down, left, and right. y1 controls the upward lift, y2 controls the downward lift, y3 controls the left lift, and y4 controls the right lift; By outputting the lift of the four motors, the flight path of the flying device is adjusted in real time to obtain a full coverage path for the entire flight area.

Images