A method and system for unmanned aerial vehicle flight control for gas pipeline inspection data collection

Abstract

The application discloses a kind of unmanned aerial vehicle flight control method and system of gas pipeline inspection data acquisition, by unmanned aerial vehicle in gas pipeline area inspection and ground sensor data collection task is modeled as Markov process model, obtain unmanned aerial vehicle flight control system state space, build actor network and critic network guide unmanned aerial vehicle to execute optimal action policy, by building sensor data state representation network, sensor historical data and current data are optimized and predicted, obtain complete sensor time series information state.The application can collect target place methane gas concentration measurement value in real time, quickly and accurately, accurately detect the working state of gas pipeline.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicle (UAV) control and relates to a UAV flight control method and system for collecting gas pipeline inspection data. Background Technology

[0002] Natural gas pipelines are a crucial component of the natural gas system, and their safe and reliable operation is directly related to the stable development of the national economy, as well as environmental protection and pollution control. Because some sections of gas pipelines are exposed to the natural environment for extended periods, they not only bear normal mechanical loads and the internal pressure of gas transmission, but also endure external damage such as rainfall, snowfall, landslides, and human-caused perforations. These factors accelerate the aging of pipeline components, and if not detected and addressed promptly, they can develop into various malfunctions, posing a serious threat to the safety and stability of the natural gas system.

[0003] Natural gas's main component is methane. Ground sensors detect methane concentrations near natural gas pipelines to determine if leaks are present. Currently, the most common method for natural gas pipeline inspection is still manual patrol by personnel along the pipeline route. This method relies heavily on manual recording, leading to incomplete inspections, missed points, and serious problems with incomplete, non-standard, and inaccurate data storage, as well as data loss. Furthermore, manual inspection suffers from long reporting times for potential hazards (such as pipeline corrosion and leaks) and unclear descriptions of the specific coordinates of hazard locations. The diverse environments of gas pipelines, including plains, mountains, hills, and deserts, further complicate inspections due to the complex terrain. Summary of the Invention

[0004] The purpose of this invention is to solve the problems of inadequate manual inspection, missed inspection points, incomplete, non-standard, inaccurate, and lost data, long reporting time, and difficult inspection in the prior art, and to provide a UAV flight control method and system for gas pipeline inspection data acquisition.

[0005] To achieve the above objectives, the present invention employs the following technical solution:

[0006] A flight control method for unmanned aerial vehicles (UAVs) for collecting gas pipeline inspection data includes:

[0007] The task of UAV inspection in the gas pipeline area and data collection from ground sensors is modeled as a Markov process model to obtain the state space of the UAV flight control system.

[0008] Build networks for actors and critics;

[0009] Train the actor network and critic network, and obtain the current system state space of the UAV and the action space of the UAV inspection flight system based on the state space of the UAV flight control system;

[0010] Collect gas pipeline inspection data sent by ground sensors and construct a sensor data status representation network;

[0011] Based on the sensor data state characterization network, historical data collected by ground sensors and current gas pipeline inspection data, optimization prediction is performed to obtain the state of data that has not yet been transmitted and complete sensor time-series state information, which is then sent to the UAV to adjust the UAV system state space and action space.

[0012] A further improvement of the present invention is that:

[0013] Furthermore, the tasks of UAV inspection in the gas pipeline area and data collection from ground sensors are modeled as a Markov process, resulting in the state space of the UAV flight control system, specifically:

[0014]

[0015] Where b UAV (t) represents the UAV battery status, and i represents the i-th sensor among N ground sensors; b i (t) and q i (t) represents the battery status of the ground sensor and the length of the data transmitted by the sensor, respectively; g i (t) represents the data transmission channel gain of the i-th sensor at time t; to estimate the energy consumed and data arrival time for collecting data from unplanned ground sensors, γ is used. i (t) represents the time span parameter, γ when the collected sensor data is within the task plan. i When the value of γ is 1, it means that the collected sensor data is not in the task schedule list. i The value of s(t) is 0; s(t) represents the timing state of the selected sensor; the UAV maintains a constant altitude flight, and (x(t),y(t),z) represents the UAV's position coordinates at time t.

[0016] Furthermore, establish actor and critic networks, specifically as follows:

[0017] Establish a current critic network and a target critic network, both using the same structure, to simulate the system state-action value function. This function represents a state space of α. k In this case, perform the action The value generated; establishing the current actor network and the target actor network, both using the same network structure, to simulate the action policy function μ{α k |w μ}, this function represents the state space α k Below, the actions performed by the drone, wμ This is the actor network parameter matrix.

[0018] Furthermore, training will be provided to actor networks and critic networks, specifically as follows:

[0019] In a simulated numerical simulation environment, the battery state of the UAV at flight time t is calculated to obtain the system state space of the UAV at the current time; the action space of the UAV inspection flight system is calculated, which includes the position and velocity of the UAV at the next moment and the sensor number of the collected data.

[0020] The algorithm for calculating the battery state of the drone at flight time t is as follows:

[0021] b UAV (t)=b UAV (t-1)+Δb UAV (t)-ΔE UAV (t) (2)

[0022] Where Δb UAV (t) represents the amount of solar energy collected by the drone from time t-1 to time t; using B UAV This indicates the power threshold required for the drone to return to the ground charging station from its current location; during the flight inspection, the drone needs to maintain b UAV (t)≥B UAV ;ΔE UAV (t) represents the amount of electricity consumed by the drone at time t, calculated as follows:

[0023]

[0024] Where P0 and P′0 are constants, ω(t) is the UAV motor speed, v0 is the average rotor speed during hovering, v(t) is the instantaneous speed of the UAV, and ξ... drag and ξ rotor These represent the fuselage drag ratio and rotor rigidity, respectively. air and S rotor These represent air density and rotor disk area, respectively.

[0025] Furthermore, the action space of the UAV inspection flight system is calculated. The action space includes the UAV's position and velocity at the next moment, as well as the sensor numbers of the collected data. Specifically:

[0026] a α =((x′(α),y′(α),z),(v x (α),v y (α)),i α (4)

[0027] Where (x′(α),y′(α),z) represents the UAV position at the next moment, and (v x (α),v y (α) represents the velocity of the UAV in the horizontal direction at the next moment, i α The sensor number to be collected in the current state space; a α ∈A, A is the action space set, which is the collection of all actions that the drone can take to optimize the drone's next position and speed, as well as the selected ground sensors used for data collection.

[0028] Further training for actor and critic networks also includes:

[0029] The current commentator network is trained using the Bellman algorithm, and the loss function is expressed as:

[0030]

[0031] Where δ is the discount factor, Q ′ Let Q be the output of the target critic network and Q be the output of the current critic network. The parameters w of the current critic network are updated using backpropagation based on the loss function. Q A soft update algorithm is used to periodically update the target commentator network parameters w. Q′ :

[0032] w Q′ ←τw Q +(1-τ)w Q′ (6)

[0033] Where τ is the soft update operator;

[0034] The gradient ascent method is used to optimize the above equation and update the gradient of the current actor network parameters:

[0035]

[0036] A soft update algorithm is used to periodically update the target commenter network parameters w. Q′ :

[0037] w μ′ ←τw μ +(1-τ)w μ′ (8).

[0038] Furthermore, gas pipeline inspection data transmitted by ground sensors are collected to construct a sensor data status representation network, specifically:

[0039] The sensor state representation network consists of long short-term memory (LSTM) network modules. The output of the last block in the network serves as the output of the state representation network. The input sequence is processed by adding new sensor information to the memory. The output of the state representation network is calculated through gates that control the degree to which new information is stored, old information is discarded, and current information is utilized.

[0040]

[0041] Where σ is the sigmoid function, {W o W c W f W p} is the weight matrix, {e o ,e c ,e f ,e p} is the offset matrix; whenever the UAV selects a sensor to communicate with, the device reports its past and unreported states, and the report is input as a timing signal into the sensor state characterization network to obtain the complete sensor timing information state.

[0042] Furthermore, the drone also includes information interaction with the ground station, which receives and analyzes the drone's flight data and obtains the sensor status of the pipeline inspection area.

[0043] A UAV flight control system for gas pipeline inspection data acquisition includes:

[0044] The construction module is used to model the UAV's inspection of gas pipeline areas and data collection from ground sensors as a Markov process model to obtain the state space of the UAV flight control system.

[0045] A module for building actor networks and critic networks;

[0046] The training module is used to train the actor network and the critic network, and obtains the current system state space of the UAV and the action space of the UAV inspection flight system based on the state space of the UAV flight control system.

[0047] The data acquisition module is used to collect gas pipeline inspection data sent by ground sensors and construct a sensor data status representation network.

[0048] The acquisition module optimizes and predicts based on the sensor data state characterization network, historical data collected by ground sensors, and current gas pipeline inspection data. It acquires the state of data that has not yet been transmitted and complete sensor time-series state information, and sends it to the UAV to adjust the UAV system state space and action space.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] This invention models the tasks of UAV inspection of gas pipeline areas and ground sensor data collection as a Markov process model, obtaining the state space of the UAV flight control system. An actor network and a commentator network are built to guide the UAV in executing optimal action strategies. By constructing a sensor data state representation network, historical and current sensor data are optimized and predicted to obtain complete sensor temporal information. This invention collects methane gas concentration measurements at target locations in real time, quickly, and accurately, enabling precise detection of the operational status of gas pipelines. Attached Figure Description

[0051] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 This is a schematic diagram of the structure of the method of the present invention;

[0053] Figure 2 This is a schematic diagram of the sensor state representation network model proposed in this invention;

[0054] Figure 3 This is a structural diagram of the UAV flight control system for gas pipeline inspection data acquisition according to the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0056] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0057] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0058] In the description of the embodiments of the present invention, it should be noted that if terms such as "upper," "lower," "horizontal," or "inner" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of the invention is in use, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. Furthermore, terms such as "first" and "second" are only used to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0059] Furthermore, the use of the term "horizontal" does not imply that the component must be absolutely horizontal, but rather that it can be slightly tilted. For example, "horizontal" simply means that its direction is more horizontal than "vertical," and does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0060] In the description of the embodiments of the present invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in the present invention according to the specific circumstances.

[0061] The present invention will now be described in further detail with reference to the accompanying drawings:

[0062] Reference Figure 1 and Figure 2 This invention discloses a flight control method for a gas pipeline inspection data acquisition drone, comprising the following steps:

[0063] Step 1: Model the UAV's inspection of the gas pipeline area and the data collection from ground sensors as a Markov process model. First, the system state space is represented as:

[0064]

[0065] Among them, b UAV (t) represents the UAV battery status, and i represents the i-th sensor among N ground sensors; b i (t) and q i(t) represents the battery status of the ground sensor and the length of the data transmitted by the sensor, respectively; g i (t) represents the data transmission channel gain of the i-th sensor at time t; to estimate the energy consumed and data arrival time for collecting data from unplanned ground sensors, γ is used. i (t) represents the time span parameter, γ when the collected sensor data is within the task plan. i When the value of γ is 1, it means that the collected sensor data is not in the task schedule list. i The value of s(t) is 0; s(t) represents the timing state of the selected sensor; the UAV maintains a constant altitude flight, and (x(t),y(t),z) represents the UAV's position coordinates at time t.

[0066] Preferably, the drone is an octocopter drone, the drone battery is a lithium battery connected in series, and the NVIDIA TX2 board is used as the onboard data processing computer.

[0067] Step 2: Begin the training process. In the simulated numerical environment, obtain the current system state space of the UAV. During training, the parameter values ​​of each component in the system state space are obtained by sensor measurements and calculations in the simulation environment. In particular, the battery state of the UAV at flight time t is calculated using the following formula, and the battery state is part of the system state space:

[0068] b UAV (t)=b UAV (t-1)+Δb UAV (t)-ΔE UAV (t) (2)

[0069] Where, Δb UAV (t) represents the amount of solar energy collected by the drone from time t-1 to time t; using B UAV This indicates the power threshold required for the drone to return to the ground charging station from its current location. During the flight inspection, the drone needs to maintain a [value / value]. UAV (t)≥B UAV ΔE UAV (t) represents the amount of electricity consumed by the drone at time t, calculated as follows:

[0070]

[0071] Where P0 and P0′ are constants, ω(t) is the UAV motor speed, v0 is the average rotor speed during hovering, v(t) is the instantaneous speed of the UAV, and ξ is the speed of the UAV motor. drag and ξ rotor These represent the fuselage drag ratio and rotor rigidity, respectively. air and S rotorThese represent air density and rotor disk area, respectively.

[0072] Step 3: In the numerical simulation environment, calculate the motion space of the UAV inspection flight system. The motion space includes the UAV's position and velocity at the next moment, as well as the sensor numbers of the collected data.

[0073] a α =((x′(α),y′(α),z),(v x (α),v y (α)),i α (4)

[0074] Where (x′(α),y′(α),z) represents the UAV position at the next moment, and (v x (α),v y (α) represents the velocity of the UAV in the horizontal direction at the next moment, i α This refers to the sensor number to be collected in the current state space. α ∈A, A is the action space set, which is the collection of all actions that the drone can take to optimize the drone's next position and speed, as well as the selected ground sensors used for data collection.

[0075] Using L{β|α,a α} indicates that the system takes action a in the current state space. α At that time, the amount of packet loss in ground sensor data transmission.

[0076] Step 4: The UAV flight control system consists of a critic network and an actor network. The actor network outputs the action space used by the UAV to perform the flight measurement task at the current moment. The critic network evaluates the output of the actor network during training, thereby guiding the network update process. A current critic network and a target critic network are established, both using the same structure, to simulate the system state-action value function. This function represents a state space of α. k In this case, perform the action The value generated; establishing the current actor network and the target actor network, both using the same network structure, to simulate the action policy function μ{α k |w μ}, this function represents the state space α k Below, the actions performed by the drone, w μ Let be the parameter matrix of the actor network. The critic network takes the system state space and action space as input and outputs the action value Q function; the actor network takes the system state space as input and outputs the optimal action space to be executed by the drone.

[0077] Step 5: Train the current commentator network using the Bellman algorithm. The loss function is expressed as:

[0078]

[0079] Where δ is the discount factor, Q′ is the target critic network output, and Q is the current critic network output. The parameters w of the current critic network are updated using the backpropagation algorithm based on the loss function. Q A soft update algorithm is used to periodically update the target commenter network parameters w. Q′ :

[0080] w Q′ ←τw Q +(1-τ)w Q′ (6)

[0081] Where τ is the soft update operator.

[0082] The gradient ascent method is used to optimize the above equation and update the gradient of the current actor network parameters:

[0083]

[0084] A soft update algorithm is used to periodically update the target commenter network parameters w. Q′ :

[0085] w μ′ ←τw μ +(1-τ)w μ′ (8)

[0086] Preferably, the critic network contains a 5-layer hidden network, and the actor network structure contains a 7-layer hidden network.

[0087] Based on the above algorithm iteration steps, the training process of the critic network and the actor network is completed, and the model network parameters are obtained.

[0088] Step 6: Deploy the trained critic network and actor network on a drone. The drone flies over the gas pipeline area. Ground sensors transmit pipeline status data to the communication channel. The drone receives sensor data from the current communication channel and establishes a ground sensor state representation network. The function of the state representation network is to predict the state of data that has not yet been transmitted by capturing the long-term representation characteristics of time-series data, and to generate the predicted state and complete time-series sensor state information. Provided to drones. The sensor state representation network consists of long short-term memory network modules, with the output of the last block in the network serving as the output of the state representation network. The input sequence is processed by adding new sensor information to memory. The output of the state representation network is calculated through gates that control the degree to which new information is stored, old information is discarded, and current information is utilized. The selected sensor timing state is part of the system state space:

[0089]

[0090] Where σ is the sigmoid function, {W o W c W f W p} is the weight matrix, {e o ,e c ,e f ,e p} is the offset matrix. Whenever the UAV selects a sensor to communicate with, the device reports its past and unreported states (associated with each time slot since the last report from the device), and the report is input as a timing signal into the sensor state characterization network to obtain the complete sensor timing information state.

[0091] Preferably, the state representation network consists of 50 long short-term memory blocks.

[0092] Step 7: Communication between the UAV and the ground station utilizes a 5G airborne network communication system. The airborne system is centered around a single-board computer, communicating with the UAV flight control computer via a serial port, and employs a Balong baseband chip 5G module. The ground-side system runs low-level communication algorithms, capable of reconstructing the transmission sequence from multiple messages. Based on the characteristics of UAV-ground station data transmission, it maximizes system performance to meet the transmission requirements between the UAV and ground station, parses airborne data, and allows the ground station to obtain sensor status information for the pipeline inspection area.

[0093] Preferably, the drone and the ground communication system adopt the MH5000-31p 5G communication module and the DIGI XBee data transmission module. The 5G communication module requires external independent power supply, with a maximum input voltage of 4.2V, a minimum input voltage of 3.7V, and a typical value of 4.0V. For the data communication protocol algorithm, the data stream is divided into data packets with a size not exceeding 200 bytes, and a unique sequence number (seq) is assigned to each data packet. The sequence number uniquely identifies the data packet and increases in the order of the sending time. The data output of the communication fusion algorithm must be in the order of the data sending time. Specifically, if the sequence numbers corresponding to the data packet sequence output by the communication data fusion algorithm are {seq1, seq2,..., seqn}, then it is necessary to ensure that seq1 < seq2 <... < seqn. Within a certain time window, the received data packets are not output. Only when there are data packets exceeding the time window, the data packets are output in the specified order (seq order) until there are no data packets exceeding the time window. By adjusting the size of the time window, a balance can be found between low latency and reliability, so as to restore the sending sequence with high accuracy.

[0094] See Figure 3 , the present invention discloses an unmanned aerial vehicle flight control system for collecting inspection data of gas pipelines, including:

[0095] A construction module, which is used to model the inspection of the drone in the gas pipeline area and the task of collecting ground sensor data as a Markov process model to obtain the state space of the drone flight control system;

[0096] A building module, which is used to build an actor network and a critic network;

[0097] A training module, which is used to train the actor network and the critic network, and based on the state space of the drone flight control system, obtain the current system state space of the drone and the action space of the drone inspection flight system;

[0098] An acquisition module, which is used to collect the inspection data of the gas pipeline sent by the ground sensor and build a sensor data state characterization network;

[0099] An acquisition module, which is used to optimize and predict based on the sensor data state characterization network, the historical data collected by the ground sensor, and the current inspection data of the gas pipeline, obtain the state of the data not yet transmitted and the complete sensor timing state information, and send it to the drone to adjust the system state space and action space of the drone.

[0100] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for unmanned aerial vehicle flight control for gas pipeline inspection data collection, characterized in that, The application relates to a gas pipeline unmanned aerial vehicle (UAV) patrol system and method. The UAV patrol and ground sensor data collection task are modeled as a Markov process model, so as to obtain a UAV flight control system state space, specifically as follows: wherein represents the UAV battery status, i represents the i-th sensor numbered in the N ground sensors; and respectively represent the ground sensor battery status and the data length transmitted by the sensor; represents the data transmission channel gain of the i-th sensor at time t; in order to estimate the energy consumed by the collected ground sensor data in the non-planned list and the data arrival time, the following formula is used represents the time span parameter, when the collected sensor data is in the task plan, the value is 1, when the collected sensor data is not in the task plan list, the value is 0; is the selected sensor timing state; the UAV keeps constant height flight, and the following formula is used represents the UAV position coordinates at time t; An actor network and a critic network are built, specifically as follows: A current critic network and a target critic network are established, both of which have the same structure and are used to simulate a system state-action value function , which represents a state space as , and the value generated by performing an action ; a current actor network and a target actor network are established, both of which have the same network structure and are used to simulate an action policy function , which represents an action performed by the UAV in a state space , is an actor network parameter matrix; The actor network and the critic network are trained based on the UAV flight control system state space, so as to obtain a current system state space of the UAV and an action space of the UAV patrol flight system; the training of the actor network and the critic network is specifically as follows: In a simulation numerical simulation environment, the battery state of the UAV at a t flight moment is calculated, so as to obtain the current system state space of the UAV; the action space of the UAV patrol flight system is calculated, and the action space comprises the position and speed of the UAV at a next moment and the sensor number of the collected data. The algorithm for calculating the battery state of the UAV at the t flight moment is as follows: wherein represents the solar energy collected by the UAV at time t-1 to t; using B UAV represents the threshold of the energy required for the UAV to return to the ground charging station from the current location; during the flight inspection process, the UAV needs to maintain ; represents the energy consumed by the UAV at time t, which is calculated as: wherein and are constants, is the UAV motor speed, is the average rotor speed in hover, is the instantaneous speed of the UAV, and are the fuselage drag ratio and rotor solidity, respectively, and are the air density and rotor disc area, respectively; The action space of the UAV patrol flight system comprises the position and speed of the UAV at a next moment and the sensor number of the collected data, and the calculation of the action space is specifically as follows: wherein, is the position of the drone at the next time instant, is the velocity of the drone in the horizontal plane at the next time instant, is the number of sensors to be collected under the current state space; , a set of action spaces by collecting all the actions that the drone can take to optimize the next position and velocity of the drone and the selected ground sensors for data collection; The training of the actor network and the critic network further comprises the following steps: The current critic network is trained by using a Bellman algorithm, and a loss function is represented as follows: wherein is a discount factor, is a target critic network output, is a current critic network output, the current critic network parameters are updated using a backpropagation algorithm according to a loss function ; the target critic network parameters are periodically updated using a soft update algorithm : wherein is a soft update operator; The current actor network parameter gradient is updated by using a gradient ascent method to optimize and calculate the above formula. The soft update algorithm is adopted to periodically update the parameters of the target critic network : Gas pipeline patrol data transmitted by a ground sensor are collected, and a sensor data state representation network is constructed. Based on the sensor data state representation network, historical data collected by the ground sensor and current gas pipeline patrol data, optimization prediction is carried out to obtain the state of data that has not been transmitted and complete sensor time sequence state information, which are transmitted to the UAV to adjust the system state space and the action space of the UAV.

2. The gas pipeline inspection data collection drone flight control method of claim 1, wherein, The gas pipeline patrol data transmitted by the ground sensor are collected, and the sensor data state representation network is constructed, specifically as follows: The sensor state representation network is composed of a long short-term memory network module, and an output of a last block in the network is taken as an output of the state representation network; the input sequence is processed by adding new information of the sensor to the memory, and the state representation network output is calculated through a gate that controls a degree to which the new information is stored, the old information is discarded, and the current information is utilized : wherein is a sigmoid function, is a weight matrix, is a bias matrix; whenever a UAV selects a sensor to communicate, the device reports its past and unreported state, reporting as a time series signal input to the sensor state representation network, resulting in a complete sensor time series information state.

3. The gas pipeline inspection data collection drone flight control method of claim 2, wherein, The UAV further comprises a ground station for information interaction, the ground station receives flight data of the UAV and carries out analysis, and the ground station obtains the sensor state of the pipeline patrol area.

4. A UAV flight control system for gas pipeline inspection data collection, based on the UAV flight control method for gas pipeline inspection data collection of claim 1, characterized in that, The application relates to a gas pipeline unmanned aerial vehicle (UAV) patrol system and method. The application relates to a gas pipeline unmanned aerial vehicle (UAV) patrol system and method. The application relates to a gas pipeline unmanned aerial vehicle (UAV) patrol system and method. The application relates to a gas pipeline unmanned aerial vehicle (UAV) patrol system and method. The application relates to a gas pipeline unmanned aerial vehicle (UAV) patrol system and method. The application relates to a gas pipeline unmanned aerial vehicle (UAV) patrol system and method.

Images