Unmanned aerial vehicle autonomous navigation remote sensing acquisition system and method

Abstract

The invention provides an unmanned aerial vehicle autonomous navigation remote sensing acquisition system and method, and the system comprises a task planning module, a navigation and control module, a remote sensing data acquisition module, a communication monitoring module, a data processing module and an emergency guarantee module. The navigation and control module obtains accurate position information by using a GPS global satellite navigation system, and keeps stable flight in combination with an inertial measurement unit and an auxiliary sensor. By introducing an intelligent dynamic task re-planning mechanism and a multi-agent system, the unmanned aerial vehicle group can share information in real time and synchronously adjust the flight path, the execution efficiency and flexibility of the whole task are remarkably improved, and meanwhile, in the autonomous obstacle avoidance process, the unmanned aerial vehicle group can perform autonomous obstacle avoidance on the basis of collaborative operation and alternate replacement of local information. Secondary dynamic adjustment of the unmanned aerial vehicle in the emergency adjustment state is achieved, and the overall intelligent planning performance is higher.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle control, and in particular to an unmanned aerial vehicle autonomous navigation remote sensing acquisition system and method. Background Art

[0002] With the development of drone technology and the continuous expansion of its application fields, traditional single drone operations have been unable to meet the increasingly complex mission requirements, especially in scenarios such as wide-area monitoring and remote sensing data collection. Group collaboration consisting of multiple drones is particularly important. However, most current drone systems still rely on preset paths or manual intervention to complete tasks, lacking sufficient autonomous decision-making capabilities and effective response to dynamic environmental changes. In the event of emergencies, the obstacle avoidance measures taken by different drones will result in different judgments due to the different angles of the detected obstacles, which can easily cause mutual interference between drones and cause data anomalies.

[0003] Therefore, in response to the above problems, a UAV autonomous navigation remote sensing acquisition system and method are now developed. Summary of the invention

[0004] In order to overcome the shortcomings of existing UAV control technology that lacks sufficient autonomous decision-making ability and effective response to dynamic environmental changes, in emergencies, the obstacle avoidance measures taken by different UAVs will be different due to the different angles of the detected obstacles, which may easily cause mutual interference between UAVs and cause data anomalies. The present invention provides a UAV autonomous navigation remote sensing acquisition system and method.

[0005] The technical solution of the present invention is: an unmanned aerial vehicle autonomous navigation remote sensing acquisition system, comprising:

[0006] A mission planning module, which is used to determine the geographical scope to be monitored according to user needs, and simultaneously set flight parameters, and to monitor surrounding factors in real time during autonomous flight to optimize the flight path;

[0007] A navigation and control module, which uses the GPS global satellite navigation system to obtain precise location information, and combines an inertial measurement unit and auxiliary sensors to maintain stable flight, and relies on laser radar and ultrasonic sensors to perceive the surrounding environment during flight;

[0008] Remote sensing data acquisition module, which relies on the remote sensing carried by the drone to collect data, ensuring that the timestamp and geographic location information match the collected data;

[0009] The communication monitoring module is used to transmit the collected key data back to the ground station and simultaneously receive manual operation instructions from the ground station.

[0010] As a preferred embodiment of the present invention, it also includes an intelligent dynamic task re-planning mechanism, which adopts the following method: adopting a reinforcement learning framework to allow drones to learn the optimal strategy by continuously trying different decisions, and then introducing a multi-agent system to use multiple drones to work together, share information synchronously, optimize the overall task architecture, and perform spatiotemporal prediction and adjustment in this process to predict changes in environmental conditions in the future and achieve secondary dynamic processing during the adjustment process.

[0011] As a preferred embodiment of the present invention, it also includes an adaptive environmental perception and response mechanism, which utilizes situational awareness to schedule sensor combinations, extracts and classifies features of monitored environmental factors, improves the accuracy of obstacle recognition, and uses a Kalman filter to fuse different data within the sensor combination to reduce noise interference.

[0012] As a preferred embodiment of the present invention, it also includes high-dimensional space path planning and risk avoidance mechanism, which performs path planning based on sampling method, creates a virtual force field during the calculation process to guide the drone to avoid dangerous areas, uses evolutionary algorithm to search for the optimal solution, and uses local improvement strategy to accelerate convergence action.

[0013] As a preferred embodiment of the present invention, a data processing module is also included, which is used to correct, splice, classify and other processes on the collected data, extract useful information, and generate maps, reports and other forms to intuitively display the results after the processing is completed to support decision-making.

[0014] As a preferred embodiment of the present invention, an emergency protection module is also included. The emergency protection module is provided with functions such as automatic return and landing to protect the safety of the equipment when a fault occurs.

[0015] As a preferred embodiment of the present invention, the intelligent dynamic task replanning mechanism adopts the following formula for calculation:

[0016] Q(s t , a t )←Q(s t , a t )+α[r t+1 +γmax a Q(s t+1 , a)-Q(s t , a t )],

[0017] Among them, s t is the state data at time step t, a t is the action data set that needs to be taken, r t+1is the immediate reward, α is the learning rate, and γ is the discount factor;

[0018] After the calculation is completed, the loss function is introduced for definition and planning, so as to express the current network parameters and the target network parameters, and the policy gradient update is performed after the data is obtained. The policy gradient update rule is:

[0019] where π i represents the strategy of the ith agent, Q π (s, a) is the value function of the joint action of all intelligent agents, that is, multiple drones conduct rapid collaborative cooperation to evaluate the overall performance of the collaboration between multiple drones, and continuously adjust the collaborative operation status between drone groups based on the obtained data, so as to make secondary dynamic adjustments under collaborative operation according to the predicted environmental status during the initial autonomous adjustment of the drones.

[0020] As a preferred embodiment of the present invention, the adaptive environmental perception and response mechanism is implemented by the following method: the most suitable sensor combination is selected through the context awareness technology, and each UAV dynamically determines which sensors should be activated to obtain the best data according to the current mission requirements, environmental characteristics and its own state. The sensor scheduling can be evaluated by the set utility function U(s, θ), where s is the environmental state and θ is the sensor configuration parameter. The specific formula is as follows:

[0021] U(s,θ)=∑ i w i f i (s,θ i ), where w i is the set weight, which needs to be dynamically evaluated according to the drone status. i is the performance index function of different sensors;

[0022] After determining the scheduling of the sensor, the command output of feature extraction and classification is sent to the sensor. The trained model is used to determine the input image and model parameters, and the category probability distribution calculation is performed on the new data obtained by the sensor to determine whether there are obstacles or other objects of interest.

[0023] Then, the system decides whether to take action based on the judgment result. If it takes action, it uses the following formula to calculate and obtain the average consensus of the drone group, update the environmental understanding and make corresponding dynamic adjustments. The formula is as follows:

[0024] Where i represents different references to drones, and the initial local estimate is N iis the set of drone swarms to which the mechanism is applied, and α is the set step size parameter;

[0025] When performing emergency obstacle avoidance, the decision rules are used to execute obstacle avoidance actions. The adjustment area range needs to be set between 0-1. When it is within this area, emergency obstacle avoidance will be performed, and in this process, dynamic adjustments will be made continuously based on environmental perception. When it is not within this area, it will fly normally.

[0026] By adopting the above technical solution, the present invention has the following advantages:

[0027] By introducing an intelligent dynamic task re-planning mechanism and a multi-agent system, the present invention enables a group of drones to share information in real time and synchronously adjust the flight path, significantly improving the overall task execution efficiency and flexibility. At the same time, in the process of autonomous obstacle avoidance, the invention relies on collaborative operations and alternating replacement of local information to achieve secondary dynamic adjustment of drones in emergency adjustment states. The overall intelligent planning is stronger, and the problem of mutual interference between drones can also be avoided during group operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a schematic diagram of the first partial structure of the UAV autonomous navigation remote sensing acquisition system of the present invention.

[0029] Figure 2 This is a schematic diagram of the second partial structure of the UAV autonomous navigation remote sensing acquisition system of the present invention. DETAILED DESCRIPTION

[0030] Reference to an embodiment herein means that a particular feature, structure, or characteristic described in conjunction with the embodiment may be included in at least one embodiment of the present invention. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0031] An autonomous navigation remote sensing acquisition system for unmanned aerial vehicles, such as Figure 1As shown, it includes a mission planning module, a navigation and control module, a remote sensing data acquisition module, a communication monitoring module, a data processing module and an emergency support module. The mission planning module is used to determine the geographical range to be monitored according to user needs, and to set flight parameters synchronously. During autonomous flight, the surrounding factors are monitored in real time to optimize the flight path. The navigation and control module uses the GPS global satellite navigation system to obtain accurate location information, and combines the inertial measurement unit and auxiliary sensors to maintain stable flight. During flight, it relies on laser radar and ultrasonic sensors to perceive the surrounding environment. The remote sensing data acquisition module relies on the remote sensing carried by the drone itself to collect data, ensuring that the timestamp and geographic location information correspond to the collected data. The communication monitoring module is used to transmit the collected key data back to the ground station and receive manual operation instructions from the ground station at the same time. The data processing module is used to correct, splice, classify and other processes on the collected data, extract useful information, and generate maps, reports and other forms to intuitively display the results after the processing is completed, support decision making, and the emergency support module sets automatic return and landing functions to protect the safety of the equipment in the event of a failure.

[0032] like Figure 2 As shown in FIG. 1 , it also includes an intelligent dynamic task re-planning mechanism. The method adopted by the intelligent dynamic task re-planning mechanism is: using a reinforcement learning framework to allow drones to learn the optimal strategy by constantly trying different decisions, then introducing a multi-agent system, using multiple drones to work together, synchronously share information, optimize the overall task architecture, and perform spatiotemporal prediction and adjustment in this process, predict the changes in environmental conditions in the future, and realize secondary dynamic processing in the adjustment process. The intelligent dynamic task re-planning mechanism uses the following formula for calculation:

[0033] Q(s t , a t )←Q(s t , a t )+α[r t+1 +γmax a Q(s t+1 , a)-Q(s t , a t )],

[0034] Among them, s t is the state data at time step t, a t is the action data set that needs to be taken, r t+1 is the immediate reward, α is the learning rate, and γ is the discount factor;

[0035] After the calculation is completed, the loss function is introduced for definition and planning, so as to express the current network parameters and the target network parameters, and the policy gradient is updated after obtaining the data. The policy gradient update rule is:

[0036] where π i represents the strategy of the ith agent, Q π (s, a) is the value function of the joint action of all intelligent agents, that is, multiple drones conduct rapid collaborative cooperation to evaluate the overall performance of the collaboration between multiple drones, and continuously adjust the collaborative operation status between drone groups based on the obtained data, so as to make secondary dynamic adjustments under collaborative operation according to the predicted environmental status during the initial autonomous adjustment of the drones.

[0037] like Figure 2 As shown, it also includes an adaptive environmental perception and response mechanism, which uses situational awareness to schedule sensor combinations, extracts and classifies features of monitored environmental factors, improves the accuracy of obstacle recognition, and uses a Kalman filter to fuse different data within the sensor combination to reduce noise interference.

[0038] The following method is used to implement the adaptive environmental perception and response mechanism: the most appropriate sensor combination is selected through contextual awareness technology. Each drone dynamically decides which sensors should be activated to obtain the best data based on the current mission requirements, environmental characteristics and its own status. Sensor scheduling can be evaluated by the set utility function U(s, θ), where s is the environmental state and θ is the sensor configuration parameter. The specific formula is as follows:

[0039] U(s,θ)=∑ i w i f i (s,θ i ), where w i is the set weight, which needs to be dynamically evaluated according to the drone status. i is the performance index function of different sensors;

[0040] After determining the scheduling of the sensor, the command output of feature extraction and classification is sent to the sensor. The trained model is used to determine the input image and model parameters, and the category probability distribution calculation is performed on the new data obtained by the sensor to determine whether there are obstacles or other objects of interest.

[0041] Then, the system decides whether to take action based on the judgment result. If it takes action, it uses the following formula to calculate and obtain the average consensus of the drone group, update the environmental understanding and make corresponding dynamic adjustments. The formula is as follows:

[0042] Where i represents different references to drones, and the initial local estimate is N iis the set of drone swarms to which the mechanism is applied, and α is the set step size parameter;

[0043] When performing emergency obstacle avoidance, the decision rules are used to execute obstacle avoidance actions. The adjustment area range needs to be set between 0-1. When it is within this area, emergency obstacle avoidance will be performed, and in this process, dynamic adjustments will be made continuously based on environmental perception. When it is not within this area, it will fly normally.

[0044] like Figure 2 As shown, it also includes high-dimensional space path planning and risk avoidance mechanism. The high-dimensional space path planning and risk avoidance mechanism are based on the sampling method for path planning. During the calculation process, a virtual force field is created to guide the UAV to avoid dangerous areas. An evolutionary algorithm is used to search for the optimal solution, and a local improvement strategy is used to accelerate the convergence action.

[0045] The above embodiments are provided for persons familiar with the art to implement or use the present invention. Personnel familiar with the art can make various modifications or changes to the above embodiments without departing from the concept of the present invention. Therefore, the protection scope of the present invention is not limited to the above embodiments, but should be the maximum scope of the innovative features mentioned in the claims.

Claims

1. An unmanned aerial vehicle autonomous navigation remote sensing acquisition system, characterized in that: Included are: A mission planning module, which is used to determine the geographical scope to be monitored according to user needs, and simultaneously set flight parameters, and to monitor surrounding factors in real time during autonomous flight to optimize the flight path; A navigation and control module, which uses the GPS global satellite navigation system to obtain precise location information, and combines an inertial measurement unit and auxiliary sensors to maintain stable flight, and relies on laser radar and ultrasonic sensors to perceive the surrounding environment during flight; Remote sensing data acquisition module, which relies on the remote sensing carried by the drone to collect data, ensuring that the timestamp and geographic location information match the collected data; The communication monitoring module is used to transmit the collected key data back to the ground station and simultaneously receive manual operation instructions from the ground station.

2. An unmanned aerial vehicle autonomous navigation remote sensing acquisition system according to claim 1, characterized in that: It also includes an intelligent dynamic task re-planning mechanism, which adopts the following method: using a reinforcement learning framework to allow drones to learn the optimal strategy by continuously trying different decisions, then introducing a multi-agent system, using multiple drones to work together, synchronously share information, optimize the overall task architecture, and perform spatiotemporal prediction and adjustment in this process, predict changes in environmental conditions in the future, and achieve secondary dynamic processing during the adjustment process.

3. An unmanned aerial vehicle autonomous navigation remote sensing acquisition system according to claim 1, characterized in that: It also includes an adaptive environmental perception and response mechanism, which uses situational awareness to schedule sensor combinations, extracts and classifies features of monitored environmental factors, improves the accuracy of obstacle recognition, and uses a Kalman filter to fuse different data within the sensor combination to reduce noise interference.

4. The UAV autonomous navigation remote sensing acquisition system according to claim 1 is characterized in that: It also includes high-dimensional space path planning and risk avoidance mechanism. The high-dimensional space path planning and risk avoidance mechanism performs path planning based on the sampling method, creates a virtual force field during the calculation process to guide the drone to avoid dangerous areas, uses evolutionary algorithms to search for the optimal solution, and uses local improvement strategies to accelerate convergence.

5. The UAV autonomous navigation remote sensing acquisition system according to claim 1 is characterized in that: It also includes a data processing module, which is used to correct, splice, classify and other processes on the collected data, extract useful information, and generate maps, reports and other forms to intuitively display the results after the processing is completed to support decision-making.

6. The UAV autonomous navigation remote sensing acquisition system according to claim 1 is characterized in that: It also includes an emergency support module, which is equipped with functions such as automatic return and landing to protect the safety of the equipment in case of failure.

7. The UAV autonomous navigation remote sensing acquisition system according to claim 2 is characterized in that: The intelligent dynamic task replanning mechanism is calculated using the following formula: Q(s t , a t )←Q(s t , a t )+α[r t+1 +γmax a Q(s t+1 , a)-Q(s t , a t )), where s t is the state data at time step t, a t is the action data set that needs to be taken, r t+1 is the immediate reward, α is the learning rate, and γ is the discount factor; After the calculation is completed, the loss function is introduced for definition and planning, so as to express the current network parameters and the target network parameters, and the policy gradient update is performed after the data is obtained. The policy gradient update rule is: where π i represents the strategy of the i-th agent, Q π (s, a) is the value function of the joint action of all intelligent agents, that is, multiple drones conduct rapid collaborative cooperation to evaluate the overall performance of the collaboration between multiple drones, and continuously adjust the collaborative operation status between drone groups based on the obtained data, so as to make secondary dynamic adjustments under collaborative operation according to the predicted environmental status during the initial autonomous adjustment of the drones.

8. An unmanned aerial vehicle autonomous navigation remote sensing acquisition system according to claim 2, characterized in that: The adaptive environmental perception and response mechanism is implemented by the following method: the most suitable sensor combination is selected through contextual perception technology. Each UAV dynamically determines which sensors should be activated to obtain the best data according to the current mission requirements, environmental characteristics and its own state. Sensor scheduling can be evaluated by the set utility function U(s, θ), where s is the environmental state and θ is the sensor configuration parameter. The specific formula is as follows: U(s,θ)=∑ i w i f i (s,θ i ), where w i is the set weight, which needs to be dynamically evaluated according to the drone status. i is the performance index function of different sensors; After determining the scheduling of the sensor, the command output of feature extraction and classification is sent to the sensor. The trained model is used to determine the input image and model parameters, and the category probability distribution calculation is performed on the new data obtained by the sensor to determine whether there are obstacles or other objects of interest. Then, the system decides whether to take action based on the judgment result. If it takes action, it uses the following formula to calculate and obtain the average consensus of the drone group, update the environmental understanding and make corresponding dynamic adjustments. The formula is as follows: Where i represents different references to drones, and the initial local estimate is N i is the set of drone swarms to which the mechanism is applied, and α is the set step size parameter; When performing emergency obstacle avoidance, the decision rules are used to execute obstacle avoidance actions. The adjustment area range needs to be set between 0-1. When it is within this area, emergency obstacle avoidance will be performed, and in this process, dynamic adjustments will be made continuously based on environmental perception. When it is not within this area, it will fly normally.

Images