An agent-based autonomous control method for safe flight of unmanned aerial vehicles

Abstract

The present invention discloses an agent-based autonomous control method for safe flight of a UAV, which involves autonomous multi-source data fusion detection, autonomous environmental situation perception, autonomous path planning, and autonomous control strategy generation. When the UAV is in flight, it uses low-altitude intelligent networking technology to detect and perceive changes in the surrounding environment, including the spatiotemporal trajectories of other UAVs, obstacles, and wind shear. According to the starting and ending points of the mission, a reliable path and control strategy that avoids UAV conflicts, obstacle distribution, and wind shear are autonomously generated, thereby assisting the safe and efficient operation level of the low-altitude airspace.

Description

Technical Field

[0001] The present invention relates to the field of low-altitude airspace, and in particular to an intelligent agent-based autonomous control method for safe flight of a drone. Background Art

[0002] Low-altitude intelligent network refers to a digital network system built by integrating networking, digitalization and intelligent technologies in the airspace below 3,000 meters. With the rapid development of fifth-generation communication technology and drone technology, promoting the development of low-altitude networked drones is a major opportunity to build new infrastructure and develop the low-altitude economy. Obviously, low-altitude intelligent networking makes drone emergency rescue safer, more efficient, green and intelligent. Drone emergency rescue for low-altitude intelligent networking is a complex large system with dynamic and static coordinated operations of multiple subject resources such as flight environment perception, drone positioning and navigation, emergency rescue missions, air traffic networks, air traffic control and environment. The main core tasks include:

[0003] (1) Flight path planning. Low-altitude intelligent networking is used to achieve low-altitude linkage. Based on the distribution of obstacles, the risks of collision and falling between drones, drones and obstacles are considered. A feasible flight channel is sought for drones to visit adjacent demand points and the dispatch center, and the aircraft's flight range is minimized.

[0004] (2) Control strategy. Based on the low-altitude intelligent network dynamic tracking of the UAV trajectory, considering the performance of the UAV, such as turning and climbing, etc., on the basis of the established route design and trajectory planning, find a feasible control strategy, including speed, angle, etc., to ensure that the UAV can reach the key waypoints as planned.

[0005] It is in an uncertain environment, and its characteristics determine the complexity, randomness, suddenness, multi-objective and constraint, large-scale nature of scheduling. The scheduling plan compiled manually at the planning and real-time operation level is time-consuming and laborious, which only approximately reflects a specific state at a certain moment. It not only fails to give full play to the system efficiency, but also has a certain deviation from the actual operation scheduling. Therefore, with the help of advanced technologies such as the Internet of Things, big data and artificial intelligence, it is urgent to study the automatic generation technology of trajectory planning and control strategy collaborative planning driven by UAV emergency rescue tasks under low-altitude intelligent networking. According to the location of demand points and dispatch centers, considering the three-dimensional spatial distribution of obstacles, taking into account green, safety and efficiency, design rescue routes to traverse all demand points, and find the best executable trajectory and control strategy plan to transport emergency materials from the starting point to the destination in a timely manner.

[0006] At present, through studying and combing a large number of domestic and foreign literature, the existing theoretical research and practice of low-altitude intelligent networked unmanned aerial vehicle trajectory planning and control strategy collaborative planning mainly have two deficiencies: on the one hand, some studies establish mathematical models of the problem and construct solution algorithms from the perspective of operations optimization, which is insufficient in that they can only solve static, deterministic small-scale problems and it is difficult to add artificial experience to decision-making; on the other hand, existing studies all assume that the environment is determined to design the trajectory and control strategy of the unmanned aerial vehicle, and lack the relationship between environmental changes and dynamic unmanned aerial vehicle trajectories and control strategies. Therefore, the present invention carries out an agent-based unmanned aerial vehicle safe flight autonomous control method, involving autonomous multi-source data fusion detection, autonomous environmental situation perception, autonomous path planning, and autonomous control strategy generation. When the unmanned aerial vehicle is in flight, it uses low-altitude intelligent network technology to detect and perceive other unmanned aerial vehicles, obstacles, wind shear and other factors in its surrounding environment. According to the starting and ending points of the mission, it autonomously generates reliable paths and control strategies to avoid unmanned aerial vehicle conflicts, obstacle distribution, and wind shear, thereby assisting the safe and efficient operation level of low-altitude airspace. Summary of the invention

[0007] The present invention provides an agent-based autonomous control method for safe flight of unmanned aerial vehicles (UAVs), which involves autonomous multi-source data fusion detection, autonomous environmental situation perception, autonomous path planning, and autonomous control strategy generation. When the UAV is in flight, the low-altitude intelligent network connection technology is used to detect and perceive changes in the surrounding environment, including the spatiotemporal trajectories of other UAVs, obstacles, and wind shear. According to the starting and ending points of the mission, a reliable path and control strategy that avoids UAV conflicts, obstacle distribution, and wind shear are autonomously generated, thereby assisting the safe and efficient operation level of the low-altitude airspace.

[0008] The present application provides an agent-based autonomous control method for safe flight of a drone, comprising the following steps:

[0009] Step 1: Autonomous multi-source data fusion detection. Based on the current position of the UAV, a three-dimensional low-altitude intelligent network operation environment is generated and gridded. Each grid is abstracted into a global and local agent to detect and perceive other UAVs, obstacles, wind shear distribution and other safety risk events in its surrounding environment.

[0010] Step 2: Autonomous environmental situation awareness. Each grid overall and local agent combines historical and real-time data to dynamically perceive the future low-altitude intelligent network operation environment's safety spatiotemporal evolution trend.

[0011] Step 3: Autonomous path planning. The drone is abstracted as an agent, and the drone agent interacts with the overall and local agents of the grid to autonomously generate a reliable path that avoids drone conflicts, obstacle distribution, and wind shear.

[0012] Step 4: Autonomous control strategy generation. The drone agent autonomously generates a control strategy based on its performance data and reliable paths.

[0013] Furthermore, the step 1 performs autonomous multi-source data fusion detection, grids the three-dimensional low-altitude intelligent network operation environment, abstracts any grid into the grid as a whole and a local agent, detects and perceives other drones, obstacles and wind shear distribution in its surrounding environment, specifically including:

[0014] Step 1.1: Generate a three-dimensional low-altitude intelligent network operation environment and grid it according to the current position of the UAV, and abstract each grid into a grid overall agent.

[0015] Step 1.2: Each grid-wide agent builds grid-local agents to monitor safety risk events such as other UAVs, obstacles, and wind shear in the grid based on a variety of low-altitude intelligent networking technologies. Combined with their safety risk thresholds, the grid-local agents obtain the traffic status of any grid under different safety risk event categories at the current moment.

[0016] Step 1.3: The grid-wide agent comprehensively determines the traffic status of any grid at the current moment based on the total results of multiple grid-local agents.

[0017] Furthermore, in the step 2 of autonomous environmental situation awareness, each grid overall and local agent, combined with historical and real-time data, dynamically perceives the future safe spatiotemporal evolution trend of the low-altitude intelligent network operation environment, specifically including:

[0018] Step 2.1: Each local agent in each grid detects and perceives other drones, obstacles, wind shear and other information in its surrounding environment based on low-altitude intelligent network technology, combines historical and real-time data, takes the risk monitoring value of a certain security risk event in the current grid as output, and the risk monitoring value of a certain security risk event in several similar networks as input. At the same time, the traffic flow density, meteorological information, and air traffic control information of several similar networks are also used as input. Based on this, a spatiotemporal prediction model based on multiple inputs and single output is constructed to predict the risk monitoring value of a certain security risk event in the current grid in the future.

[0019] Step 2.2: Each grid overall Agent detects and perceives other drones, obstacles, wind shear and other information in its surrounding environment based on low-altitude intelligent network technology, combines historical and real-time data, takes the risk monitoring value of all safety risk events of the current grid as output, and the risk monitoring value of all safety risk events of several similar networks as input. At the same time, the traffic flow density, meteorological information, and air traffic control information of several similar networks are also used as input. Based on this, a spatiotemporal prediction model based on multi-input and multi-output is constructed to predict the comprehensive safety risk monitoring value of the current grid in the future.

[0020] Step 2.3: Combine their security risk thresholds to predict the traffic status of all grids under different security risk event categories at different times in the future, as well as the comprehensive traffic status.

[0021] Furthermore, in step 3, autonomous path planning, the UAV is abstracted as an agent, and the UAV agent interacts with the grid overall and local agents to autonomously generate a reliable path that avoids UAV conflicts, obstacle distribution, and wind shear, specifically including:

[0022] Step 3.1: The UAV Agent interacts with the grid-based and local Agents of its mission start and end points, finds a virtual connection line based on the start and end points, and determines the set of grid-based and local Agents along the virtual connection line.

[0023] Step 3.2: In the above set, check whether the traffic status of any two adjacent grids and local agents on the virtual connection line is prohibited from flying. If they are not prohibited from flying, connect them as part of the path of the drone agent.

[0024] Step 3.3: If the path of the drone agent is incomplete, for the prohibited flight grid and local agent, check the adjacent non-prohibited flight grid and local agent, add them to the drone agent's path, and register them in the routing table.

[0025] Step 3.4: In the routing table, determine the best grid and local agent for non-prohibited flight according to the best path preference of the drone agent, replace the original grid and local agent, and update the routing table. If the best grid and local agent for non-prohibited flight cannot be found in the routing table, go to step 3.3, use different strategies, and regenerate the routing table registration.

[0026] Step 3.5: If the path of the drone agent is complete, calculate the objective function and output the result; otherwise, go to step 3.3.

[0027] Furthermore, in step 4, autonomous control strategy generation, the drone agent autonomously generates a control strategy based on its performance data and reliable paths, specifically including:

[0028] Step 4.1: The UAV Agent interacts with the relevant grid global and local agents in the reliable path, and collects information such as their dynamic limited time window, the distance, direction and position difference of the interaction between the adjacent grid global and local agents.

[0029] Step 4.2: The UAV Agent considers the performance of the UAV, such as the maximum and minimum speed, turning angle, and elevation angle range, based on the dynamic limited time window of the interaction between the grid as a whole and the local agent, and establishes a mathematical model of the speed, turning angle, elevation angle, hovering and other optimization decision variables of the execution-related grid as a whole and the local agent flying in a reliable path. The optimization objectives are total cost, control smoothness, and handling stability.

[0030] Step 4.3: The UAV Agent uses Lagrangian relaxation technology to add the relevant constraints of the mathematical model as penalty terms into its objective function, and determines the existence of the optimal control strategy by taking the first-order and second-order derivatives of the objective function.

[0031] Step 4.4: If the best control strategy is found, output the speed, turning angle, elevation angle, hovering time, etc. of any grid and local agent in the reliable path of the UAV; otherwise, output the recommended maximum and minimum speed, turning angle, and elevation angle range, and go to step 4.2.

[0032] Beneficial effects achieved by the present invention:

[0033] The present invention integrates the functions of autonomous multi-source data fusion detection, autonomous environmental situation perception, autonomous path planning, and autonomous control strategy generation for safe flight control of unmanned aerial vehicles based on intelligent agents. When the unmanned aerial vehicle is in flight, it uses low-altitude intelligent networking technology to detect and perceive changes in the surrounding environment, and autonomously generates reliable paths and control strategies that avoid unmanned aerial vehicle conflicts, obstacle distribution, and wind shear according to the starting and ending points of the mission, thereby assisting the safe and efficient operation level of low-altitude airspace. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 The present invention is a flow chart of an autonomous control method for safe flight of a UAV based on an intelligent agent.

[0035] Figure 2 It is a schematic diagram of the autonomous multi-source data fusion detection process of the present invention.

[0036] Figure 3 This is a schematic diagram of the autonomous environmental situation awareness process proposed by the present invention.

[0037] Figure 4 This is a schematic diagram of the autonomous path planning process proposed by the present invention.

[0038] Figure 5 A schematic diagram of a flow chart for generating the autonomous control strategy proposed in the present invention. DETAILED DESCRIPTION

[0039] The technical solution of the present invention is described in detail below in conjunction with the specific drawings.

[0040] like Figure 1 As shown, this embodiment provides an agent-based autonomous control method for safe flight of unmanned aerial vehicles, which involves autonomous multi-source data fusion detection, autonomous environmental situation awareness, autonomous path planning, and autonomous control strategy generation. When the unmanned aerial vehicle is in flight, it uses low-altitude intelligent network technology to detect and perceive changes in the surrounding environment, including the spatiotemporal trajectories of other unmanned aerial vehicles, obstacles, and wind shear. According to the starting and ending points of the mission, it autonomously generates reliable paths and control strategies to avoid unmanned aerial vehicle conflicts, obstacle distribution, and wind shear, thereby assisting the safe and efficient operation level of low-altitude airspace. Specifically, it includes the following steps:

[0041] S1: Autonomous multi-source data fusion detection. Based on the current position of the UAV, a three-dimensional low-altitude intelligent network operation environment is generated and gridded, and each grid is abstracted into a global and local agent to detect and perceive other UAVs, obstacles, wind shear distribution and other safety risk events in its surrounding environment.

[0042] S 2: Autonomous environmental situation awareness. Each grid overall and local agent combines historical and real-time data to dynamically perceive the safety spatiotemporal evolution trend of the future low-altitude intelligent network operation environment.

[0043] S3: Autonomous path planning. The drone is abstracted as an agent, and the drone agent interacts with the grid overall and local agents to autonomously generate a reliable path that avoids drone conflicts, obstacle distribution, and wind shear.

[0044] like Figure 2 As shown, the autonomous multi-source data fusion detection process of the present invention includes:

[0045] S1.1: Generate a three-dimensional low-altitude intelligent network operation environment and grid it according to the current position of the UAV.

[0046] S1.2: Use low-altitude intelligent network technology to monitor other drones, obstacles, wind shear and other safety risk events in the grid.

[0047] S1.3: Each grid is abstracted into a grid-wide agent and a local agent, where the grid-wide agent is responsible for building multiple local agents based on different low-altitude intelligent network connection technologies.

[0048] S1.4: Combined with their safety risk thresholds, the local agent of any grid obtains the traffic status of any grid under different safety risk event categories (other drones, obstacles and wind shear) at the current moment.

[0049] S1.5: The grid-wide agent comprehensively determines the traffic status of any grid at the current moment based on the total results of multiple grid-local agents.

[0050] like Figure 3 As shown, the autonomous environmental situation awareness process proposed by the present invention specifically includes:

[0051] S2.1: Detect and perceive other drones, obstacles, wind shear and other information in the surrounding environment based on low-altitude intelligent network technology, and collect historical and real-time data on different safety risk events in each grid

[0052] S2.2: Each local agent in each grid takes the risk monitoring value of a security risk event of the current grid as output and the risk monitoring value of a security risk event of several similar networks as input. At the same time, it also takes the traffic flow density, meteorological information, and air traffic control information of several similar networks as input. Based on this, a spatiotemporal prediction model based on multiple inputs and single output is constructed to predict the risk monitoring value of a security risk event of the current grid in the future.

[0053] S2.3: Each grid overall Agent takes the risk monitoring value of all safety risk events of the current grid as output and the risk monitoring value of all safety risk events of several similar networks as input. At the same time, it also takes the traffic flow density, meteorological information, and air traffic control information of several similar networks as input. Based on this, a spatiotemporal prediction model based on multi-input and multi-output is constructed to predict the comprehensive safety risk monitoring value of the current grid in the future.

[0054] S2.4: Combined with their safety risk thresholds, predict the traffic status of all grids under different safety risk event categories at different times in the future, as well as the comprehensive traffic status.

[0055] like Figure 4 As shown, the autonomous path planning process proposed by the present invention specifically includes:

[0056] S3.1: The UAV Agent interacts with the grid-based and local Agents of its mission start and end points, finds a virtual connection line based on the start and end points, and determines the set of grid-based and local Agents along the virtual connection line.

[0057] S3.2: In the above set, check whether the traffic status of any two adjacent grids on the virtual connection line is prohibited from flying. If they are not prohibited from flying, connect them as part of the path of the drone agent.

[0058] S3.3: If the path of the drone agent is incomplete, for the prohibited flight grid and local agent, check the adjacent non-prohibited flight grid and local agent, add them to the path of the drone agent, and register them in the routing table.

[0059] S3.4: In the routing table, according to the best path preference of the drone agent, determine the best grid overall and local agent for non-prohibited flight, replace the original grid overall and local agent, and update the routing table. If the best grid overall and local agent for non-prohibited flight cannot be found in the routing table, go to step S3.3, adopt different strategies, and regenerate the routing table registration.

[0060] S3.5: If the path of the drone agent is complete, calculate the objective function and output the result; otherwise, go to step S3.3.

[0061] like Figure 5 As shown, the autonomous control strategy generation process proposed by the present invention specifically includes:

[0062] S4.1: UAV Agent reads the path.

[0063] S4.2: The UAV Agent interacts with the relevant grid global and local agents in the reliable path, and collects information such as their dynamic limited time window, the distance, direction and position difference of the interaction between the adjacent grid global and local agents.

[0064] S4.3: The UAV Agent considers the performance of the UAV, such as the maximum and minimum speed, turning angle, and elevation angle range, based on the dynamic limited time window of the interaction between the grid as a whole and the local agent, and establishes a mathematical model of the speed, turning angle, elevation angle, hovering and other optimization decision variables of the execution-related grid as a whole and the local agent in a reliable path. The optimization objectives are total cost, control smoothness, and handling stability.

[0065] S4.4: The UAV Agent uses Lagrangian relaxation technology to add the relevant constraints of the mathematical model as penalty terms into its objective function, and determines the existence of the optimal control strategy by taking the first-order and second-order derivatives of the objective function.

[0066] S4.5: If the best control strategy is found, output the speed, turning angle, elevation angle, hovering time, etc. of any grid and local agent in the reliable path of the UAV; otherwise, output the recommended maximum and minimum speed, turning angle, and elevation angle range, and go to step S4.3.

[0067] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process or method including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the existence of other identical elements in the process or method including the element.

[0068] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. An agent-based autonomous control method for safe flight of unmanned aerial vehicles, characterized in that: The following steps are involved: Step 1: Autonomous multi-source data fusion detection: Based on the current position of the UAV, a three-dimensional low-altitude intelligent network operation environment is generated and gridded, and each grid is abstracted into a global and local agent to detect and perceive safety risk events in its surrounding environment, where the safety risk events include other UAVs, obstacles, and wind shear distribution; Step 2: Autonomous environmental situation awareness: Each grid’s overall and local agents combine historical and real-time data to dynamically perceive the future safe spatiotemporal evolution trend of the low-altitude intelligent network operation environment; Step 3: Autonomous path planning: Abstract the drone as an agent, and use the drone agent to interact with the grid as a whole and local agents to autonomously generate a reliable path that avoids drone conflicts, obstacle distribution, and wind shear; Step 4: Autonomous control strategy generation: The drone agent autonomously generates a control strategy based on its performance data and reliable paths; In step 1, the autonomous multi-source data fusion detection includes the following steps: Step 1.1: Generate a three-dimensional low-altitude intelligent network operation environment based on the current position of the UAV and grid it, abstracting each grid into a grid overall agent; Step 1.2: Each grid overall agent builds a grid local agent to monitor grid safety risk events based on a variety of different low-altitude intelligent network connection technologies. Combined with their safety risk thresholds, the grid local agent obtains the traffic status of any grid under different safety risk event categories at the current moment; Step 1.3: The grid-wide agent comprehensively determines the traffic status of any grid at the current moment based on the total results of multiple grid-local agents; In step 3, the autonomous path planning includes the following steps: Step 3.1: The UAV Agent interacts with the grid-based and local Agents of its mission start and end points, finds a virtual connection line based on the start and end points, and determines a number of grid-based and local Agent sets along the virtual connection line. Step 3.2: among the several grid whole and local agent sets along the virtual link, check whether the traffic status of any two adjacent grid whole and local agents on the virtual link is prohibited from flying. If both are not prohibited from flying, connect them as part of the path of the drone agent; Step 3.3: If the path of the drone agent is incomplete, for the prohibited flight grid and local agent, check the adjacent non-prohibited flight grid and local agent, add them to the drone agent's path, and register them in the routing table; Step 3.4: In the routing table, determine the best grid overall and local agent for non-prohibited flight according to the best path preference of the drone agent, replace the original grid overall and local agent, and update the routing table; if the best grid overall and local agent for non-prohibited flight cannot be found in the routing table, go to step 3.3, adopt different strategies, and regenerate the routing table registration; Step 3.5: If the path of the drone agent is complete, calculate the objective function and output the result; otherwise, go to step 3.

3.

2. The method for autonomous control of safe flight of unmanned aerial vehicle based on intelligent agent according to claim 1, characterized in that: In step 2, the autonomous environmental situation awareness includes the following steps: Step 2.1: Each local agent in each grid detects and perceives the safety risk events in its surrounding environment based on the low-altitude intelligent network technology. Combining historical and real-time data, it takes the risk monitoring value of a certain safety risk event in the current grid as output and the risk monitoring value of a certain safety risk event in several similar networks as input. At the same time, it also takes the traffic flow density, meteorological information, and air traffic control information of several similar networks as input. Based on this, a spatiotemporal prediction model based on multiple inputs and single outputs is constructed to predict the risk monitoring value of a certain safety risk event in the current grid in the future. Step 2.2: Each grid overall agent detects and perceives the safety risk events in its surrounding environment based on the low-altitude intelligent network technology, combines historical and real-time data, takes the risk monitoring value of all safety risk events in the current grid as output, and the risk monitoring value of all safety risk events in several similar networks as input. At the same time, the traffic flow density, meteorological information, and air traffic control information of several similar networks are also used as input. Based on this, a spatiotemporal prediction model based on multi-input and multi-output is constructed to predict the comprehensive safety risk monitoring value of the current grid in the future; Step 2.3: Combine their security risk thresholds to predict the traffic status of all grids under different security risk event categories at different times in the future, as well as the comprehensive traffic status.

3. The method for autonomous control of safe flight of unmanned aerial vehicle based on intelligent agent according to claim 1, characterized in that: In step 4, the autonomous control strategy generation includes the following steps: Step 4.1: The UAV Agent interacts with the relevant grid global and local agents in the reliable path, and collects the distance, direction and position difference of the interaction between the dynamic limited time window and the adjacent grid global and local agents; Step 4.2: The UAV Agent considers the performance of the UAV, including the maximum and minimum speed, turning angle and elevation angle range, based on the dynamic limited time window of the interaction between the grid as a whole and the local agent, and establishes a mathematical model of the optimization decision variables of the execution-related grid as a whole and the local agent flying in a reliable path. The optimization decision variables include speed, turning angle, elevation angle and hovering, and the optimization objectives are total cost, control smoothness and handling stability; Step 4.3: The UAV Agent uses Lagrangian relaxation technology to add the relevant constraints of the mathematical model as penalty terms into its objective function, and determines the existence of the optimal control strategy by taking the first-order and second-order derivatives of the objective function; Step 4.4: If the optimal control strategy is found, output the speed, turning angle, elevation angle and hovering time of any grid and local agent in the reliable path of the UAV; otherwise, output the recommended maximum and minimum speed, turning angle, and elevation angle ranges, and go to step 4.2.

Images