Unmanned aerial vehicle landing method and device

Abstract

The invention provides an unmanned aerial vehicle landing method and device, relates to the field of unmanned aerial vehicles, and can realize accurate landing of the unmanned aerial vehicle. The method comprises the steps that parameter information of the unmanned aerial vehicle, rotation information of a plurality of propellers of the unmanned aerial vehicle and position information of a plurality of landing point positions in the flight process of the unmanned aerial vehicle are obtained, the parameter information of the unmanned aerial vehicle comprises space position information of the propellers, and the rotation information comprises rotation directions and rotation speeds of the propellers; one landing point corresponds to one propeller; determining a landing strategy of the unmanned aerial vehicle based on the parameter information of the unmanned aerial vehicle, the rotation information of the plurality of propellers of the unmanned aerial vehicle and the position information of the plurality of landing point positions; the landing strategy is used for adjusting the action posture of the unmanned aerial vehicle in the landing process; first landing indication information is sent to the unmanned aerial vehicle, and the first landing indication information is used for indicating the landing strategy, so that the unmanned aerial vehicle uses the landing strategy to land in the landing process.

Description

Technical Field

[0001] The present application relates to the technical field of unmanned aerial vehicles, and in particular to a method and device for landing an unmanned aerial vehicle. Background Art

[0002] With the rapid development of drone technology, drones are used in various occasions. For example, in agricultural applications, drones are used to spray pesticides; in filming applications, drones are used for aerial photography; in firefighting applications, drones are used for firefighting, etc.

[0003] Current drones require a strong contrast reference object to use image recognition technology. The reference object must have multiple levels of patterns, otherwise a single-color drone cannot find a landing location, and the pattern cannot be symmetrical. If it is symmetrical, the direction cannot be identified. When the lighting conditions at the target landing location are poor, or the target reference object is not clear, or there are other things on the drone that mistakenly block the recognition system, the drone's landing position will be inaccurate. Summary of the invention

[0004] The present application provides a method and device for landing a drone, which can achieve accurate landing of the drone.

[0005] In order to achieve the above objectives, this application adopts the following technical solutions:

[0006] In a first aspect, the present application provides a method for landing a drone, the method comprising: obtaining parameter information of the drone, rotation information of multiple propellers of the drone, and position information of multiple landing points during the flight of the drone, the parameter information of the drone including the spatial position information of the propeller, the rotation information including the rotation direction and rotation speed of the propeller, and one landing point corresponding to one propeller; determining a landing strategy of the drone based on the parameter information of the drone, the rotation information of multiple propellers of the drone, and the position information of the multiple landing points; the landing strategy is used to adjust the movement posture of the drone during the landing process; and sending first landing indication information to the drone, the first landing indication information being used to indicate the landing strategy, so that the drone uses the landing strategy to land during the landing process.

[0007] Based on the method provided in the first aspect above, the landing point of the drone is not determined through image recognition, but through the precise positioning of a base to achieve automatic parking of the drone, without the need for reference objects; the landing location is determined based on the positioning of multiple propeller positions of the drone and the corresponding landing points of the base, and multi-point positioning has higher accuracy.

[0008] In a possible implementation manner, the parameter information of the UAV also includes the flight direction and flight speed of the UAV.

[0009] Based on the above method, the precise landing of the drone can be achieved through the flight direction and speed of the drone, the positioning of the propeller and the positioning of the base, with multi-point positioning and higher accuracy.

[0010] A possible implementation method is to determine the landing strategy of the UAV based on the parameter information of the UAV, the rotation information of multiple propellers of the UAV, and the position information of the multiple landing points, including: determining multiple landing methods of the UAV based on the parameter information of the UAV, the rotation information of multiple propellers of the UAV, and the position information of the multiple landing points; and selecting a target landing method from the multiple landing methods as the landing strategy of the UAV.

[0011] Based on the above method, a suitable landing method can be selected from a variety of landing methods to improve landing accuracy.

[0012] A possible implementation method, determining multiple landing methods of the drone based on the parameter information of the drone, the rotation information of multiple propellers of the drone, and the position information of the multiple landing points, includes: obtaining the parameter information of the drone at a second moment based on the parameter information of the drone at a first moment and the rotation information of the multiple propellers of the drone; determining multiple rotation information of the multiple propellers of the drone at the second moment based on the parameter information of the drone at the second moment and the position information of the multiple landing points; determining multiple landing methods of the drone based on the multiple rotation information of the multiple propellers of the drone at the second moment, one rotation information corresponding to one landing method; the second moment is a moment after the first moment.

[0013] A possible implementation method, wherein the target landing method is selected from the multiple landing methods as the landing strategy of the UAV, including: obtaining a state value function and an action value function corresponding to each landing method; the state value function is used to determine the value of parameter information, and the action value function is used to determine the value of rotation information; for each landing method, determining the parameter information and rotation information corresponding to the landing method when the state value function and the action value function corresponding to the landing method take maximum values ​​respectively; based on the parameter information and rotation information corresponding to each landing reflection method, selecting the target landing method from the multiple landing methods.

[0014] A possible implementation method is to determine the landing strategy of the drone based on the parameter information of the drone, the rotation information of multiple propellers of the drone, and the position information of the multiple landing points, including: inputting the parameter information of the drone, the rotation information of multiple propellers of the drone, and the position information of the multiple landing points into a preset deep Q network model to obtain the landing strategy of the drone.

[0015] Based on the above method, artificial intelligence can be used to learn algorithms and train the safest and most accurate landing model, which can then be used to meet the needs of automatic landing, ensuring the safe landing of drones in different environments.

[0016] In a second aspect, the present application provides a drone landing device for implementing the above method. The drone landing device includes a module, unit, or means corresponding to the above method, and the module, unit, or means can be implemented by hardware, software, or by hardware executing the corresponding software. The hardware or software includes one or more modules or units corresponding to the above functions.

[0017] In a third aspect, the present application provides a device for landing a drone, the device comprising: a processor and a communication interface; the communication interface and the processor are coupled, and the processor is used to run a computer program or instructions to implement a method for landing a drone as described in the first aspect and any possible implementation of the first aspect.

[0018] In a fourth aspect, the present application provides a computer-readable storage medium, which stores instructions. When the instructions are executed on a terminal, the terminal executes the method for landing a drone as described in the first aspect and any possible implementation of the first aspect.

[0019] In a fifth aspect, an embodiment of the present application provides a computer program product comprising instructions. When the computer program product is run on a drone landing device, the drone landing device executes the drone landing method as described in the first aspect and any possible implementation of the first aspect.

[0020] In a sixth aspect, an embodiment of the present application provides a chip, the chip comprising a processor and a communication interface, the communication interface and the processor are coupled, and the processor is used to run a computer program or instructions to implement a method for landing a drone as described in the first aspect and any possible implementation of the first aspect.

[0021] Specifically, the chip provided in the embodiment of the present application also includes a memory for storing computer programs or instructions. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is a schematic diagram of a drone landing control system provided in an embodiment of the present application;

[0023] Figure 2 is a flow chart of a method for landing a drone provided in an embodiment of the present application;

[0024] Figure 3 is a flow chart of another method for landing a drone provided in an embodiment of the present application;

[0025] Figure 4 This is a schematic diagram of environmental feedback during the landing process of a drone provided in an embodiment of the present application;

[0026] Figure 5 is a schematic diagram of a drone landing strategy provided by an embodiment of the present application;

[0027] Figure 6 is a flow chart of another method for landing a drone provided in an embodiment of the present application;

[0028] Figure 7 It is a structural schematic diagram of a drone landing device provided in an embodiment of the present application;

[0029] Figure 8 This is a schematic diagram of the hardware structure of a drone landing device provided in an embodiment of the present application;

[0030] Fig. 9 It is a schematic diagram of the structure of a chip provided in an embodiment of the present application. DETAILED DESCRIPTION

[0031] The method and device for monitoring sensitive data provided by the embodiments of the present application are described in detail below in conjunction with the accompanying drawings.

[0032] The term "and / or" in this article is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone.

[0033] The terms "first" and "second" and the like in the specification and drawings of this application are used to distinguish different objects, or to distinguish different processing of the same object, rather than to describe a specific order of objects.

[0034] In addition, the terms "including" and "having" and any variations thereof mentioned in the description of the present application are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device comprising a series of steps or units is not limited to the listed steps or units, but may optionally include other steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or devices.

[0035] It should be noted that, in the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0036] The implementation of the embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0037] With the rapid development of modern technology, the application of artificial intelligence and 5G technology has become the mainstream of the times. This application is mainly based on deep learning algorithms. It is a series of algorithms that simulate the human brain to process data and make corresponding decisions based on its data. It is also a new research direction in the field of machine learning. In addition, deep learning is based on artificial neural networks, so the main purpose of studying it is to establish a neural network system that simulates human brain analysis and learning, so as to simulate the thinking mode of the human brain to interpret data such as images, sounds, and texts. The current deep learning network model can understand and recognize general natural images. The deep learning model not only greatly improves the accuracy of image recognition, but also avoids the need to consume a lot of time for artificial feature extraction, so that the online operation efficiency is improved. As described in the background technology, the prior art uses deep learning algorithms to realize automatic landing of drones mainly through image recognition. It is a new technology based on image recognition technology. A strong contrast reference is set at the landing point of the drone, and it can contain a variety of information together and has a strong contrast reference. Only in this way can the relative position of the target landing point and the drone be obtained, thereby realizing automatic landing. Image recognition technology requires a strong contrast reference object with multi-level patterns. Otherwise, a single-color drone cannot find a landing location, and the pattern cannot be symmetrical, and the direction cannot be distinguished if it is symmetrical. When the lighting conditions at the target landing location are poor, or the target reference object is not clear, or there are other things on the drone that mistakenly block the recognition system, it cannot land automatically.

[0038] In order to solve the above problems, an embodiment of the present application provides a method for landing a drone, the method comprising: obtaining parameter information of the drone during flight, rotation information of multiple propellers of the drone, and position information of multiple landing points, the parameter information of the drone including spatial position information of the propeller, the rotation information including the rotation direction and rotation speed of the propeller, and one landing point corresponding to one propeller; based on the parameter information of the drone, the rotation information of multiple propellers of the drone, and the position information of multiple landing points, determining the landing strategy of the drone; the landing strategy is used to adjust the movement posture of the drone during the landing process; sending first landing indication information to the drone, the first landing indication information is used to indicate the landing strategy, so that the drone uses the landing strategy to land during the landing process.

[0039] It is understandable that the above method can be performed by any device with computing capabilities. For example, it can be performed by a server, a computer or a computing device (hereinafter referred to as a device for landing a drone). Among them, the server includes but is not limited to: a tower server, a blade server, a rack server, a physical server, a virtual host, a virtual private server (Virtual Private Server, VPS), a cloud server, a home server, an enterprise server, etc.

[0040] See also Figure 1 , Figure 1 Schematic diagram of a drone landing control system provided in an embodiment of the present application. The drone landing control system includes a drone 1 and a controller 2. The drone 1 includes four propellers. The controller 2 is used to control the rotation of the propellers to control the flight of the drone.

[0041] In one scenario, a base 3 is provided on the drone landing platform, and the base 3 is used to carry the drone after landing. The base 3 is provided with four drone landing points, and each landing point corresponds to a propeller. After the drone 1 lands, the propeller of the drone 1 will land at the corresponding landing point.

[0042] The following is a detailed description of the method for landing a drone provided in an embodiment of the present application in conjunction with the accompanying drawings.

[0043] like Figure 2 As shown, Figure 2 : is a flow chart of a method for landing a drone provided in an embodiment of the present application, and the method for landing a drone comprises the following steps:

[0044] S201: Obtaining parameter information of the drone, rotation information of multiple propellers of the drone, and location information of multiple landing points during the flight of the drone.

[0045] The drone includes four propellers. The parameter information of the drone includes spatial position information of the propellers. The spatial position information of the propellers represents the position information of the drone.

[0046] Optionally, the spatial position information of the propeller is the spatial position coordinates of the propeller.

[0047] In a possible implementation, the controller obtains spatial position information of the propellers of the drone during flight, rotation information of multiple propellers of the drone, and position information of multiple landing points of the base.

[0048] Exemplarily, the spatial position coordinates of the four propellers are (x1, y1, z1), (x2, y2, z2), (x3, y3, z3), and (x4, y4, z4), where x and y are the horizontal and vertical coordinates of the propeller, and z is the height of the propeller. The height of the drone can be obtained, and the height of the drone can be used as the height of the four propellers, that is, z1, z2, z3, and z4 are all the same as the height of the drone. The height of each propeller can also be obtained separately.

[0049] In one implementation, the parameter information of the drone also includes the rotation state of each propeller in the multiple propellers, and the rotation information of each propeller includes the rotation direction and rotation speed of the propeller. The rotation direction of the propeller includes four directions: east, south, west, and north. The controller obtains the spatial position information of the propellers of the drone during flight, the rotation state of each propeller in the multiple propellers, the rotation information of the multiple propellers of the drone, and the position information of the multiple landing points of the base.

[0050] In another implementation, the parameter information of the drone also includes the flight speed and flight direction of the drone. The controller obtains the spatial position information of the propellers of the drone during flight, the rotation state of each of the multiple propellers, the rotation information of the multiple propellers of the drone, the flight speed and flight direction of the drone, and the position information of the multiple landing points of the base.

[0051] The drone landing platform is provided with a base, and four landing points are set on the base according to the structure of the four propellers of the drone. The position information of multiple landing points of the base is obtained, and one landing point corresponds to one propeller. The position information of multiple landing points is a spatial coordinate, which is used to land the drone at a specified position. Exemplarily, the spatial position coordinates of the four landing points are (X1, Y1, Z1), (X2, Y2, Z2), (X3, Y3, Z3), and (X4, Y4, Z4). Among them, X and Y are the horizontal coordinates and vertical coordinates of the landing point in the horizontal direction, and Z is the height of the landing point. When the base of the drone landing platform is flat, the spatial position information of the four landing points is set to Z1=Z2=Z3=Z4=0. When the base of the drone landing platform is uneven and has an inclination angle, the three-dimensional coordinates of the four landing points are obtained respectively.

[0052] S202: Determine a landing strategy of the drone based on the parameter information of the drone, the rotation information of multiple propellers of the drone, and the position information of multiple landing points.

[0053] The landing strategy of the drone is a strategy for each propeller of the drone to land at the corresponding point. The landing strategy is used to adjust the action posture of the drone during the landing process. The parameter information of the drone can also be called the state information S of the drone, and the rotation information of multiple propellers of the drone can also be called the action information A of the drone.

[0054] The state information S includes the spatial position information of the multiple propellers of the drone, the flight direction and the flight speed of the drone, and the action information A includes the rotation information of each of the multiple propellers of the drone, wherein the rotation information includes the rotation direction and the rotation speed of the propeller. Under the spatial position information of the multiple propellers of the drone at the current moment, the multiple propellers rotate according to the rotation information of each propeller, then the spatial position information of the multiple propellers of the drone at the next moment can be determined.

[0055] For example, the state of the drone at time t is St, and the action of the propeller is at, so that the state of the drone at time t+1 is St+1.

[0056] In one implementation, the best landing strategy is selected from multiple landing methods by constructing an action value function and a state value function of the drone flight. The state value function is used to characterize the value of the drone parameter information to the drone landing, and the action value function is used to characterize the value of the rotation information of the drone's multiple propellers to the drone landing. Figure 3 The method shown.

[0057] In another implementation, the parameter information of the drone, the rotation information of the multiple propellers of the drone, and the location information of the multiple landing points are input into a preset Deep Q-Network (DQN) model to obtain the landing strategy of the drone. Figure 4 The method shown.

[0058] S203: Send a first landing instruction to the drone, so that the drone uses a landing strategy to land during the landing process.

[0059] The first landing instruction is used for the landing strategy. The network device sends the optimal state and the optimal action in the calculated optimal strategy as the first landing instruction to the drone.

[0060] exist Figure 2 In the embodiment shown, the landing point of the drone is not determined by image recognition, but by the precise positioning of a base to achieve automatic parking of the drone, without the need for reference objects; the landing location is determined based on the positioning of multiple propeller positions of the drone and the corresponding landing points of the base, and multi-point positioning has higher accuracy.

[0061] like Figure 3 As shown, Figure 3 : is a flow chart of another method for landing a drone provided in an embodiment of the present application, the method for landing a drone comprises the following steps:

[0062] S301: Determine multiple landing modes of the drone based on parameter information of the drone, rotation information of multiple propellers of the drone, and position information of multiple landing points.

[0063] It should be noted that the parameter information of the drone can be called the drone state, the rotation information of the propeller can be called the drone action, and the location information of multiple landing points of the base can also be called the drone landing environment.

[0064] Based on the parameter information of the drone at the first moment and the rotation information of the multiple propellers of the drone, the parameter information of the drone at the second moment is obtained. Based on the parameter information of the drone at the second moment and the position information of multiple landing points, the multiple rotation information of the multiple propellers of the drone at the second moment is determined; based on the multiple rotation information of the multiple propellers of the drone at the second moment, the multiple landing methods of the drone are determined, one rotation information corresponds to one landing method, and the second moment is the moment after the first moment. The state value function and action value function corresponding to each landing method are obtained; the state value function is used to determine the value of the parameter information, and the action value function is used to determine the value of the rotation information. For each landing method, the parameter information and rotation information corresponding to the landing method are determined when the state value function and action value function corresponding to the landing method are respectively maximized; based on the parameter information and rotation information corresponding to each landing reflection method, the target landing method is selected from multiple landing methods.

[0065] For example, see Figure 4 , Figure 4 It is a schematic diagram of environmental feedback during the landing process of a drone provided in an embodiment of the present application. The drone performs action a, and according to the base environment, the drone obtains the state s at the next moment, and the environmental feedback is r. Optionally, based on the Bellman equation, the state of the drone at time t is St, and the propeller performs an action at. According to the coordinate information of multiple landing points of the base, the reward Rt+1 of the drone in state St and action at is obtained. This affects the reward Rt+2, Rt+3, Rt+4, etc. at subsequent moments. Among them, the reward is used to characterize the value of the drone's flight state and action to precise landing. The higher the reward, the higher the value of the flight state and flight action, and the more effective it is for precise landing.

[0066] In order to land on the base, the propeller of the drone may make multiple actions at time t+1. One action corresponds to a state of the drone at time t+2. Similarly, multiple possible paths for the drone to fly to the base station can be obtained.

[0067] For example, see Figure 5 , Figure 5This is a schematic diagram of a drone landing strategy provided by an embodiment of the present application. At time t, the drone's state is s and its action is a. Therefore, at time t+1, the drone's state becomes s1. In order to reach the base, the drone may perform a variety of actions a1, a2, a3, etc. When the action performed by the drone at time t+1 is a1, the drone's state at time t+2 is s21. When the action performed by the drone at time t+1 is a2, the drone's state at time t+2 is s22. And so on, multiple states of the drone at time t+2 are obtained. Similarly, when the drone's state at time t+2 is s21, in order to reach the base, the drone may perform a variety of actions a1, a2, a3, etc. Multiple states of the drone at time t+3 are obtained, such as s31, s32, s33, etc. It is urgent to obtain the drone's state and drone action at later moments. As Figure 3 As shown, each flight path of the UAV is regarded as a landing strategy of the UAV, and multiple landing strategies are obtained using the above method.

[0068] Exemplarily, the state of the drone at the first moment is state 1, and the action is action 1. It is estimated that the state of the drone at the second moment is state 2, and the possible actions include action 21 and action 22. The corresponding drone state at the third moment is state 31 and state 32. When the drone state is state 31, the possible actions include action 311 and action 322. When the drone state is state 32, the possible actions include action 321. At this time, three drone flight strategies are obtained, each strategy includes a combination of multiple drone states and drone actions. Strategy 1 is: (state 1, action 1) (state 2, action 21) (state 31, action 311); Strategy 2 is: (state 1, action 1) (state 2, action 21) (state 31, action 312); Strategy 3 is: (state 1, action 1) (state 2, action 22) (state 32, action 321).

[0069] S302: Obtain the state value function and action value function corresponding to each landing method.

[0070] It should be noted that the present application solves the prediction problem of drone flight. The prediction problem is to solve the value function based on the flight strategy given by the flight strategy π, the state space S of the drone, the action A of the drone, the return reward R, the attenuation coefficient γ, and the state transition probability P. Each path strategy is a flight strategy for the drone reward, and each strategy includes the spatial position information of multiple propellers of the drone at each moment and the rotation information of each propeller in the multiple propellers. According to the state and action at each moment, the state value function and action value function of the drone are determined.

[0071] The state value function of the drone is:

[0072] v π (s) = E π (R t+1 +γR t+2 +γ 2 R t+3 +…|S t =s).

[0073] Where γ∈(0,1) is the decay coefficient.

[0074] The action value function of the drone is:

[0075] q π (s, a) = E π (R t+1 +γq π (S t+1 , A t+1 )|S t =s,A t =a).

[0076] S303: Determine the corresponding parameter information and rotation information when the state value function and the action value function take the maximum value respectively; based on the parameter information and the rotation information, select the target landing method from multiple landing methods.

[0077] It should be noted that this step solves the optimal strategy π by given the drone’s state space S, the drone’s action A, the return reward R, the decay coefficient γ, and the state transition probability P * and the optimal value function.

[0078] State value function v π (s) represents the value of the drone’s state s, and the action value function q π (s,a) represents the value of the drone taking action a in state s. The optimal strategy for drone flight is obtained using the optimization method, that is, the maximum value of the drone's state value function is determined as the optimal state in the optimal strategy, and the maximum value of the drone's action value function is determined as the optimal action of the optimal strategy.

[0079] The maximum value of the drone's state value function is:

[0080]

[0081] The maximum value of the drone's action value function is:

[0082]

[0083] After the propeller of the drone performs action a in state s, the propeller changes to state s′. After obtaining the states s and s′ of the drone, the action a of the drone between the two states can be obtained. The optimal strategy function of the drone is expressed as:

[0084]

[0085] After calculating the state value function at any moment, the state value function at each subsequent moment can be iterated from the state value function at this moment. Therefore, after the state value function of the kth round is determined, the state value function of the k+1th round is expressed as:

[0086]

[0087] By continuously iterating the above state-value function, when the value function value of each state changes less than a certain value, the reward value of each state after this moment is similar, that is, the drone is very close to the landing point. At this time, stop the iteration, get the optimal state, use the optimal state to obtain the optimal action, and finally get the optimal strategy.

[0088] exist Figure 3 In the embodiment shown, the landing point of the drone is not determined by image recognition, but by the precise positioning of a base to achieve automatic parking of the drone, without the need for reference objects; the landing location is determined based on the positioning of multiple propeller positions of the drone and the corresponding landing points of the base, and multi-point positioning has higher accuracy.

[0089] In one implementation, the above iterative solution process is solved by a neural network algorithm, and the algorithm parameters are adjusted so that the algorithm adjustment obtains the best parameter model, so that the drone can complete precise landing in a fast and safe manner every time. Optionally, the neural network algorithm can be a trained Deep Q-Network (DQN) algorithm.

[0090] In a specific embodiment, the state space of the drone includes the spatial position information of the propellers in the drone, the flight speed of the drone, and the flight direction. The action space of the drone includes the rotation state of the propellers of the drone. A value function is constructed based on the state space of the drone, the action space of the drone, and the spatial position information of the base, and a DQN algorithm is used to plan a fast and safe landing strategy for the drone.

[0091] The following is a detailed description of the method for landing a drone provided in an embodiment of the present application in conjunction with the accompanying drawings.

[0092] like Figure 6 As shown, Figure 6: is a flowchart of another method for landing a drone provided in an embodiment of the present application, and the method for landing a drone includes the following steps:

[0093] S601: Obtaining drone parameter information, rotation information of each of a plurality of propellers, and spatial position information of a plurality of landing points of a base during the flight of the drone.

[0094] The parameter information of the drone includes the parameter information of the propeller and the flight direction and speed of the drone. The parameter information of the propeller includes the spatial position information of the propeller. The parameter information of the drone is taken as the state space S of the drone, the rotation information of the propeller is taken as the action space A of the propeller, and the spatial position information of the base is taken as the environment space.

[0095] In one embodiment, the state space S of the drone is:

[0096] (x1, y1, x2, y2, x3, y3, x4, y4, V x , V y , θ, h).

[0097] Among them, (x1, y1), (x2, y2), (x3, y3), (x4, y4) represent the position information of the four propellers of the UAV flight plane; V x Indicates the horizontal flight speed in unmanned space; V y represents the vertical flight speed of the UAV in space; θ represents the flight direction of the UAV in space; h represents the relative height of the UAV in space.

[0098] In another embodiment, the state space S of the drone is:

[0099] (x1, y1, z1, x2, y2, z2, x3, y3, z3, x4, y4, z4, V x , V y , θ).

[0100] Among them, (x1, y1, z1), (x2, y2, z2), (x3, y3, z3), (x4, y4, z4) represent the position information of the four propellers of the UAV flight plane; V x Indicates the horizontal flight speed in unmanned space; V y represents the vertical flight speed of the UAV in space; θ represents the flight direction of the UAV in space; h represents the relative height of the UAV in space.

[0101] The action space A of the propeller is (a1, a2, a3, a4), where ai represents the rotation direction of each propeller, and the rotation direction includes four directions: east, south, west, and north.

[0102] S602: Inputting parameter information of the drone, rotation information of each of the multiple propellers, and spatial position information of multiple landing points of the base into the DQN model to determine the landing strategy of the drone.

[0103] Among them, the Q function in the DQN model of this application is composed of various parameterized functions, and the definition function of the Q function is: The N network model in the DQN model is a neural network: N(s): N(s) is used as a scoring function to calculate a score vector for each input spatial state, and each score in this vector is the score of the corresponding action. This application uses a feedforward neural network (FNN) as N in the formula.

[0104] Use the pre-training data to train the DQN model, obtain the data set through learning, and then use the obtained data set to train the DQN model. During the training process, the loss function between the true value and the predicted value of the training data is calculated, and then multiplied by a learning rate, and the network weights of the DQN model are gradually updated forward using back propagation. Optionally, define a loss function L, which is used to measure whether the prediction made by the Q function is standard. The loss function L can be defined as:

[0105] L=[Q(s t , a t )-(r t +γmax a Q(s t+1 , a))| 2 .

[0106] Receiving the input state information s and action information a, and making predictions about the state information and action information at later moments, is to train and optimize the parameters of the Q function by calculating the gradient of the loss function L.

[0107] Where S is the state space, A is the action set, and |A| is the number of actions in the action set. By forward passing |A|, we can find the best action a and get the maximum max a Q(s t+1 , a) value.

[0108] Among them, FNN includes a fully connected layer (Feedforward, FFW), which is composed of multiple linear or nonlinear "layers". Its H-layer hidden layer formula can be expressed as follows:

[0109]

[0110] in, Represents a combination of functions, the input layer f input Get the input state vector and generate an n1-dimensional state vector; hidden layer is the transformation function f i And the activation function a i Each hidden layer inputs an n i dimensional vector; in f i When it is a linear function or has a linear deviation, f i is a linear layer. When i≤H-1, the hidden layer f i Output an n i+1 dimensional vector to the next hidden layer. When i = H, the hidden layer outputs a vector to the output layer f output , the output layer f output Generate a final state vector. In general, f input , f1 ， ..., f h , f output There are learnable parameters b in both and when f i Use an n i When a dimensional vector is used as its input, the hidden layer has n i Hidden units.

[0111] In one embodiment, the state space S of the drone is: (x1, y1, x2, y2, x3, y3, x4, y4, V x , V y , θ, h), the propeller action space A is (a1, a2, a3, a4). Input layer f input The state vector n1 is a 12-dimensional vector corresponding to the state space S of the drone. According to the rotation information of the propeller, the output layer f output The generated output vector is a 4-dimensional vector. The feed-forward network of the neural network is designed as FFW with a linear hidden layer of two units, one unit includes 384 units and the other unit includes 128 units.

[0112] In another embodiment, the state space S of the drone is: (x1, y1, z1, x2, y2, z2, x3, y3, z3, x4, y4, z4, V x , V y ,θ), the propeller action space A is (a1, a2, a3, a4). Input layer f input The state vector n1 is a 15-dimensional vector corresponding to the state space S of the drone. According to the rotation information of the propeller, the output layer foutput The resulting output vector is a 4-dimensional vector.

[0113] Each score in the output vector is the score of the corresponding action. Let N(s)[a] be the score of the corresponding action a in N(s), and Q(s, a) is:

[0114] Q(s,a)=N(s)[a].

[0115] The gain for each N is generated by a single forward pass over all the scores, yielding:

[0116] max N(s)=max a Q(s t+1 , a).

[0117] Among them, the activation function is the corrector, f output As the function of identity recognition, we get:

[0118] a i (x) = relu(x) = max(0, x).

[0119] N(s)=A2relu(A1relu(A input s+b input )+b1)+b2.

[0120] Through the above algorithm, the experience replay technology and ε-greedy method are used to adjust the algorithm to the optimal parameter model, so as to achieve accurate positioning of the drone through the Internet of Things, allowing the drone to automatically learn to find the best parking method and achieve automatic, safe and accurate landing.

[0121] exist Figure 6 In the illustrated implementation, the drone landing point is not determined by image recognition, but by the precise positioning of a base to achieve automatic parking of the drone, without the need for reference objects; the precise landing of the drone is achieved by fixed-point positioning of the base and four points of the drone propeller, which solves the problem of requiring special reference objects and the inaccurate automatic landing position in the prior art. This patent uses artificial intelligence to learn algorithms, train the safest and most accurate landing model, and then uses this model to achieve the demand for automatic landing, ensuring the safe landing of drones in different environments. The algorithm used in this application uses experience replay technology and ε-greedy method, which improves the stability of the algorithm and the balance between exploration and development of the algorithm. It combines key technologies in the fields of 5G applications and artificial intelligence to achieve automatic landing of drones.

[0122] The embodiment of the present application can divide the drone landing device into functional modules or functional units according to the above method example. For example, each functional module or functional unit can be divided according to each function, or two or more functions can be integrated into one processing module. The above integrated module can be implemented in the form of hardware or in the form of software functional modules or functional units. Among them, the division of modules or units in the embodiment of the present application is schematic, which is only a logical functional division. There may be other division methods in actual implementation.

[0123] like Figure 7 As shown, Figure 7 70 is a schematic diagram of the structure of a drone landing device provided in an embodiment of the present application, wherein the device includes an acquisition module 701 and a processing module 702.

[0124] The acquisition module 701 is used to acquire the parameter information of the drone, the rotation information of multiple propellers of the drone, and the position information of multiple landing points during the flight of the drone. The parameter information of the drone includes the spatial position information of the propellers, and the rotation information includes the rotation direction and rotation speed of the propellers. One landing point corresponds to one propeller. For example, the acquisition module 701 is used to execute the above S201.

[0125] The processing module 702 determines the landing strategy of the drone based on the parameter information of the drone, the rotation information of multiple propellers of the drone, and the position information of multiple landing points; the landing strategy is used to adjust the action posture of the drone during the landing process. For example, the processing module 702 is used to execute the above S202.

[0126] The processing module 702 is further configured to send first landing instruction information to the UAV, where the first landing instruction information is used to indicate a landing strategy so that the UAV uses the landing strategy to land during the landing process. For example, the processing module 702 is configured to execute the above S103.

[0127] One possible implementation method is that the processing module 702 is specifically used to determine the landing strategy of the drone based on the spatial position information of the drone's multiple propellers, the drone's flight direction and flight speed, the rotation information of the drone's multiple propellers, and the position information of multiple landing points.

[0128] In one possible implementation, processing module 702 is further specifically used to determine multiple landing methods of the UAV based on parameter information of the UAV, rotation information of multiple propellers of the UAV, and position information of multiple landing points; processing module 702 is further specifically used to select a target landing method from multiple landing methods as the landing strategy of the UAV.

[0129] In a possible implementation method, the processing module 702 is further specifically used to obtain the parameter information of the UAV at a second moment based on the parameter information of the UAV at a first moment and the rotation information of multiple propellers of the UAV; the processing module 702 is further specifically used to determine multiple rotation information of multiple propellers of the UAV at a second moment based on the parameter information of the UAV at the second moment and the position information of multiple landing points, and determine multiple landing methods of the UAV, one rotation information corresponds to one landing method; the second moment is a moment after the first moment.

[0130] In a possible implementation, processing module 702 is also used to obtain the state value function and action value function corresponding to each landing method; the state value function is used to determine the value of parameter information, and the action value function is used to determine the value of rotation information; processing module 702 is also used to determine the corresponding parameter information and rotation information when the state value function and the action value function take the maximum value respectively; processing module 702 is also used to select a target landing method from multiple landing methods based on the parameter information and the rotation information.

[0131] In a possible implementation, the processing module 702 is further specifically used to input the parameter information of the drone, the rotation information of multiple propellers of the drone, and the position information of multiple landing points into a preset deep Q network model to obtain the landing strategy of the drone.

[0132] It is understandable that the above-mentioned drone landing device can also be implemented by hardware. For example, when implemented by hardware, the acquisition module 701 and the non-processing module 702 in the embodiment of the present application are integrated on the processor. The structure of the hardware can be as follows: Figure 8 shown.

[0133] Figure 8 A possible hardware structure diagram of the device for landing a drone involved in the above embodiment is shown. The device for landing a drone includes: a processor 802. Optionally, the device for landing a drone also includes: a communication interface 803, a memory 801 and a bus 804.

[0134] The processor 802 is used to control and manage the actions of the device for landing the drone, for example, to execute the steps performed by the above-mentioned processing module 702, and / or to perform other processes of the technology described herein. Optionally, the processor 802 can also execute the steps performed by the above-mentioned processing module 702. The above-mentioned processor 802 can be a variety of exemplary logic blocks, modules and circuits that implement or execute the contents described in conjunction with the disclosure of this application. The processor can be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field programmable gate array or other programmable logic device, a transistor logic device, a hardware component or any combination thereof. It can implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the contents disclosed in this application. The processor can also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0135] The communication interface 803 is used to support the communication between the device for landing the drone and other network entities, for example, executing the steps performed by the above-mentioned acquisition module 701.

[0136] The memory 801 is used to store program codes and data of the device for landing the drone. For example, the memory 801 may be a memory in the device for landing the drone, etc. The memory may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a read-only memory, a flash memory, a hard disk or a solid-state drive; the memory may also include a combination of the above-mentioned types of memory.

[0137] The bus 804 may be an Extended Industry Standard Architecture (EISA) bus, etc. The bus 804 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 8 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0138] Fig. 9 90 is a schematic diagram of the structure of a chip 90 provided in an embodiment of the present application. The chip 90 includes one or more (including two) processors 901. Optionally, the chip 90 also includes a communication interface 903, a bus 902 and a memory 904.

[0139] Among them, the above-mentioned processor 901 can implement or execute various exemplary logic blocks, units and circuits described in conjunction with the disclosure of this application. The processor can be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field programmable gate array or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It can implement or execute various exemplary logic blocks, units and circuits described in conjunction with the disclosure of this application. The processor can also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of DSP and microprocessors, etc.

[0140] The memory 904 may include a read-only memory and a random access memory, and provides operation instructions and data to the processor 901. A portion of the memory 904 may also include a non-volatile random access memory (NVRAM).

[0141] In some implementations, the memory 904 stores the following elements, execution modules or data structures, or a subset thereof, or an extended set thereof.

[0142] In the embodiment of the present application, the corresponding operation is performed by calling the operation instruction stored in the memory 904 (the operation instruction may be stored in the operating system).

[0143] The memory 904 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a read-only memory, a flash memory, a hard disk or a solid-state drive; the memory may also include a combination of the above types of memory.

[0144] The bus 902 may be an Extended Industry Standard Architecture (EISA) bus, etc. The bus 902 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Fig. 9 The fact that only one line is used in the diagram does not mean that there is only one bus or only one type of bus.

[0145] Through the description of the above implementation methods, technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device and unit described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0146] An embodiment of the present application provides a computer program product including instructions. When the computer program product is run on a computer, the computer is enabled to execute the method in the above method embodiment.

[0147] An embodiment of the present application also provides a computer-readable storage medium, in which instructions are stored. When the instructions are executed on a computer, the computer executes the method in the method flow shown in the above method embodiment.

[0148] Among them, the computer readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a register, a hard disk, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above, or any other form of computer readable storage medium known in the art. An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an application-specific integrated circuit (ASIC). In the embodiments of the present application, a computer-readable storage medium may be any tangible medium that contains or stores a program, which may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0149] Since the device, computer-readable storage medium, and computer program product in the embodiments of the present application can be applied to the above-mentioned method, the technical effects that can be obtained can also refer to the above-mentioned method embodiments, and the embodiments of the present application will not be repeated here.

[0150] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0151] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0152] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0153] The above are only specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for landing a drone, characterized in that: The method comprises: Acquire parameter information of the drone, rotation information of multiple propellers of the drone, and position information of multiple landing points during the flight of the drone, wherein the parameter information of the drone includes spatial position information of the propeller, and the rotation information includes rotation direction and rotation speed of the propeller, and one landing point corresponds to one propeller; Determine a landing strategy of the drone based on the parameter information of the drone, the rotation information of the multiple propellers of the drone, and the position information of the multiple landing points; the landing strategy is used to adjust the action posture of the drone during the landing process; Sending first landing instruction information to the drone, where the first landing instruction information is used to indicate the landing strategy, so that the drone uses the landing strategy to land during the landing process.

2. The method according to claim 1, characterized in that The parameter information of the drone also includes the flight direction and flight speed of the drone.

3. The method according to claim 2, characterized in that The determining of the landing strategy of the drone based on the parameter information of the drone, the rotation information of the plurality of propellers of the drone, and the position information of the plurality of landing points includes: Determining multiple landing modes of the drone based on the parameter information of the drone, the rotation information of multiple propellers of the drone, and the position information of the multiple landing points; A target landing method is selected from the multiple landing methods as the landing strategy of the UAV.

4. The method according to claim 3, characterized in that The determining of multiple landing modes of the drone based on the parameter information of the drone, the rotation information of multiple propellers of the drone, and the position information of the multiple landing points includes: Based on the parameter information of the drone at the first moment and the rotation information of the plurality of propellers of the drone, obtaining the parameter information of the drone at the second moment; Determine multiple rotation information of multiple propellers of the drone at the second moment based on the parameter information of the drone at the second moment and the position information of the multiple landing points; Based on the multiple rotation information of the multiple propellers of the drone at the second moment, multiple landing modes of the drone are determined, and one rotation information corresponds to one landing mode; the second moment is a moment after the first moment.

5. The method according to claim 3 or 4, characterized in that: The step of selecting a target landing method from the multiple landing methods as a landing strategy for the UAV includes: Obtaining a state value function and an action value function corresponding to each landing mode; the state value function is used to determine the value of parameter information, and the action value function is used to determine the value of rotation information; For each landing mode, determining the parameter information and rotation information corresponding to the landing mode when the state value function and the action value function corresponding to the landing mode respectively take the maximum value; The target landing mode is selected from the plurality of landing modes based on the parameter information and the rotation information corresponding to each landing reflection mode.

6. The method according to claim 1 or 2, characterized in that: The determining of the landing strategy of the drone based on the parameter information of the drone, the rotation information of the plurality of propellers of the drone, and the position information of the plurality of landing points includes: The parameter information of the drone, the rotation information of the multiple propellers of the drone, and the position information of the multiple landing points are input into a preset deep Q network model to obtain the landing strategy of the drone.

7. A device for landing a drone, characterized in that: The device comprises an acquisition module and a processing module; The acquisition module is used to acquire parameter information of the drone, rotation information of multiple propellers of the drone, and position information of multiple landing points during the flight of the drone, wherein the parameter information of the drone includes spatial position information of the propeller, and the rotation information includes rotation direction and rotation speed of the propeller. The base is used to carry the drone, and one landing point corresponds to one propeller; The processing module is further used to determine a landing strategy of the drone based on the parameter information of the drone, the rotation information of the multiple propellers of the drone, and the position information of the multiple landing points; the landing strategy is used to adjust the action posture of the drone during the landing process; The processing module is further used to send first landing instruction information to the drone, where the first landing instruction information is used to indicate the landing strategy so that the drone uses the landing strategy to land during the landing process.

8. The device according to claim 7, characterized in that The parameter information of the drone also includes the flight direction and flight speed of the drone.

9. The device according to claim 8, characterized in that The processing module is further specifically used to determine multiple landing modes of the drone based on the parameter information of the drone, the rotation information of multiple propellers of the drone, and the position information of the multiple landing points; The processing module is further specifically configured to select a target landing method from the multiple landing methods as a landing strategy for the UAV.

10. The device according to claim 7, characterized in that The processing module is further specifically configured to obtain parameter information of the drone at a second moment based on the parameter information of the drone at a first moment and the rotation information of a plurality of propellers of the drone; The processing module is further specifically configured to determine a plurality of rotation information of a plurality of propellers of the drone at the second moment based on the parameter information of the drone at the second moment and the position information of the plurality of landing points; The processing module is also specifically used to determine multiple landing methods of the drone based on multiple rotation information of multiple propellers of the drone at the second moment, one rotation information corresponds to one landing method; the second moment is a moment after the first moment.

11. The device according to claim 9 or 10, characterized in that The processing module is further specifically used to obtain a state value function and an action value function corresponding to each landing mode; the state value function is used to determine the value of the parameter information, and the action value function is used to determine the value of the rotation information; The processing module is further specifically used to determine, for each landing mode, parameter information and rotation information corresponding to the landing mode when the state value function and the action value function corresponding to the landing mode are respectively maximized; The processing module is further specifically used to select the target landing mode from the multiple landing modes based on the parameter information and rotation information corresponding to each landing reflection mode.

12. The device according to claim 7 or 8, characterized in that The processing module is also specifically used to input the parameter information of the drone, the rotation information of multiple propellers of the drone, and the position information of the multiple landing points into a preset deep Q network model to obtain the landing strategy of the drone.

13. A device for landing a drone, characterized in that: include: A processor and a communication interface; the communication interface is coupled to the processor, and the processor is used to run a computer program or instruction to implement the method for landing a drone as described in any one of claims 1 to 6.

14. A computer-readable storage medium, wherein instructions are stored in the computer-readable storage medium, characterized in that: When the computer executes the instruction, the computer executes the method for landing the drone as described in any one of claims 1 to 6 above.

Images