Decision method and system for low-power autonomous landing of vertical take-off and landing unmanned aerial vehicle

Abstract

This invention discloses a decision-making method and system for low-battery autonomous landing of a vertical take-off and landing (VTOL) drone, relating to the field of drone landing decision-making technology. The method includes the following steps: acquiring terrain point cloud data, establishing a neural network model, and extracting and predicting terrain feature parameters, including obstacle density and maximum height; combining cruise area point cloud data and real-time environmental parameters, including ambient temperature and wind speed, to calculate the obstacle impact index and vertical energy consumption impact index, and correcting the return-to-home battery threshold; further, dynamically correcting the return-to-home battery threshold by combining battery health parameters, including storage time and number of charge cycles, and ambient temperature; comparing the dynamic return-to-home battery threshold with the real-time battery level to generate a landing or return-to-home decision command. By comprehensively considering the energy consumption of the drone during horizontal flight and VTOL under different terrain conditions, and simultaneously adjusting based on battery performance, the drone's battery management strategy becomes more intelligent and precise.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) landing decision technology, specifically to a decision-making method and system for low-battery autonomous landing of a vertical take-off and landing (VTOL) UAV. Background Technology

[0002] With the rapid development of drone technology, vertical takeoff and landing (VTOL) drones, possessing both the efficient cruising capabilities of fixed-wing flight and the vertical takeoff and landing capabilities of multi-rotor flight, are widely used in logistics, disaster relief, and topographic mapping. However, when drones perform missions over long distances and in complex environments, the limitation of battery life becomes a key factor affecting their mission completion rate and safety. Especially during mission execution, drones need to make decisions about returning to base and landing based on their remaining battery power, and unplanned crashes due to insufficient battery power can have serious consequences for the drone itself, the mission target, and the surrounding environment.

[0003] Existing low-battery return-to-home or landing decision-making methods primarily rely on monitoring only the remaining power (SOC, State of Charge) or battery voltage. However, these methods have significant limitations in practical applications. Firstly, SOC or voltage estimations are easily affected by ambient temperature, flight load, and battery aging, leading to misjudgments. For example, high-load flight or low-temperature environments may cause a sudden drop in battery voltage, triggering premature return-to-home; while battery aging may cause the estimated power level to differ from the actual remaining capacity, thus delaying the return-to-home time. Secondly, existing methods typically assume a relatively simple mission environment and do not adequately consider complex terrain and dynamic environmental factors. In practical applications, the UAV's return-to-home or landing path may traverse areas with dense obstacles, such as tall buildings, mountains, or forests, which places higher demands on the UAV's path planning and energy consumption calculations.

[0004] Furthermore, traditional return-to-home battery thresholds are typically statically set, failing to dynamically adapt to changes in various environmental parameters. For example, strong winds significantly increase flight power consumption, while undulating terrain can lead to higher vertical takeoff and landing energy consumption; static return-to-home battery thresholds often fall short of the requirements for these scenarios. This decision-making model not only limits the flexibility of drones but also increases the risk of mission failure due to insufficient return-to-home battery estimation.

[0005] Therefore, how to accurately determine the best time to return home or land in complex terrain and dynamic environments, by combining the battery health status of the drone, terrain features and real-time environmental parameters, has become a key technical challenge for low battery management of vertical take-off and landing drones.

[0006] In the prior art, CN117055616A discloses a method for low-battery autonomous landing decision-making based on a vertical take-off and landing (VTOL) UAV. The method includes: setting a battery-based return-to-home (ROW) mode and a voltage-based ROW mode for the UAV; determining whether the battery or voltage meets the take-off conditions; if so, proceeding to the next step; otherwise, terminating take-off; determining the ROW mode; if the battery-based ROW mode is selected, the voltage-based ROW mode is also retained, and a safe landing battery percentage is determined; the actual battery level is acquired in real-time during UAV flight, and the ROW decision is based on the battery level; when the ROW condition is triggered, the UAV executes the battery-based ROW decision. If the voltage-based ROW mode is selected, the actual voltage is acquired in real-time during UAV flight, and the ROW decision is based on the voltage; when the ROW condition is triggered, the UAV executes the voltage-based ROW decision; if the battery data is read incorrectly, the voltage-based ROW function is automatically activated. However, this method does not mention how to handle obstacles (such as buildings, power lines, trees, terrain undulations, etc.) on the ROW path. During the ROW process, especially in a low-battery state, the UAV may fail to avoid obstacles in the ROW path, leading to a collision risk. Furthermore, only the influence of wind speed was considered, without taking into account the impact of factors such as temperature on power consumption. Therefore, relying solely on a single environmental factor reduces the accuracy and effectiveness of judgment and decision-making.

[0007] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0008] The purpose of this invention is to provide a decision-making method and system for low-battery autonomous landing based on a vertical take-off and landing unmanned aerial vehicle (UAV) to solve the problems mentioned in the background art.

[0009] To achieve the above objectives, the present invention provides the following technical solution:

[0010] A decision-making method for low-battery autonomous landing based on a vertical take-off and landing (VTOL) UAV, comprising the following steps:

[0011] A terrain point cloud data with several known terrain feature parameters is obtained as sample point cloud data. A neural network model is established based on the sample point cloud data. The sample point cloud data is used as the input of the model, and the corresponding terrain feature parameters are used as labels to train the neural network model to obtain a terrain feature prediction model. The terrain feature parameters include obstacle density and maximum height.

[0012] Collect terrain point cloud data of the UAV flight cruise area, input the terrain point cloud data of the UAV flight cruise area into the trained terrain feature prediction model, and obtain the predicted values ​​of terrain feature parameters of the UAV flight cruise area. The predicted values ​​of terrain feature parameters of the UAV flight cruise area include obstacle density prediction value and maximum height prediction value.

[0013] The system acquires real-time flight altitude data of the UAV during flight cruise, and simultaneously acquires real-time environmental parameters at the current flight altitude and the horizontal distance to the target take-off and landing point. Based on the obtained real-time environmental parameters, the horizontal distance to the take-off and landing point, and the predicted obstacle density, the system calculates the obstacle impact index. The environmental parameters include ambient temperature, wind speed, and wind direction angle.

[0014] Obtain the vertical height of the target take-off and landing point. Using the obtained vertical height of the target take-off and landing point and the current flight altitude, combined with the total mass of the UAV, calculate and generate the vertical energy consumption impact index. Based on the vertical energy consumption impact index and the obstacle impact index, correct the set return-to-home power threshold to obtain the return-to-home power threshold.

[0015] The system collects battery health parameters of the drone and dynamically adjusts the return-to-home power threshold based on these parameters and ambient temperature. This dynamic return-to-home power threshold is then compared with the drone's real-time power level. Based on the comparison, a corresponding decision command is issued. The battery health parameters include battery storage time and the number of charge cycles.

[0016] Furthermore, based on the sample point cloud data, a neural network model is established. Specifically, a terrain feature prediction model is built based on the Long Short-Term Memory (LSTM) network model. Activation functions and optimization algorithms are selected, with the Tanh function chosen as the activation function and Adam as the optimization algorithm for the LSTM model. The formula for the Tanh function is:

[0017]

[0018] In the formula, f(r) represents the Tanh function, and the independent variable r represents the weighted sum of the neuron's inputs, that is, the result of the weighted sum of the inputs received by the neuron from the previous layer.

[0019] Simultaneously, the hyperparameters of the LSTM model are set, including: number of network layers, number of iterations, learning rate, batch size, number of training iterations, number of batches, and number of hidden layer neurons.

[0020] The network is set to a 3-layer network structure, the number of iterations is set to 200, the learning rate is set to 0.001, the batch size is set to 32, the number of training iterations is set to 100, the batch size is set to 256, and the number of hidden layer neurons is set to 32.

[0021] The trained terrain feature prediction model takes terrain point cloud data as input and outputs terrain feature parameters corresponding to the terrain.

[0022] Furthermore, the obstacle impact index is calculated based on the obtained real-time environmental parameters, the horizontal distance to the take-off and landing point, and the predicted obstacle density. The specific formula used to calculate the obstacle impact index is as follows:

[0023]

[0024] In the formula, IES represents the obstacle impact index at the current moment, λ is the obstacle density in the flight cruise area, and P wind This represents the wind resistance power consumption experienced by the drone at the current moment, and L represents the horizontal distance between the drone's current position and the take-off and landing point.

[0025] Wherein is the wind resistance power P experienced by the drone at the current moment. wind The calculations are performed using real-time environmental parameters collected, and the specific calculation formula is as follows:

[0026]

[0027] In the formula, ρ is the air density at the current flight altitude, A is the effective frontal area of ​​the UAV at the current moment, and C... d V is the air drag coefficient. eff This indicates the wind speed in the environment at the current moment, at the drone's current flight altitude;

[0028] The formula used to calculate the effective windward area of ​​the drone at the current moment is as follows:

[0029]

[0030] In the formula, θ is the angle between the current wind direction and the flight direction of the UAV, and A on and A bise These are the projected areas from the top-down view and the projected areas from the front view of the drone, respectively.

[0031] Furthermore, by combining the obtained vertical altitude of the target take-off and landing point and the current flight altitude with the total mass of the UAV, a vertical energy consumption impact index is calculated. The formula used to calculate the vertical energy consumption impact index is as follows:

[0032]

[0033] In the formula, E clb The vertical energy consumption impact index is given by m, where m is the total mass of the UAV, g is the acceleration due to gravity, and h is the altitude of the target take-off and landing point. f h represents the current altitude of the drone.for η is the predicted maximum height of the obstacle, and η is the efficiency of the UAV rotor motor.

[0034] The efficiency η of the UAV rotor motor is obtained by correcting for ambient temperature and current flight altitude. The specific formula used to calculate the efficiency η of the UAV rotor motor is as follows:

[0035]

[0036] In the formula, η0 is the standard efficiency of the UAV rotor motor, T0 is the reference temperature, and T hf β represents the temperature at the current location of the drone, and β is the altitude correction factor.

[0037] Furthermore, based on the vertical energy consumption impact index and the obstacle impact index, the set return-to-base power threshold is corrected to obtain the power threshold affecting return-to-base. The formula used to calculate the power threshold affecting return-to-base is as follows:

[0038]

[0039] In the formula, yz represents the threshold of the return-to-home power consumption at the current moment, IES is the obstacle impact index at the current moment during the drone's flight cruise, yz0 is the set initial threshold of the return-to-home power consumption, and ω1 and ω2 are the weighting coefficients of the obstacle impact index and the vertical energy consumption impact index, respectively, where ω1 < ω2 and both ω1 and ω2 are greater than 0.

[0040] Furthermore, based on battery health parameters, the threshold affecting the return-to-home battery level is dynamically corrected to obtain the dynamic return-to-home battery level threshold. The formula used to calculate the dynamic return-to-home battery level threshold is as follows:

[0041]

[0042] In the formula, yz′ is the dynamic return-to-home power threshold, and cs and cj are the battery storage time and the number of charge cycles, respectively.

[0043] The dynamic return-to-home battery threshold is compared with the drone's real-time battery level. Based on the comparison result, a corresponding decision command is issued. The specific logic for issuing the corresponding decision command is as follows:

[0044] When ZH≥1.0*yz′, the cruise status is determined to be healthy, indicating that sufficient power is currently available for cruise control;

[0045] When 0.4*yz′≤ZH<1.0*yz′, the cruise mode is determined to be in the middle, and a prompt is issued to pay attention to the remaining battery power;

[0046] When 0≤ZH<0.4*yz′, the cruise status is judged to be low, indicating that the return trip should be initiated immediately.

[0047] ZH represents the current battery level displayed on the drone.

[0048] This invention also provides a decision-making system for low-battery autonomous landing based on a vertical takeoff and landing (VTOL) unmanned aerial vehicle (UAV). This system is used to execute the aforementioned decision-making method for low-battery autonomous landing based on a VTOL UAV, and includes:

[0049] The prediction model training module is used to acquire terrain point cloud data with several known terrain feature parameters as sample point cloud data. Based on the sample point cloud data, a neural network model is established. The sample point cloud data is used as the input of the model, and the corresponding terrain feature parameters are used as labels to train the neural network model to obtain a terrain feature prediction model. The terrain feature parameters include obstacle density and maximum height.

[0050] The cruise altitude confirmation module is used to collect terrain point cloud data of the UAV's flight cruise area, input the terrain point cloud data of the UAV's flight cruise area into the trained terrain feature prediction model, and obtain the predicted values ​​of terrain feature parameters of the UAV's flight cruise area. The predicted values ​​of terrain feature parameters of the UAV's flight cruise area include obstacle density prediction values ​​and maximum height prediction values.

[0051] The environmental disturbance analysis module is used to acquire real-time flight altitude data of the UAV during flight cruise, and at the same time acquire real-time environmental parameters at the current flight altitude and the horizontal distance to the target take-off and landing point. Based on the obtained real-time environmental parameters, the horizontal distance to the take-off and landing point and the predicted value of obstacle density, the obstacle impact index is calculated. The environmental parameters include ambient temperature, wind speed and wind direction angle.

[0052] The vertical energy consumption analysis module is used to obtain the vertical height of the target take-off and landing point. By using the obtained vertical height of the target take-off and landing point and the current flight altitude, combined with the total mass of the UAV, a vertical energy consumption impact index is calculated and generated. Based on the vertical energy consumption impact index and the obstacle impact index, the set return-home power threshold is corrected to obtain the return-home power threshold.

[0053] The landing decision generation module is used to collect the battery health parameters of the UAV, and dynamically correct the return-to-home power threshold based on the battery health parameters and ambient temperature to obtain the dynamic return-to-home power threshold. The dynamic return-to-home power threshold is compared with the real-time power of the UAV, and the corresponding decision command is issued based on the comparison result. The battery health parameters include battery storage time and number of charge cycles.

[0054] Compared with the prior art, the beneficial effects of the present invention are:

[0055] First, this solution fully considers the complexity of the terrain in the UAV's flight area. By collecting terrain point cloud data and using a neural network model to predict terrain feature parameters, it can accurately obtain core terrain information such as obstacle density and maximum altitude in the flight area. During UAV cruise flight or return flight, this method effectively avoids the problem of traditional low-battery decision-making methods lacking awareness of obstacles in the flight path. By predicting terrain feature parameters, the system can dynamically adjust flight altitude and cruise path, thereby reducing the risk of obstacle collisions and improving the safety and mission completion rate of the UAV in complex environments.

[0056] Secondly, real-time environmental parameters, including key variables such as ambient temperature, wind speed, and wind direction, are incorporated into the calculation of the return-to-home battery threshold. This allows for dynamic assessment of the environment's impact on flight energy consumption. By comprehensively considering the energy consumption of the UAV during horizontal flight and vertical takeoff and landing under different terrain conditions, the return-to-home battery threshold is made more consistent with actual flight requirements. Finally, this solution fully considers the impact of battery health status on the return-to-home decision. By combining health parameters such as battery storage time and number of charge cycles, as well as the dynamic impact of ambient temperature on battery performance, the return-to-home battery threshold is further refined. This significantly reduces the error in return-to-home judgment caused by battery aging or temperature changes, making the UAV's battery management strategy more intelligent and precise. Attached Figure Description

[0057] Figure 1 This is a schematic diagram of the overall method flow of the present invention;

[0058] Figure 2 This is a schematic diagram of the overall system structure of the present invention. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0060] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0061] Example:

[0062] Please see Figure 1 The present invention provides a technical solution:

[0063] A decision-making method for low-battery autonomous landing based on a vertical take-off and landing (VTOL) UAV, comprising the following steps:

[0064] Step 1: Obtain terrain point cloud data with known terrain feature parameters as sample point cloud data. Based on the sample point cloud data, establish a neural network model, using the sample point cloud data as the input to the model and the corresponding terrain feature parameters as labels, to train the neural network model and obtain a terrain feature prediction model. The terrain feature parameters include obstacle density and maximum height.

[0065] The obstacle density is specifically the number of obstacle points per unit volume or unit area, calculated through the following steps: Divide the terrain point cloud data into several spatial units of fixed size (such as 3D grid volume or 2D grid area); determine the ground points and obstacle points in the terrain point cloud data; count the number of obstacle point clouds in each spatial unit; regard the number of obstacle point clouds as the number of obstacle points in that unit; and the obstacle density is the ratio of the number of point clouds to the voxel volume or grid area.

[0066] The specific method for determining ground points and obstacle points in the terrain point cloud data is as follows: Flat ground detection: The RANSAC algorithm is used to perform planar fitting on the point cloud. Multiple points are randomly selected from the point cloud data, and a planar model is fitted. The distance from each point to the fitted plane is calculated. Points with a distance less than a certain error threshold (such as 0.05 meters) are classified as ground points. Through iterative optimization, the main plane containing the most points is found and used as the ground plane. After removing the ground from the point cloud, the remaining point cloud is basically the obstacle point cloud. Based on the height characteristics of the obstacles, points with a height exceeding a certain threshold (such as 0.2 meters) are marked as obstacle points.

[0067] Based on sample point cloud data, a neural network model is established. Specifically, a terrain feature prediction model is built using a Long Short-Term Memory (LSTM) network model. Activation functions and optimization algorithms are selected, with the Tanh function chosen as the activation function and Adam as the optimization algorithm for the LSTM model. The formula for the Tanh function is:

[0068]

[0069] In the formula, f(r) represents the Tanh function, and the independent variable r represents the weighted sum of the neuron's inputs, that is, the result of the weighted sum of the inputs received by the neuron from the previous layer.

[0070] Simultaneously, the hyperparameters of the LSTM model are set, including: number of network layers, number of iterations, learning rate, batch size, number of training iterations, number of batches, and number of hidden layer neurons.

[0071] The network is set to a 3-layer network structure, the number of iterations is set to 200, the learning rate is set to 0.001, the batch size is set to 32, the number of training iterations is set to 100, the batch size is set to 256, and the number of hidden layer neurons is set to 32.

[0072] The trained terrain feature prediction model takes terrain point cloud data as input and outputs terrain feature parameters corresponding to the terrain.

[0073] Terrain point cloud data often exhibits relatively continuous changes in height and density, such as a gradual transition from flat terrain to steep terrain. This continuous trend can be regarded as a sequence characteristic, and LSTM models can effectively capture these trends to predict the distribution density and maximum height of obstacles.

[0074] Topographic point cloud data often has complex distribution characteristics, which may require consideration of both short-term features of local areas and long-term features of a wider range. In a complex canyon, the obstacle density in some areas may depend not only on neighboring points but also on topographic changes in more distant areas. Similarly, the prediction of maximum height may involve cumulative height change trends rather than just local height values.

[0075] Topographic point cloud data is typically irregular, with point distribution potentially varying in density due to terrain complexity. LSTM models, through their sequence processing capabilities, can flexibly adapt to this irregularity after input data preprocessing. By organizing point cloud data into a spatial sequence as an input sequence, LSTM can effectively learn the data's feature distribution without relying on a regular grid structure.

[0076] Step 2: Collect terrain point cloud data of the UAV flight cruise area, input the terrain point cloud data of the UAV flight cruise area into the trained terrain feature prediction model, and obtain the predicted values ​​of terrain feature parameters of the UAV flight cruise area. The predicted values ​​of terrain feature parameters of the UAV flight cruise area include obstacle density prediction value and maximum height prediction value.

[0077] The specific steps for acquiring terrain point cloud data of the drone's flight cruise area include: the drone carrying sensor equipment suitable for terrain data acquisition, such as lidar, which is currently the preferred choice for acquiring high-precision terrain point cloud data. Its working principle is to measure the distance to the ground through a laser beam to generate high-resolution three-dimensional point cloud data. LiDAR is suitable for complex terrain (such as forests, densely built-up areas, etc.) and can penetrate tree canopies to recover ground data; or a high-definition camera can be used, equipped with an RGB high-definition camera, to take multiple overlapping two-dimensional images from the air and generate three-dimensional point clouds using structured light technology in computer vision or photogrammetry. The camera method is suitable for open terrain, but the data recovery effect is poor in densely vegetated areas.

[0078] After acquiring terrain data, the raw point cloud data or image data needs to be preprocessed. This typically includes the following steps: Synchronizing the LiDAR scan data with the UAV's GPS and IMU data to ensure accurate geographic location of the point cloud data; the raw point cloud data or image data may contain noise (such as signal reflection errors or external environmental interference), which needs to be removed using algorithms (such as the Random Sample Consensus Algorithm RANSAC); and then using the processed point cloud data as input to the model.

[0079] Step 3: Obtain real-time flight altitude data of the UAV during flight cruise, and simultaneously obtain real-time environmental parameters at the current flight altitude and the horizontal distance to the target take-off and landing point. Calculate the obstacle impact index based on the obtained real-time environmental parameters, the horizontal distance to the take-off and landing point, and the predicted obstacle density. The environmental parameters include ambient temperature, wind speed, and wind direction angle.

[0080] The obstacle impact index is calculated based on the obtained real-time environmental parameters, the horizontal distance to the take-off and landing point, and the predicted obstacle density. The specific formula for calculating the obstacle impact index is as follows:

[0081]

[0082] In the formula, IES represents the obstacle impact index at the current moment, λ is the obstacle density in the flight cruise area, and P wind This represents the wind resistance power consumption experienced by the drone at the current moment, and L represents the horizontal distance between the drone's current position and the take-off and landing point.

[0083] It should be noted that the obstacle impact index IES at the current moment during the drone's flight cruise is characterized by a combination of obstacle density in the flight cruise area, real-time horizontal distance from the take-off and landing point, and wind resistance power consumption. The higher the obstacle impact index IES value at the current moment during the drone's flight cruise, the greater the obstacle during the drone's cruise and the more power it requires.

[0084] Among them, the obstacle density λ in the flight cruise area represents the complexity of the obstacle distribution in the environment, which directly affects the flight path planning and energy consumption of the UAV. When the obstacle density is high, the probability of an emergency increases, and the UAV needs to reserve more power to avoid emergency operations, which will increase the flight path length and energy consumption. Therefore, the obstacle density λ in the flight cruise area is proportional to the obstacle influence index IES at the current moment in the UAV's flight cruise process. The logarithmic function ln(1+λ) is used to appropriately smooth the impact of the increase in obstacle density value on the result. Obstacle density usually changes drastically due to terrain complexity, but the logarithmic transformation can avoid the unreasonable amplification of the obstacle influence index IES0 at the current moment in the UAV's flight cruise process when the density value is too large.

[0085] The horizontal distance L between the drone's current location and the takeoff and landing point directly affects the energy and time required for its return journey. As the horizontal distance increases, the drone's return path becomes longer, and energy consumption increases significantly. Therefore, the horizontal distance L between the drone's current location and the takeoff and landing point is directly proportional to the obstacle effect index IES, expressed as the square of L. 2 This reflects the nonlinear effect of distance on energy consumption. Energy consumption typically increases nonlinearly with flight distance, and this method can more accurately characterize the nonlinear relationship of energy consumption as the return path increases with distance.

[0086] Wind speed and direction are important environmental factors affecting the endurance of drones. When flying in a direction inconsistent with the wind, the drone needs to overcome higher wind resistance, resulting in significantly increased energy consumption. Therefore, the wind resistance power consumption P experienced by the drone at the current moment is... wind It is proportional to the hindering effect index IES, through an exponential function. With the denominator representing a direct proportional relationship, the wind resistance power consumption value is converted into a non-linear relationship, representing the wind resistance power consumption P experienced by the drone at the current moment. wind Significant impact on the hindering effect index IES.

[0087] Wherein is the wind resistance power P experienced by the drone at the current moment. wind The calculations are performed using real-time environmental parameters collected, and the specific calculation formula is as follows:

[0088]

[0089] In the formula, ρ is the air density at the current flight altitude, A is the effective frontal area of ​​the UAV at the current moment, and C... d V is the air drag coefficient. eff This indicates the wind speed in the environment at the current moment, at the drone's current flight altitude;

[0090] The air drag coefficient Cd The method for obtaining the drag coefficient is as follows: place the drone or its model in a wind tunnel, measure the aerodynamic force (especially drag) data by changing the wind speed, and then calculate the drag coefficient, or refer to the typical air drag coefficient value of similar aircraft, which is generally between 0.1 and 1.5.

[0091] The formula used to calculate the effective windward area of ​​the drone at the current moment is as follows:

[0092]

[0093] In the formula, θ is the angle between the current wind direction and the flight direction of the UAV, and A on and A bise These are the projected areas from the top-down view and the projected areas from the front view of the drone, respectively.

[0094] A small electronic anemometer (such as an ultrasonic anemometer or a mechanical blade anemometer) is installed on the drone. The anemometer measures the direction and magnitude of the wind speed in real time. The wind speed direction is the wind direction angle. IMU sensors (including accelerometers, gyroscopes, and magnetometers) can acquire the drone's attitude and direction of motion in real time. Based on the obtained direction of motion and wind speed, the angle between the current wind direction angle and the drone's flight direction is determined.

[0095] The effective windward area is calculated by averaging the projected area of ​​the drone in the top-down direction and the projected area in the frontal direction. The projected area in the top-down direction specifically refers to the projected area observed from the top of the drone (perpendicular to the ground). This is the horizontal projection of the drone from the top view, mainly composed of the horizontal projections of the fuselage, support, propeller, etc. The projected area in the frontal direction specifically refers to the projected area observed from the front of the drone (parallel to the flight direction). This is the vertical projection of the drone from the front view, mainly composed of the frontal projections of the fuselage, support, and other components.

[0096] Step 4: Obtain the vertical height of the target take-off and landing point. Using the obtained vertical height of the target take-off and landing point and the current flight altitude, combined with the total mass of the UAV, calculate and generate the vertical energy consumption impact index. Based on the vertical energy consumption impact index and the obstacle impact index, correct the set return-to-home power threshold to obtain the power threshold affecting the return-to-home.

[0097] By combining the vertical altitude of the target take-off and landing point and the current flight altitude with the total mass of the UAV, a vertical energy consumption impact index is calculated. The formula used to calculate the vertical energy consumption impact index is as follows:

[0098]

[0099] In the formula, E clbThe vertical energy consumption impact index is given by m, where m is the total mass of the UAV, g is the acceleration due to gravity, and h is the altitude of the target take-off and landing point. f h represents the current altitude of the drone. for η is the predicted maximum height of the obstacle, and η is the efficiency of the UAV rotor motor.

[0100] The vertical energy consumption impact index represents the energy consumed in vertical movement, where the vertical energy consumption impact index E clb The higher the value, the more electricity is consumed.

[0101] In vertical flight, a drone needs to overcome gravity to climb, or control its speed through motors or aerodynamics during descent. During climb, the drone consumes energy to counteract its own weight (i.e., overcome gravity), and this energy is directly related to changes in gravitational potential energy. The altitude difference is calculated by taking the maximum value between the current drone's flight altitude and the predicted maximum obstacle height. Using this maximum difference ensures the drone has sufficient power for a vertical landing. Therefore, the vertical energy consumption impact index is proportional to the difference between the current drone's flight altitude, the predicted maximum obstacle height, and the target takeoff / landing point altitude. The greater the mass, the greater the gravity the drone experiences, and thus the greater the power required for vertical flight. Therefore, the total mass m of the drone is directly proportional to the vertical energy consumption impact index E. clb Proportional.

[0102] The rotor motors of drones are not 100% efficient in converting electrical energy into mechanical energy; there is a certain amount of energy loss. Motor efficiency represents the ratio of consumed electrical energy to usable mechanical energy. Since energy consumption and efficiency are inversely proportional, energy loss is introduced into the formula. The gravitational potential energy consumption is corrected to reflect the actual energy consumption.

[0103] The efficiency η of the UAV rotor motor is obtained by correcting for ambient temperature and current flight altitude. The specific formula used to calculate the efficiency η of the UAV rotor motor is as follows:

[0104]

[0105] In the formula, η0 is the standard efficiency of the UAV rotor motor, T0 is the reference temperature, and T hf β represents the temperature at the current location of the drone, and β is the altitude correction factor.

[0106] Operating temperatures deviating from the reference temperature T0 will affect the operating efficiency of both mechanical and electrical components. The closer the ambient temperature is to the reference temperature T0, the closer the motor's performance is to its ideal state. Therefore, the efficiency η of the UAV rotor motor is related to |T0-T0|. hf|Inversely proportional. The reference temperature is generally taken as 25℃, and the standard efficiency η0 of the drone rotor motor can be obtained from the product parameter manual provided by the vendor.

[0107] Increased altitude leads to decreased air density, reduced cooling capacity, and poorer heat dissipation, resulting in lower motor efficiency. Therefore, the efficiency η of the drone's rotor motor is related to the drone's current flight altitude h. f Inversely proportional.

[0108] The altitude correction factor β is used to represent the impact of different altitudes on the efficiency of the drone's rotor motor. It can be set based on expert experience and is generally between 0.1 and 0.5.

[0109] Based on the vertical energy consumption impact index and the obstacle impact index, the set return-to-home power threshold is corrected to obtain the power threshold affecting return-to-home. The formula used to calculate the power threshold affecting return-to-home is as follows:

[0110]

[0111] In the formula, yz represents the threshold of the return-to-home power consumption at the current moment, IES is the obstacle impact index at the current moment during the drone's flight cruise, yz0 is the set initial threshold of the return-to-home power consumption, and ω1 and ω2 are the weighting coefficients of the obstacle impact index and the vertical energy consumption impact index, respectively, where ω1 < ω2 and both ω1 and ω2 are greater than 0.

[0112] It should be noted that the impact threshold yz of the current moment is represented by the combined obstacle impact index and vertical energy consumption impact index. The larger yz is, the more power needs to be retained for the return landing. It has been stated that both the obstacle impact index and the vertical energy consumption impact index are proportional to the energy consumption. Therefore, both the obstacle impact index and the vertical energy consumption impact index are proportional to the impact threshold yz of the current moment.

[0113] In the form of square root The vertical energy consumption impact index represents the direct proportionality between the vertical energy consumption impact index and the threshold of the return-to-home battery capacity. The square root indicates a non-linear reduction of this impact. The contribution of vertical energy consumption to the return-to-home battery capacity is significant but non-linear because the energy consumption during vertical climb is usually greater than that during horizontal flight. However, during the return-to-home process, the drone may recover some energy through gravitational potential energy during descent (e.g., some motors support energy feedback). Therefore, the impact of vertical energy consumption on the return-to-home battery capacity is not entirely linear.

[0114] IES in the form of square 2 This represents the additional energy consumption during flight caused by external environmental conditions (such as wind speed, obstacle density, and unexpected events), expressed in terms of the squared IES. 2Amplifying this indicates that the contribution of hindering factors to the return power is non-linear, especially under adverse conditions where its impact may be more significant.

[0115] Since vertical energy consumption is mainly related to the height difference between the take-off and landing point and the flight altitude, its value is basically fixed during flight and its impact is relatively stable. However, the impact of obstruction is dynamic and may increase rapidly due to environmental changes, so it has a greater potential impact on energy consumption. Therefore, it needs to be given a higher weight. Hence, ω1 < ω2 and both ω1 and ω2 are greater than 0.

[0116] The initial threshold yz0 for the impact on return-to-home power can be set according to the actual environment and expert experience, and is generally 10%-30%.

[0117] Step 5: Collect the battery health parameters of the drone. Based on the battery health parameters and the ambient temperature, dynamically adjust the return-to-home power threshold to obtain the dynamic return-to-home power threshold. Compare the dynamic return-to-home power threshold with the drone's real-time power level. Based on the comparison results, issue the corresponding decision command. The battery health parameters include battery storage time and number of charge cycles.

[0118] Based on battery health parameters, the threshold affecting the return-to-home battery level is dynamically adjusted to obtain the dynamic return-to-home battery level threshold. The formula used to calculate the dynamic return-to-home battery level threshold is as follows:

[0119]

[0120] In the formula, yz′ is the dynamic return-to-home power threshold, and cs and cj are the battery storage time and the number of charge cycles, respectively.

[0121] During long-term storage, batteries experience capacity decay due to self-discharge and changes in chemical properties (such as electrolyte decomposition and lithium plating on the negative electrode). The longer the storage time, the more significant the capacity loss and the lower the actual usable capacity. Each complete charge-discharge cycle causes chemical reactions in the internal active materials, leading to irreversible capacity reduction. As the number of charge-discharge cycles increases, the total battery capacity gradually decreases, internal resistance increases, and output power decreases. Therefore, it is necessary to increase the dynamic return-to-base capacity threshold to ensure sufficient actual charge for return landing. Thus, the dynamic return-to-base capacity threshold is directly proportional to battery storage time and the number of charge-discharge cycles. Taking into account both the battery's storage time and the number of charge cycles, the square root of the sum of their squares is used to quantify the overall health of the battery.

[0122] Temperature is a crucial factor affecting battery performance. Low temperatures significantly impact battery output power and capacity. Ambient temperature at cruising altitude typically decreases with increasing flight altitude. Low temperatures reduce the battery's chemical reaction rate and degrade discharge performance; therefore, the dynamic return-to-home charge threshold yz′ is inversely proportional to temperature.

[0123] Specifically, the battery storage time refers to the number of days from the first use of the drone battery to the current date, and the number of charge cycles refers to the number of times the drone battery is charged within the number of days from the first use to the current date.

[0124] The dynamic return-to-home battery threshold is compared with the drone's real-time battery level. Based on the comparison result, a corresponding decision command is issued. The specific logic for issuing the corresponding decision command is as follows:

[0125] When ZH≥1.0*yz′, the cruise status is determined to be healthy, indicating that sufficient power is currently available for cruise control;

[0126] When 0.4*yz′≤ZH<1.0*yz′, the cruise mode is determined to be in the middle, and a prompt is issued to pay attention to the remaining battery power;

[0127] When 0≤ZH<0.4*yz′, the cruise status is judged to be low, indicating that the return trip should be initiated immediately.

[0128] ZH represents the current battery level displayed on the drone.

[0129] Please see Figure 2 The present invention also provides a decision-making system for low-battery autonomous landing based on a vertical take-off and landing (VTOL) unmanned aerial vehicle (UAV). This decision-making system is used to execute the aforementioned decision-making method for low-battery autonomous landing based on a VTOL unmanned aerial vehicle, comprising:

[0130] The prediction model training module is used to acquire terrain point cloud data with several known terrain feature parameters as sample point cloud data. Based on the sample point cloud data, a neural network model is established. The sample point cloud data is used as the input of the model, and the corresponding terrain feature parameters are used as labels to train the neural network model to obtain a terrain feature prediction model. The terrain feature parameters include obstacle density and maximum height.

[0131] The cruise altitude confirmation module is used to collect terrain point cloud data of the UAV's flight cruise area, input the terrain point cloud data of the UAV's flight cruise area into the trained terrain feature prediction model, and obtain the predicted values ​​of terrain feature parameters of the UAV's flight cruise area. The predicted values ​​of terrain feature parameters of the UAV's flight cruise area include obstacle density prediction values ​​and maximum height prediction values.

[0132] The environmental disturbance analysis module is used to acquire real-time flight altitude data of the UAV during flight cruise, and at the same time acquire real-time environmental parameters at the current flight altitude and the horizontal distance to the target take-off and landing point. Based on the obtained real-time environmental parameters, the horizontal distance to the take-off and landing point and the predicted value of obstacle density, the obstacle impact index is calculated. The environmental parameters include ambient temperature, wind speed and wind direction angle.

[0133] The vertical energy consumption analysis module is used to obtain the vertical height of the target take-off and landing point. By using the obtained vertical height of the target take-off and landing point and the current flight altitude, combined with the total mass of the UAV, a vertical energy consumption impact index is calculated and generated. Based on the vertical energy consumption impact index and the obstacle impact index, the set return-home power threshold is corrected to obtain the return-home power threshold.

[0134] The landing decision generation module is used to collect the battery health parameters of the UAV, and dynamically correct the return-to-home power threshold based on the battery health parameters and ambient temperature to obtain the dynamic return-to-home power threshold. The dynamic return-to-home power threshold is compared with the real-time power of the UAV, and the corresponding decision command is issued based on the comparison result. The battery health parameters include battery storage time and number of charge cycles.

[0135] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0136] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0137] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0138] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A decision-making method for low-battery autonomous landing based on a vertical takeoff and landing unmanned aerial vehicle (UAV), characterized in that, The specific steps include: A terrain point cloud data with several known terrain feature parameters is obtained as sample point cloud data. A neural network model is established based on the sample point cloud data. The sample point cloud data is used as the input of the model, and the corresponding terrain feature parameters are used as labels to train the neural network model to obtain a terrain feature prediction model. The terrain feature parameters include obstacle density and maximum height. Collect terrain point cloud data of the UAV flight cruise area, input the terrain point cloud data of the UAV flight cruise area into the trained terrain feature prediction model, and obtain the predicted values ​​of terrain feature parameters of the UAV flight cruise area. The predicted values ​​of terrain feature parameters of the UAV flight cruise area include obstacle density prediction value and maximum height prediction value. The system acquires real-time flight altitude data of the UAV during flight cruise, and simultaneously acquires real-time environmental parameters at the current flight altitude and the horizontal distance to the target take-off and landing point. Based on the obtained real-time environmental parameters, the horizontal distance to the take-off and landing point, and the predicted obstacle density, the system calculates the obstacle impact index. The environmental parameters include ambient temperature, wind speed, and wind direction angle. Obtain the vertical height of the target take-off and landing point. Using the obtained vertical height of the target take-off and landing point and the current flight altitude, combined with the total mass of the UAV, calculate and generate the vertical energy consumption impact index. Based on the vertical energy consumption impact index and the obstacle impact index, correct the set return-to-home power threshold to obtain the return-to-home power threshold. The system collects battery health parameters of the drone and dynamically adjusts the return-to-home power threshold based on these parameters and ambient temperature. This dynamic return-to-home power threshold is then compared with the drone's real-time power level. Based on the comparison, a corresponding decision command is issued. The battery health parameters include battery storage time and the number of charge cycles.

2. The decision-making method for low-battery autonomous landing based on a vertical takeoff and landing UAV according to claim 1, characterized in that: Based on sample point cloud data, a neural network model is established. Specifically, a terrain feature prediction model is built using a Long Short-Term Memory (LSTM) network model. Activation functions and optimization algorithms are selected, with the Tanh function chosen as the activation function and Adam as the optimization algorithm for the LSTM model. The formula for the Tanh function is: In the formula, f(r) represents the Tanh function, and the independent variable r represents the weighted sum of the neuron's inputs, that is, the result of the weighted sum of the inputs received by the neuron from the previous layer. Simultaneously, the hyperparameters of the LSTM model are set, including: number of network layers, number of iterations, learning rate, batch size, number of training iterations, number of batches, and number of hidden layer neurons. The network is set to a 3-layer network structure, the number of iterations is set to 200, the learning rate is set to 0.001, the batch size is set to 32, the number of training iterations is set to 100, the batch size is set to 256, and the number of hidden layer neurons is set to 32. The trained terrain feature prediction model takes terrain point cloud data as input and outputs terrain feature parameters corresponding to the terrain.

3. The decision-making method for low-battery autonomous landing based on a vertical takeoff and landing UAV according to claim 2, characterized in that: The obstacle impact index is calculated based on the obtained real-time environmental parameters, the horizontal distance to the take-off and landing point, and the predicted obstacle density. The specific formula for calculating the obstacle impact index is as follows: In the formula, IES represents the obstacle impact index at the current moment, λ is the obstacle density in the flight cruise area, and P wind This represents the wind resistance power consumption experienced by the drone at the current moment, and L represents the horizontal distance between the drone's current position and the take-off and landing point. Wherein the wind resistance power consumption P at the current moment wind The calculations are performed using real-time environmental parameters collected, and the specific calculation formula is as follows: In the formula, ρ is the air density at the current flight altitude, A is the effective frontal area of ​​the UAV at the current moment, and C... d V is the air drag coefficient. eff This indicates the wind speed in the environment at the current moment, at the drone's current flight altitude; The formula used to calculate the effective windward area of ​​the drone at the current moment is as follows: In the formula, θ is the angle between the current wind direction and the flight direction of the UAV, and A on and A bise These are the projected areas from the top-down view and the projected areas from the front view of the drone, respectively.

4. The decision-making method for low-battery autonomous landing based on a vertical takeoff and landing UAV according to claim 1, characterized in that: By combining the vertical altitude of the target take-off and landing point and the current flight altitude with the total mass of the UAV, a vertical energy consumption impact index is calculated. The formula used to calculate the vertical energy consumption impact index is as follows: In the formula, E clb The vertical energy consumption impact index is given by m, where m is the total mass of the UAV, g is the acceleration due to gravity, and h is the altitude of the target take-off and landing point. f h represents the current altitude of the drone. for η is the predicted maximum height of the obstacle, and η is the efficiency of the UAV rotor motor. The efficiency η of the UAV rotor motor is obtained by correcting for ambient temperature and current flight altitude. The specific formula used to calculate the efficiency η of the UAV rotor motor is as follows: In the formula, η0 is the standard efficiency of the UAV rotor motor, T0 is the reference temperature, and T hf β represents the temperature at the current location of the drone, and β is the altitude correction factor.

5. The decision-making method for low-battery autonomous landing based on a vertical takeoff and landing UAV according to claim 4, characterized in that: Based on the vertical energy consumption impact index and the obstacle impact index, the set return-to-home power threshold is corrected to obtain the power threshold affecting return-to-home. The formula used to calculate the power threshold affecting return-to-home is as follows: In the formula, yz represents the threshold of the return-to-home power consumption at the current moment, IES is the obstacle impact index at the current moment during the drone's flight cruise, yz0 is the set initial threshold of the return-to-home power consumption, and ω1 and ω2 are the weighting coefficients of the obstacle impact index and the vertical energy consumption impact index, respectively, where ω1 < ω2 and both ω1 and ω2 are greater than 0.

6. The decision-making method for low-battery autonomous landing based on a vertical takeoff and landing UAV according to claim 5, characterized in that: Based on battery health parameters, the threshold affecting the return-to-home battery level is dynamically adjusted to obtain the dynamic return-to-home battery level threshold. The formula used to calculate the dynamic return-to-home battery level threshold is as follows: In the formula, yz′ is the dynamic return-to-home power threshold, and cs and cj are the battery storage time and the number of charge cycles, respectively. The dynamic return-to-home battery threshold is compared with the drone's real-time battery level. Based on the comparison result, a corresponding decision command is issued. The specific logic for issuing the corresponding decision command is as follows: When ZH≥1.0*yz′, the cruise status is determined to be healthy, indicating that sufficient power is currently available for cruise control; When 0.4*yz′≤ZH<1.0*yz′, the cruise mode is determined to be in the middle, and a prompt is issued to pay attention to the remaining battery power; When 0≤ZH<0.4*yz′, the cruise status is judged to be low, indicating that the return trip should be initiated immediately. ZH represents the current battery level displayed on the drone.

7. A decision-making system for low-battery autonomous landing based on a vertical takeoff and landing unmanned aerial vehicle (UAV), characterized in that: The decision system for low-battery autonomous landing based on a vertical takeoff and landing (VTOL) UAV is used to execute the decision method for low-battery autonomous landing based on a VTOL UAV as described in any one of claims 1-6, comprising: The prediction model training module is used to acquire terrain point cloud data with several known terrain feature parameters as sample point cloud data. Based on the sample point cloud data, a neural network model is established. The sample point cloud data is used as the input of the model, and the corresponding terrain feature parameters are used as labels to train the neural network model to obtain a terrain feature prediction model. The terrain feature parameters include obstacle density and maximum height. The cruise altitude confirmation module is used to collect terrain point cloud data of the UAV's flight cruise area, input the terrain point cloud data of the UAV's flight cruise area into the trained terrain feature prediction model, and obtain the predicted values ​​of terrain feature parameters of the UAV's flight cruise area. The predicted values ​​of terrain feature parameters of the UAV's flight cruise area include obstacle density prediction values ​​and maximum height prediction values. The environmental disturbance analysis module is used to acquire real-time flight altitude data of the UAV during flight cruise, and at the same time acquire real-time environmental parameters at the current flight altitude and the horizontal distance to the target take-off and landing point. Based on the obtained real-time environmental parameters, the horizontal distance to the take-off and landing point and the predicted value of obstacle density, the obstacle impact index is calculated. The environmental parameters include ambient temperature, wind speed and wind direction angle. The vertical energy consumption analysis module is used to obtain the vertical height of the target take-off and landing point. By using the obtained vertical height of the target take-off and landing point and the current flight altitude, combined with the total mass of the UAV, a vertical energy consumption impact index is calculated and generated. Based on the vertical energy consumption impact index and the obstacle impact index, the set return-home power threshold is corrected to obtain the return-home power threshold. The landing decision generation module is used to collect the battery health parameters of the UAV, and dynamically correct the return-to-home power threshold based on the battery health parameters and ambient temperature to obtain the dynamic return-to-home power threshold. The dynamic return-to-home power threshold is compared with the real-time power of the UAV, and the corresponding decision command is issued based on the comparison result. The battery health parameters include battery storage time and number of charge cycles.

Images