An agricultural unmanned aerial vehicle and power facility conflict prediction method and system

Abstract

The application discloses an agricultural unmanned aerial vehicle and power facility conflict prediction method and system, and relates to the technical field of intelligent operation, and comprises the following steps: acquiring environment perception data of a target area, outputting a flight prediction trajectory set through a time series prediction model and in combination with dynamic constraint conditions of an unmanned aerial vehicle; performing multi-scale risk analysis on the flight prediction trajectory set, and dynamically adjusting a flight trajectory; performing hierarchical avoidance control on the flight trajectory, and outputting a control instruction sequence of the unmanned aerial vehicle; and the application solves the problems of unmanned aerial vehicle trajectory deviation and collision risk through multi-scale risk prediction and hierarchical control.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent operation, more specifically, the present application relates to a method and system for predicting conflicts between agricultural unmanned aerial vehicles and power facilities. BACKGROUND

[0002] With the rapid development of modern agriculture, unmanned aerial vehicles are increasingly used in crop spraying, plant protection, inspection and monitoring, etc. Unmanned aerial vehicles have gradually become an important tool for intelligent agriculture due to their high flexibility, high efficiency and low cost. However, in actual application, especially in farmland environments close to power facilities, unmanned aerial vehicle operations face serious safety risks. Traditional agricultural unmanned aerial vehicles mostly rely on manual remote control or pre-set fixed flight paths for operation, and have limited ability to identify and avoid dynamic changes in the environment, sudden obstacles and high-risk facilities such as power lines, which can easily result in trajectory deviation, collision or low operation efficiency.

[0003] For example, the invention patent with publication number CN120255548A discloses a system and method for optimizing the flight trajectory of an unmanned aerial vehicle based on artificial intelligence and the Internet of Things, which relates to the field of unmanned aerial vehicle control. The method includes: collecting communication delay, environmental data and unmanned aerial vehicle state in real time; using Markov chain to analyze historical communication delay data to predict the basic delay value in the near future; correcting the basic delay value through a machine learning model; outputting the final delay prediction result; calculating the delay risk index to determine the delay risk level; based on the delay risk level, when the delay risk level is low, the control command is sent in advance according to the final delay prediction result; when the delay risk level is medium or high, a pre-trained reinforcement learning strategy library is called to generate adaptive control commands. The invention constructs a multi-dimensional risk index and provides multiple strategy adjustments to the flight trajectory based on real-time data and risk index to improve the stability of unmanned aerial vehicle flight.

[0004] For example, the invention patent with publication number CN117434967B discloses a method, system, medium and device for detecting collisions of unmanned aerial vehicles, which relates to the field of collision detection. The method includes: obtaining the inherent flight parameters and flight trajectory of the unmanned aerial vehicle, determining the first influence area affecting the flight trajectory based on the inherent flight parameters; determining the target collision object in the first influence area and obtaining the motion information of the target collision object; determining the second influence area of the target collision object on the flight trajectory in the first influence area based on the motion information; calculating the risk score value of the target collision object on the flight of the unmanned aerial vehicle based on the motion information and the second influence area; adjusting the flight trajectory based on the risk score value to obtain the target flight trajectory, and controlling the unmanned aerial vehicle to fly according to the target flight trajectory. Thus, the collision risk in the process of unmanned aerial vehicle flight is reduced.

[0005] The above-disclosed technical solutions have at least the following technical problems:

[0006] In traditional technical solutions, the flight path planning of agricultural drones in complex power facility environments relies on static maps or simple obstacle avoidance algorithms, which cannot fully consider the dynamic constraints of drones and the uncertainty of future trajectories, resulting in low trajectory prediction accuracy and a high risk of deviation or collision.

[0007] To address the above problems, this invention proposes a solution. Summary of the Invention

[0008] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method and system for predicting conflicts between agricultural drones and power facilities, which solves the problems of drone trajectory deviation and collision risk through multi-scale risk prediction and hierarchical control.

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

[0010] A method for predicting conflicts between agricultural drones and power facilities includes: acquiring environmental perception data of the target area; using a time series prediction model and combining it with the dynamic constraints of the drone to output a set of predicted flight trajectories; performing multi-scale risk analysis on the set of predicted flight trajectories and dynamically adjusting the flight trajectories; performing hierarchical avoidance control on the flight trajectories and outputting a sequence of control commands for the drone.

[0011] In a preferred embodiment, the step of acquiring environmental perception data of the target area, using a time series prediction model, and combining the dynamic constraints of the UAV to output a set of predicted flight trajectories is as follows: acquiring real-time position, speed, and attitude data of the UAV and constructing a local environmental point cloud; matching the local environmental point cloud data with a preset power line GIS database to align the perception results with the power grid geographic information, thereby constructing an environmental state model and extracting task constraints; generating several trajectories of the UAV for any future time period using a time series prediction method based on the environmental state model, forming a trajectory candidate set; applying dynamic constraints to the trajectory candidate set to obtain the final set of predicted flight trajectories.

[0012] In a preferred embodiment, the dynamic constraint on the candidate trajectory set to obtain the final flight prediction trajectory set is specifically as follows: Dynamic constraints are applied to the candidate trajectory set using six-degree-of-freedom equations of motion to filter and obtain an initial trajectory set that meets physical feasibility requirements; the initial trajectory is recursively predicted using a particle filtering method to obtain the state mean and covariance matrix at each time step; a trajectory confidence interval is constructed based on the state mean and covariance matrix; the trajectory confidence interval is repeatedly discretely sampled to generate a set of boundary points for each time step, obtaining the set of boundary points of the trajectory on the confidence interval; all time step boundary point sets are connected to form the three-dimensional trajectory envelope of each trajectory; the three-dimensional geometry of the power facility is obtained. The model is used to discretize the power facilities into a set of obstacles composed of line segments, surfaces, or voxel meshes. The set of boundary points of the trajectory confidence interval is compared with the set of obstacles. The minimum spatial distance between the boundary points and the power facilities is output through the nearest neighbor search algorithm, and the closest approach time of the trajectory is recorded. The trajectory risk is output by linearly weighting the minimum spatial distance and the closest approach time. Based on the propeller power model and battery discharge curve, the energy consumption of each initial trajectory is output, and the task deviation of each trajectory is output according to the trajectory deviation and operation coverage. Based on the task deviation and trajectory risk, the final set of flight prediction trajectories is obtained by screening with the goal of minimizing the task deviation and the trajectory risk.

[0013] In a preferred embodiment, the multi-scale risk analysis of the flight prediction trajectory set and the dynamic adjustment of the flight trajectory are specifically as follows: The trajectory envelope boundary points of the flight prediction trajectory are obtained, and the future state probability distribution of the trajectory envelope boundary points is predicted using a recurrent neural network to generate a probability trajectory set for each time step; the probability trajectory set is divided into short-term and medium-term time scales, and the risk of each time scale is output, including short-term trajectory risk and medium-term trajectory risk; each boundary point within the trajectory confidence interval is assigned a probability weight using a Gaussian approximation algorithm to obtain the trajectory envelope probability distribution; the short-term trajectory risk score of the probability trajectory set is monitored in real time, and when the short-term trajectory risk score exceeds a threshold, an immediate collision avoidance action is triggered, and a short-term trajectory set is generated based on the trajectory envelope probability distribution and wind field data; the overall trajectory trend is analyzed based on the medium-term trajectory risk score of the probability trajectory set, conflict areas and task deviation areas are identified, and the medium-term trajectory risk score is combined with task constraints to generate a medium-term trajectory set using an MPC algorithm; the short-term trajectory set and the medium-term trajectory set are dynamically weighted and fused, and the final feasible trajectory is selected based on the probability weighting of the trajectory confidence interval, outputting the final dynamically adjusted trajectory set.

[0014] In a preferred embodiment, the output of risk at each time scale is specifically as follows: Predict several future time steps for each trajectory to generate a sequence of trajectory points; for each trajectory point, detect whether an obstacle has entered a set safe zone; if an obstacle enters the safe zone, mark a risk event at that trajectory point and count the number of risk events occurring on each trajectory within a short period; based on the number and distribution of risk events, output a short-term trajectory risk score using an event density method based on spatial distribution; merge the short-term trajectory risk scores according to the trajectory order to form a continuous short-term risk sequence; divide the continuous short-term risk sequence into multiple medium-term time periods according to the medium-term time range; for each medium-term time period, count the number and distribution of short-term risk events included, and based on the cumulative number and distribution of short-term events, output a medium-term trajectory risk score using a nearest-event time method based on time distance.

[0015] In a preferred embodiment, the step of generating a short-term trajectory set based on the trajectory envelope probability distribution and wind field data is as follows: Short-term wind field data is acquired, and the wind field is estimated using a Gaussian process regression method to obtain the wind speed distribution and its uncertainty covariance; within a short timescale, the wind field is modeled as a wind-induced displacement distribution of the UAV's position, obtaining the expected value and covariance of the displacement distribution; the trajectory envelope probability distribution and the wind-induced displacement distribution are convolved to obtain the wind-coupled trajectory distribution; Monte Carlo sampling is performed on the wind-coupled trajectory distribution to generate a set of trajectory points disturbed by wind; the disturbed trajectory points are subjected to nearest neighbor search with a set of power facility obstacles to obtain the minimum spatial distance and the closest approach time; a weighted average of the minimum spatial distance and the closest approach time is calculated based on sample weights to obtain the expected safe distance and the expected approach time, and the collision probability is obtained; a wind-coupled risk function is constructed using a weighted method based on the expected safe distance, the expected approach time, and the collision probability; based on the wind-coupled risk function, the candidate trajectory set is clustered and sorted, prioritizing the retention of high-probability, low-risk trajectories to form the final short-term trajectory set.

[0016] In a preferred embodiment, the step of combining the mid-term trajectory risk score with task constraints and generating a mid-term trajectory set through the MPC algorithm is as follows: obtaining the centerline of each trajectory in the flight prediction trajectory set and using the centerline as the reference trajectory input for the MPC algorithm to generate several executable candidate mid-term trajectories; comparing the candidate mid-term trajectories with historical flight trajectories for trajectory similarity and eliminating mismatched trajectories; and sorting the selected mid-term trajectories according to the task constraints to form a mid-term trajectory set.

[0017] In a preferred embodiment, the hierarchical avoidance control of the flight trajectory and the output of the UAV control command sequence are specifically as follows: A hierarchical control architecture is established, comprising a strategic layer, a tactical layer, and an execution layer; when the execution layer performs avoidance maneuvers, safety constraints are followed; if control commands that meet the safety constraints cannot be generated, an emergency strategy is triggered, and the execution layer sends the safety status and emergency handling results back to the tactical and strategic layers; the control commands generated by the execution layer are smoothed and rate-limited; if an emergency avoidance or upper-level command requires immediate change, the smoothing process is allowed to be interrupted, and a new executable command is immediately issued, with the adjusted control command serving as the final output command for the current cycle; the final execution layer control commands for each cycle are organized into a control command sequence in chronological order, and the control command sequence is sent to the flight controller for execution in real time.

[0018] In a preferred embodiment, the hierarchical control architecture includes a strategic layer, a tactical layer, and an execution layer, as follows: The strategic layer reads the final flight trajectory, selects a mid-term reference trajectory segment for the current time window, and selects alternative trajectories based on mission priority and remaining energy, then distributes the mid-term trajectory segment and mission constraints to the tactical layer; the tactical layer receives the mid-term reference trajectory segment and mission constraint information from the strategic layer, makes local adjustments to the mid-term trajectory based on the short-term risk score, and generates a short-term reference point sequence; if the mid-term trajectory is unavailable, a rollback request is sent to the strategic layer; the execution layer receives the short-term reference point sequence and performs preset operations in each control cycle.

[0019] A system for predicting conflicts between agricultural drones and power facilities includes an output module, an adjustment module, and a control module, which are interconnected. The output module acquires environmental perception data of the target area, uses a time series prediction model, and combines the dynamic constraints of the drone to output a set of predicted flight trajectories. The adjustment module performs multi-scale risk analysis on the set of predicted flight trajectories and dynamically adjusts the flight trajectory. The control module performs hierarchical avoidance control on the flight trajectory and outputs a sequence of control commands for the drone.

[0020] The technical effects and advantages of the present invention, a method and system for predicting conflicts between agricultural drones and power facilities, are as follows:

[0021] 1. This invention achieves dynamic optimization of flight trajectories and conflict prevention through multi-scale risk analysis. On a short-term timescale, the system utilizes a recurrent neural network to predict the future state probability distribution of trajectory envelope boundary points, combining wind field data and a Gaussian approximation algorithm to achieve probabilistic modeling of the trajectory envelope. Through Monte Carlo sampling and minimum spatial distance calculation, potential collision risks can be identified in real time, dynamically generating a short-term trajectory set and executing immediate collision avoidance actions. On a medium-term timescale, the system uses a model predictive control (MPC) algorithm to combine medium-term trajectory risk scoring with task constraints, generating a medium-term trajectory set that meets task priority and remaining energy requirements. This dynamic weighted fusion of short-term and medium-term trajectories not only ensures the UAV's rapid response to sudden environmental changes but also optimizes the overall operational path, balancing safety and mission efficiency, and enabling stable flight of the UAV in complex power facility environments. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the process for predicting conflicts between agricultural drones and power facilities according to the present invention.

[0023] Figure 2 This is a schematic diagram of the system structure of a method for predicting conflicts between agricultural drones and power facilities according to the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0025] Example 1, Figure 1 This invention provides a method for predicting conflicts between agricultural drones and power facilities, comprising:

[0026] S1: Acquire environmental perception data of the target area, and output a set of predicted flight trajectories through a time series prediction model and in combination with the dynamic constraints of the UAV.

[0027] In this embodiment, environmental perception data of the target area is acquired, and a set of predicted flight trajectories is output through a time series prediction model combined with the dynamic constraints of the UAV, as detailed below:

[0028] The UAV acquires real-time position, velocity and attitude data through GNSS / RTK and IMU, and performs multi-source data fusion to reduce noise and drift.

[0029] Power facilities are perceived using airborne lidar, millimeter-wave radar, or stereo vision sensors, and local environmental point clouds are constructed.

[0030] The local environmental point cloud data is matched with the preset power line GIS database to align the perception results with the power grid geographic information, thereby constructing an environmental state model and extracting task constraints, including operation coverage, operation area range and remaining battery capacity.

[0031] Based on the environmental state model, several trajectories of UAVs for any future time period are generated using time series prediction methods (such as recurrent neural networks (RNN) or Bayesian filtering methods), forming a trajectory candidate set.

[0032] Dynamic constraints are applied to the candidate trajectory set to obtain the final set of predicted flight trajectories.

[0033] In this embodiment, dynamic constraints are applied to the candidate trajectory set to obtain the final set of predicted flight trajectories, as detailed below:

[0034] By applying dynamic constraints to candidate trajectories using six-degree-of-freedom motion equations, an initial set of trajectories that meet physical feasibility is obtained through screening. The dynamic constraints include velocity, acceleration, rate of climb, turning radius, and load limits.

[0035] The initial trajectory is recursively predicted by particle filtering to obtain the mean and covariance matrix of the state at each time step.

[0036] The trajectory confidence interval is constructed based on the state mean and covariance matrix. The trajectory confidence interval is repeatedly discretely sampled to generate the boundary point set for each time step, thus obtaining the boundary point set of the trajectory on the confidence interval.

[0037] Connect all time step boundary points to form a three-dimensional trajectory envelope for each trajectory, which is used to characterize the uncertainty range of the predicted trajectory;

[0038] A three-dimensional geometric model of the power facility is obtained, and the power facility is discretized into a set of obstacles composed of line segments, curved surfaces or voxel meshes. The three-dimensional geometric model includes the location of the transmission tower, the height of the conductor, and the shape of the tension arc.

[0039] The set of boundary points of the trajectory confidence interval is compared with the set of obstacles. The minimum spatial distance between the boundary points and the power facilities is output through the nearest neighbor search algorithm, and the time point when the minimum spatial distance is generated is recorded as the closest approach time of the trajectory.

[0040] The trajectory risk is output by using a linear weighting method to combine the minimum spatial distance and the closest approach time.

[0041] Based on the propeller power model and battery discharge curve, the energy consumption of each initial trajectory is output, and the task deviation of each trajectory is output according to the trajectory deviation and the operation coverage.

[0042] Based on mission deviation and trajectory risk, the final set of flight prediction trajectories is obtained by screening with the goal of minimizing mission deviation and trajectory risk.

[0043] The confidence interval is as follows:

[0044]

[0045] In the formula: This is the chi-square distribution threshold, corresponding to a 95% confidence level, used to control the size of the trajectory confidence interval. and This represents the actual position and state variables of the UAV at a certain predicted time step. and This represents the predicted mean (center trajectory point) of the drone at the same time step. It is the transpose of the matrix. It is the state covariance matrix, used to describe the uncertainty of the predicted state. It is the inverse of the covariance matrix, used to normalize the deviations in the state space to standardized units.

[0046] S2 performs multi-scale risk analysis on the flight prediction trajectory set and dynamically adjusts the flight trajectory.

[0047] In this embodiment, multi-scale risk analysis is performed on the flight prediction trajectory set, and the flight trajectory is dynamically adjusted, as follows:

[0048] Obtain the trajectory envelope boundary points of the flight prediction trajectory, and predict the future state probability distribution of the trajectory envelope boundary points through a recurrent neural network to generate the probability trajectory edge set for each time step.

[0049] The probabilistic trajectory set is divided into a short-term time scale (1–2 seconds, used to capture immediate collision risks) and a medium-term time scale (5–10 seconds, used to assess flight path trends and mission continuity), and the risk of each time scale is output, including short-term trajectory risk and medium-term trajectory risk.

[0050] Each boundary point within the trajectory confidence interval is assigned a probability weight using the Gaussian approximation algorithm to obtain the trajectory envelope probability distribution.

[0051] The short-term trajectory risk score of the probability trajectory set is monitored in real time. When the short-term trajectory risk score exceeds the threshold, an immediate collision avoidance action is triggered, and a short-term trajectory set is generated based on the trajectory envelope probability distribution and wind field data.

[0052] Based on the mid-term trajectory risk score analysis of the probability trajectory set, the overall trajectory trend is analyzed, potential conflict areas and task deviation areas are identified, and the mid-term trajectory risk score is combined with task constraints (job coverage, energy consumption budget) to generate a mid-term trajectory set through the MPC algorithm.

[0053] The short-term trajectory set and the medium-term trajectory set are dynamically weighted and fused, and the final feasible trajectory is selected based on the probability weight of the trajectory confidence interval, and the final dynamically adjusted trajectory set is output.

[0054] In this embodiment, the risk at each time scale is output as follows:

[0055] For each trajectory, predict several future time steps to generate a sequence of trajectory points;

[0056] For each trajectory point, check whether the obstacles detected by the sensors have entered the set safety range;

[0057] If an obstacle enters the safe zone, a "risk event" is marked at that trajectory point, and the number of risk events occurring on each trajectory within a short period of time is counted.

[0058] Based on the number and distribution of risk events, a short-term trajectory risk score is output using the spatial distribution-based event density method.

[0059] Short-term trajectory risk scores are merged according to trajectory order to form a continuous short-term risk sequence;

[0060] The continuous short-term risk sequence is divided into multiple medium-term time periods according to the medium-term time range (e.g., 5–10 seconds);

[0061] For each medium-term period, the number and distribution of short-term risk events are counted. Based on the cumulative number and distribution of short-term events, the medium-term trajectory risk score is output using the nearest event time method based on time distance.

[0062] In this embodiment, a short-term trajectory set is generated based on the trajectory envelope probability distribution and wind field data, as follows:

[0063] Short-term wind field data, including gridded wind speed and direction information, are obtained, and the wind field is estimated by Gaussian process regression, Kriging interpolation or extended Kalman filter method to obtain the wind speed distribution and its uncertainty covariance.

[0064] In the short timescale, the wind field is modeled as the wind-induced displacement distribution of the UAV's location, and the expected value and covariance of the displacement distribution are obtained.

[0065] The wind-coupled trajectory distribution is obtained by convolving the trajectory envelope probability distribution with the wind-induced displacement distribution.

[0066] Monte Carlo sampling is performed on the wind-coupled trajectory distribution to generate a set of trajectory points affected by wind disturbance;

[0067] By performing a nearest neighbor search on the trajectory points after wind disturbance and the set of power facility obstacles, the minimum spatial distance and the closest approach time are obtained;

[0068] The minimum spatial distance and the closest approach time are weighted and averaged based on the sample weights to calculate the expected safe distance and the expected approach time, and then the collision probability is obtained. The collision probability is determined based on the safety threshold conditions (safe distance threshold, time threshold) to determine whether the sample constitutes a collision event. The probability weights of the trajectories determined to be collisions are accumulated to obtain the collision probability.

[0069] Based on the expected safe distance, expected approach time, and collision probability, a wind-coupled risk function is constructed using a weighted method.

[0070] Based on the wind-coupled risk function, the candidate trajectory set is clustered and sorted, and high-probability, low-risk trajectories are retained first to form the final short-term trajectory set.

[0071] In this embodiment, the mid-term trajectory risk score is combined with task constraints, and a mid-term trajectory set is generated using the MPC algorithm, as detailed below:

[0072] The centerline of each trajectory in the flight prediction trajectory set is obtained and used as the reference trajectory input for the MPC algorithm to generate several executable candidate intermediate trajectories. The MPC algorithm automatically adjusts the trajectory according to the UAV dynamics constraints and the task constraints.

[0073] The candidate intermediate trajectory is compared with the historical flight trajectory (distance between trajectory points, similarity of path shape), and the mismatched trajectory is removed;

[0074] The selected intermediate trajectories are sorted according to the task constraints to form an intermediate trajectory set.

[0075] S3 performs layered avoidance control on the flight trajectory and outputs a sequence of control commands for the UAV.

[0076] In this embodiment, the flight trajectory is subjected to layered avoidance control, and the control command sequence of the UAV is output as follows:

[0077] Establish a hierarchical control architecture, which includes a strategic layer, a tactical layer, and an execution layer;

[0078] When performing avoidance maneuvers at the execution layer, safety constraints (such as maximum G-load, minimum flight altitude, and minimum spatial distance) must be observed. If control commands that meet the safety constraints cannot be generated, the execution layer will trigger emergency strategies, such as hovering, deceleration and return to base, or safe landing. The execution layer will then transmit the safety status and emergency handling results back to the tactical and strategic layers for trajectory adjustment or mission plan updates.

[0079] The control commands generated by the execution layer are processed smoothly and their rate is limited to ensure that command changes are stable and can be executed by the UAV.

[0080] If an emergency avoidance or an upper-level instruction requires an immediate change, the process is allowed to be interrupted and smoothed out. New executable instructions are issued immediately, and the smoothed or immediately adjusted control instructions are used as the final output instructions for the current cycle.

[0081] The final execution layer control commands for each cycle are organized into a control command sequence in chronological order, and the control command sequence is sent to the flight controller for execution in real time. The control command sequence includes thrust, angular velocity, position / velocity target and trigger source identifier.

[0082] In this embodiment, a hierarchical control architecture is established, which includes a strategic layer, a tactical layer, and an execution layer, as detailed below:

[0083] The strategic layer reads the final flight trajectory, selects the intermediate reference trajectory segment for the current time window, and selects alternative trajectories based on mission priority and remaining energy. The intermediate trajectory segment and mission constraints are then distributed to the tactical layer.

[0084] The tactical layer receives the mid-term reference trajectory segment and mission constraint information from the strategic layer, and performs local replacements or adjustments to the mid-term trajectory based on the short-term risk score to generate a short-term reference point sequence.

[0085] If the intermediate trajectory is unavailable, a rollback or replacement request is sent to the strategy layer;

[0086] The execution layer receives short-term reference point sequences and real-time sensor / status information, and performs preset operations in each control cycle.

[0087] The preset operations include:

[0088] Read the current position, velocity, and attitude information;

[0089] Detect real-time risk events and determine whether to trigger an avoidance maneuver;

[0090] Short-term control targets (position / velocity / attitude) are generated based on the execution layer reference point and avoidance actions.

[0091] Convert the control target into angular velocity and thrust commands that the flight controller can execute;

[0092] Control commands are sent to the flight controller via the interface.

[0093] Example 2, Figure 2 The present invention provides a system for predicting conflicts between agricultural drones and power facilities, comprising an output module, an adjustment module, and a control module, with connections between the modules;

[0094] The output module is used to acquire environmental perception data of the target area, and output a set of predicted flight trajectories through a time series prediction model and in combination with the dynamic constraints of the UAV.

[0095] The adjustment module is used to perform multi-scale risk analysis on the flight prediction trajectory set and dynamically adjust the flight trajectory.

[0096] The control module is used to perform layered avoidance control of the flight trajectory and output the control command sequence of the UAV.

[0097] 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.

[0098] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0099] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in 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. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0100] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0101] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0102] Finally, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting conflicts between agricultural drones and power facilities, characterized in that, include: The system acquires environmental perception data of the target area, uses a time series prediction model, and combines the dynamic constraints of the UAV to output a set of predicted flight trajectories. Perform multi-scale risk analysis on the flight prediction trajectory set and dynamically adjust the flight trajectory; The flight trajectory is subjected to layered avoidance control, and a sequence of control commands for the UAV is output. The process involves acquiring environmental perception data of the target area, using a time series prediction model, and combining it with the dynamic constraints of the UAV to output a set of predicted flight trajectories, as detailed below: Acquire real-time position, velocity, and attitude data of the UAV, and construct a local environmental point cloud; The local environmental point cloud data is matched with the preset power line GIS database to align the perception results with the power grid geographic information, thereby constructing an environmental state model and extracting task constraints. Based on the environmental state model, several trajectories of UAVs for any future time period are generated using time series prediction methods, forming a trajectory candidate set; Dynamic constraints are applied to the candidate trajectory set to obtain the final set of predicted flight trajectories; The dynamic constraints are applied to the candidate trajectory set to obtain the final set of predicted flight trajectories, as detailed below: By applying dynamic constraints to the candidate trajectory set using six-degree-of-freedom equations of motion, an initial trajectory set that satisfies physical feasibility is obtained through screening. The initial trajectory is recursively predicted by particle filtering to obtain the mean and covariance matrix of the state at each time step. The trajectory confidence interval is constructed based on the state mean and covariance matrix. The trajectory confidence interval is repeatedly discretely sampled to generate the boundary point set for each time step, thus obtaining the boundary point set of the trajectory on the confidence interval. Connect all time step boundary points to form the three-dimensional trajectory envelope of each trajectory; Obtain a three-dimensional geometric model of the power facility and discretize the power facility into a set of obstacles composed of line segments, surfaces, or voxel meshes; The set of boundary points of the trajectory confidence interval is compared with the set of obstacles. The minimum spatial distance between the boundary points and the power facilities is output through the nearest neighbor search algorithm, and the closest approach time of the trajectory is recorded. The trajectory risk is output by using a linear weighting method to combine the minimum spatial distance and the closest approach time. Based on the propeller power model and battery discharge curve, the energy consumption of each initial trajectory is output, and the task deviation of each trajectory is output according to the trajectory deviation and the operation coverage. Based on mission deviation and trajectory risk, the final set of flight prediction trajectories is obtained by screening with the goal of minimizing mission deviation and trajectory risk.

2. The method for predicting conflicts between agricultural drones and power facilities according to claim 1, characterized in that, The process of performing multi-scale risk analysis on the flight prediction trajectory set and dynamically adjusting the flight trajectory is as follows: Obtain the trajectory envelope boundary points of the flight prediction trajectory, and predict the future state probability distribution of the trajectory envelope boundary points through a recurrent neural network to generate the probability trajectory edge set for each time step. The set of probability trajectories is divided into short-term and medium-term time scales, and the risk of each time scale is output, including short-term trajectory risk and medium-term trajectory risk. Each boundary point within the trajectory confidence interval is assigned a probability weight using the Gaussian approximation algorithm to obtain the trajectory envelope probability distribution. The short-term trajectory risk score of the probability trajectory set is monitored in real time. When the short-term trajectory risk score exceeds the threshold, an immediate collision avoidance action is triggered, and a short-term trajectory set is generated based on the trajectory envelope probability distribution and wind field data. Based on the mid-term trajectory risk score of the probability trajectory set, the overall trajectory trend is analyzed, conflict areas and task deviation areas are identified, and the mid-term trajectory risk score is combined with task constraints to generate a mid-term trajectory set through the MPC algorithm. The short-term trajectory set and the medium-term trajectory set are dynamically weighted and fused, and the final feasible trajectory is selected based on the probability weight of the trajectory confidence interval, and the final dynamically adjusted trajectory set is output.

3. The method for predicting conflicts between agricultural drones and power facilities according to claim 2, characterized in that, The risk output for each time scale is as follows: For each trajectory, predict several future time steps to generate a sequence of trajectory points; For each trajectory point, detect whether the obstacle has entered the set safe range; If an obstacle enters the safe zone, a risk event is marked at that trajectory point, and the number of risk events occurring on each trajectory within a short period of time is counted. Based on the number and distribution of risk events, a short-term trajectory risk score is output using the spatial distribution-based event density method. Short-term trajectory risk scores are merged according to trajectory order to form a continuous short-term risk sequence; The continuous short-term risk sequence is divided into multiple medium-term time periods according to the medium-term time range; For each medium-term period, the number and distribution of short-term risk events are counted. Based on the cumulative number and distribution of short-term events, the medium-term trajectory risk score is output using the nearest event time method based on time distance.

4. The method for predicting conflicts between agricultural drones and power facilities according to claim 3, characterized in that, The short-term trajectory set is generated based on the trajectory envelope probability distribution and wind field data, as detailed below: Short-term wind field data were acquired, and the wind field was estimated using the Gaussian process regression method to obtain the wind speed distribution and its uncertainty covariance. In the short timescale, the wind field is modeled as the wind-induced displacement distribution of the UAV's location, and the expected value and covariance of the displacement distribution are obtained. The wind-coupled trajectory distribution is obtained by convolving the trajectory envelope probability distribution with the wind-induced displacement distribution. Monte Carlo sampling is performed on the wind-coupled trajectory distribution to generate a set of trajectory points affected by wind disturbance; By performing a nearest neighbor search on the trajectory points after wind disturbance and the set of power facility obstacles, the minimum spatial distance and the closest approach time are obtained; The minimum spatial distance and the closest approach time are weighted and averaged based on the sample weights to calculate the expected safe distance and the expected approach time, and the collision probability is obtained. Based on the expected safe distance, expected approach time, and collision probability, a wind-coupled risk function is constructed using a weighted method. Based on the wind-coupled risk function, the candidate trajectory set is clustered and sorted, and high-probability, low-risk trajectories are retained to form the final short-term trajectory set.

5. The method for predicting conflicts between agricultural drones and power facilities according to claim 4, characterized in that, The process of combining mid-term trajectory risk scoring with task constraints and generating a mid-term trajectory set using the MPC algorithm is as follows: Obtain the centerline of each trajectory in the flight prediction trajectory set, and use the centerline as the reference trajectory input for the MPC algorithm to generate several executable candidate intermediate trajectories; The candidate intermediate trajectory is compared with the historical flight trajectory for similarity, and the mismatched trajectory is eliminated; The selected intermediate trajectories are sorted according to the task constraints to form an intermediate trajectory set.

6. The method for predicting conflicts between agricultural drones and power facilities according to claim 5, characterized in that, The layered avoidance control of the flight trajectory outputs a sequence of control commands for the UAV, as follows: Establish a hierarchical control architecture, which includes a strategic layer, a tactical layer, and an execution layer; When performing avoidance actions at the execution layer, safety constraints must be followed. If control instructions that meet the safety constraints cannot be generated, an emergency strategy is triggered, and the execution layer sends the safety status and emergency handling results back to the tactical and strategic layers. The control instructions generated by the execution layer are smoothed and rate-limited. If an emergency avoidance or higher-level instructions require immediate changes, a smooth interruption is permitted, and new executable instructions are immediately issued. The adjusted control instructions are then used as the final output instructions for the current cycle. The final execution layer control commands for each cycle are organized into a control command sequence in chronological order, and the control command sequence is sent to the flight controller for execution in real time.

7. The method for predicting conflicts between agricultural drones and power facilities according to claim 6, characterized in that, The establishment of a layered control architecture, comprising a strategic layer, a tactical layer, and an execution layer, is detailed below: The strategic layer reads the final flight trajectory, selects the intermediate reference trajectory segment for the current time window, and selects alternative trajectories based on mission priority and remaining energy. The intermediate trajectory segment and mission constraints are then distributed to the tactical layer. The tactical layer receives the mid-term reference trajectory segment and mission constraint information from the strategic layer, makes local adjustments to the mid-term trajectory based on the short-term risk score, and generates a short-term reference point sequence. If the intermediate trajectory is unavailable, a rollback request is sent to the strategy layer; The execution layer receives a short-term reference point sequence and performs preset operations in each control cycle.

8. A system using the conflict prediction method between agricultural drones and power facilities as described in any one of claims 1-7, characterized in that, It includes an output module, an adjustment module, and a control module, and the modules are interconnected. The output module is used to acquire environmental perception data of the target area, and output a set of predicted flight trajectories through a time series prediction model and in combination with the dynamic constraints of the UAV. The adjustment module is used to perform multi-scale risk analysis on the flight prediction trajectory set and dynamically adjust the flight trajectory. The control module is used to perform layered avoidance control of the flight trajectory and output the control command sequence of the UAV.

Images