Unmanned aerial vehicle autonomous return control method in interference environment based on multi-source information

Abstract

This invention belongs to the field of unmanned aerial vehicle (UAV) autonomous control technology, and relates to an autonomous return-to-home control method for UAVs under interference environments based on multi-source information. It acquires raw observation data from multiple sensors and the UAV's self-checked motion state in real time, constructs a navigation reliability time-series evaluation set for uncertainty quantification, generates a confidence space boundary characterizing the potential deviation range of the UAV's position and velocity within a future preset time window, and adaptively constructs the UAV's return-to-home channel by combining the return-to-home target point location and real-time environmental obstacle information. It also integrates the UAV's motion laws to establish an objective function with dynamically adjusted weights based on the uncertainty quantification results, solves for the return-to-home control command sequence with the future preset time window as the optimized time domain, and executes it cyclically until return-to-home. This achieves adaptive coordination between the return-to-home channel and control strategy under interference environments, helping to improve the environmental adaptability and mission reliability of the UAV during the return-to-home process in complex interference scenarios.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicle (UAV) autonomous control technology, and relates to a method for autonomous return-to-home control of UAVs under interference environments based on multi-source information. Background Technology

[0002] Autonomous return-to-home functionality is a core technology module ensuring flight safety for drones and is crucial for their stable operation in complex scenarios. In actual operations, when drones encounter satellite navigation signal blockage, multipath interference, or severe electromagnetic disturbances, traditional satellite-based return-to-home modes are prone to failure, potentially leading to drone loss of contact, crashes, and other safety incidents. Therefore, overcoming the bottlenecks in positioning and navigation technology under complex environments to achieve high-precision, high-reliability autonomous return-to-home is a critical technical challenge that the drone industry urgently needs to address.

[0003] In existing technologies, methods for improving return-to-home reliability under interference environments mainly focus on path planning and initial positioning optimization. For example, the return-to-home control method, device, chip, and UAV disclosed in Chinese Patent Publication No. CN119987433A, by comprehensively utilizing satellite positioning and multi-sensor information after the UAV takes off, delays calculation and back-derives a high-precision initial takeoff point position, aiming to solve the problem of large errors in the return-to-home reference point caused by signal instability during the takeoff phase, thereby improving the landing accuracy of the return-to-home endpoint.

[0004] Another Chinese patent publication, CN117369494A, describes an autonomous return method, device, and drone for unmanned aerial vehicles (UAVs). It focuses on dynamic obstacle avoidance during the return process. By recording the flight trajectory and setting key points, it can reverse-search historical key points as a reference when an obstacle is detected, and dynamically adjust local segments of the return path to achieve obstacle avoidance.

[0005] While the aforementioned methods improve return-to-home performance in certain aspects, their control logic still suffers from inherent defects in existing technologies, making it difficult to adapt to complex interference environments where navigation performance dynamically declines in real time. Specifically: 1. Existing technologies lack an adaptive mechanism for real-time decline in navigation performance. Their control logic typically assumes that the tracking accuracy of the return-to-home path is continuously reliable. Their path planning and tracking control are not deeply coupled with the real-time performance indicators of the navigation system. When environmental interference causes dynamic decline in positioning and velocity measurement accuracy, they cannot detect and adjust the control strategy, continuing to track according to the original accuracy assumption. This makes them highly susceptible to collisions due to excessive actual control errors.

[0006] 2. Existing technologies, when performing obstacle avoidance or path adjustment, typically set a fixed width for the safety passage of the UAV, or simply expand it based on the obstacle's geometry. This fails to reflect the fluctuations in navigation uncertainty caused by changes in the strength of interference. In areas with strong interference, a fixed safety boundary may be insufficient to accommodate increased positioning drift. In areas with weak interference, an excessively large boundary may lead to an overly conservative path, affecting return-to-home efficiency. Summary of the Invention

[0007] In view of this, in order to solve the problems mentioned in the background technology, a method for autonomous return-to-home control of UAVs under interference environment based on multi-source information is proposed.

[0008] The objective of this invention can be achieved through the following technical solution: This invention provides an autonomous return-to-home control method for unmanned aerial vehicles (UAVs) under interference conditions based on multi-source information, including: S1. Real-time acquisition of raw observation data from multi-source sensors and the UAV's self-check motion state, and construction of a navigation reliability time series evaluation set.

[0009] S2. The uncertainty of the navigation confidence time series evaluation set is quantified to generate a confidence space boundary that characterizes the potential deviation range of the UAV's position and speed within a future preset time window.

[0010] S3. Based on the return-to-home target location, real-time environmental obstacle information, and the confidence space boundary, construct the UAV's return-to-home channel through adaptive rules.

[0011] S4. Based on the spatial boundary defined by the UAV return channel, and combined with the UAV motion law, establish an objective function with dynamically adjusted weights based on the uncertainty quantification results. The objective function includes at least the UAV channel centerline deviation term and the remaining flight distance term, and solves the return control command sequence with a future preset time window as the optimization time domain.

[0012] S5. Execute the immediate control commands in the return control command sequence, update all input information in the next control cycle, and repeat the S1-S4 process until the return target point is reached.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. The present invention constructs the state estimation covariance inside the navigation system into a navigation reliability time series evaluation set, and quantifies it into a confidence space boundary that characterizes the potential deviation range of future position and speed, thereby determining the return channel of the UAV including the safe passage radius of each path node, realizing the direct linkage between control strategy and navigation uncertainty, solving the problem that the existing safe channel is fixed and cannot adapt to the dynamic changes of interference, and helping to improve the UAV's adaptive capability to interference during return.

[0014] 2. This invention establishes an objective function that dynamically adjusts the weights based on the uncertainty quantification results. By acquiring position uncertainty and velocity uncertainty indices in real time, it dynamically determines the weights of the channel centerline deviation term and the remaining range term, effectively balancing the contradiction between channel maintenance and range advancement. Furthermore, by combining the objective function with the constraint optimization solution process, it helps to ensure the physical feasibility and spatial safety of the return control command sequence. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating the implementation steps of the method of the present invention.

[0017] Figure 2 This is a schematic diagram illustrating the process of generating an initial return path corresponding to a future preset time window using heuristic search, as described in this invention.

[0018] Figure 3 This is a schematic diagram illustrating the process of optimizing the solution of the return control command sequence according to the present invention. Detailed Implementation

[0019] 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 skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1 As shown, the present invention provides an autonomous return-to-home control method for UAVs under interference environment based on multi-source information, including: S1. Real-time acquisition of raw observation data from multi-source sensors and UAV self-check motion status, and construction of navigation reliability time series evaluation set.

[0021] Specifically, the process of constructing a navigation reliability time series evaluation set includes the following steps: First, the raw observation data from multiple sources such as inertial measurement units, global navigation satellite system receivers, and visual sensors are timestamped and aligned with the coordinate system to form fused observation data with consistent spatiotemporal reference, providing standardized input for subsequent state fusion.

[0022] Then, based on the spatiotemporally aligned fused observation data, existing state estimation algorithms such as Kalman filtering are used for fusion processing to output the UAV's current state estimation vector and its corresponding covariance matrix. The state estimation vector includes at least the UAV's position and velocity in three-dimensional space, and may further include state parameters such as attitude angles. The covariance matrix quantifies the uncertainty of this state estimation, with its main diagonal elements representing the variance of the self-estimated state variables, such as the X, Y, and Z axis position and velocity.

[0023] Simultaneously, a set of parameters characterizing the motion state of the UAV body are monitored in real time. These parameters include at least the body angular rate, linear acceleration amplitude, and flight altitude change rate. The real-time monitoring values ​​of each parameter are compared with the calibration range determined by the UAV's operational technical specifications.

[0024] If the real-time monitoring value of any parameter continuously deviates from its calibration range in multiple consecutive sampling periods, or if its deviation from the calibration range exceeds the preset tolerance limit in a single sampling period, the parameter is determined to be in an abnormal state, and an abnormal flag bit corresponding to the parameter is generated.

[0025] Anomaly flags are captured to clarify the type of the current abnormal state. Then, based on the physical correlation between this anomaly type and the various dimensions of the UAV state estimation, specific main diagonal elements in the covariance matrix that are directly related to the uncertainties in the three-dimensional position and three-dimensional velocity estimations are identified. The percentage of monitored values ​​deviating from the calibration range or the proportion of their duration is used as the degree of anomaly. The values ​​of these specific main diagonal elements are proportionally amplified, thereby proactively reflecting the negative impact of aircraft anomalies on navigation state reliability at the data level.

[0026] As an example, within the framework of using Kalman filtering or its variants as the state estimation algorithm, anomalous states can be characterized by dynamically increasing the corresponding elements in the process noise covariance matrix. Specifically, the anomalous type is mapped to the dimension representing the corresponding dynamic disturbance in the process noise matrix. For example, angular rate anomalies are mapped to angular velocity process noise, and linear acceleration anomalies are mapped to linear acceleration process noise.

[0027] The quantized anomaly level is used as a direct scaling factor on the base noise variance value of the mapped dimension in the process noise covariance matrix, thereby amplifying the variance value proportionally. According to the recursive formula of Kalman filtering, the increased process noise will directly cause specific main diagonal elements of the state estimation covariance matrix to be amplified in the prediction step.

[0028] Furthermore, from the covariance matrix after the aforementioned amplification process, all diagonal elements corresponding to the three-dimensional position and three-dimensional velocity are extracted. These variance values ​​are then combined to form a multi-dimensional vector, which serves as the navigation confidence feature vector for the current moment. This feature vector collectively represents the confidence level of the current estimation of the UAV's spatial motion state.

[0029] Finally, the navigation reliability feature vector at the current moment is combined with the navigation reliability feature vectors from the previous N consecutive control cycles, arranged in ascending order of time, to form a time-series data set of length N+1, namely the navigation reliability time-series evaluation set. This time-series set will be used to analyze the evolution trend of navigation reliability over time.

[0030] This completes the construction of the current cycle navigation confidence time series evaluation set, which not only contains the latest state estimation uncertainty information, but also performs adaptive correction through self-checking state, and retains the historical change sequence, providing a data foundation for subsequent uncertainty quantification.

[0031] S2. The uncertainty of the navigation confidence time series evaluation set is quantified to generate a confidence space boundary that characterizes the potential deviation range of the UAV's position and speed within a future preset time window.

[0032] Specifically, the uncertainty quantification of the navigation reliability time-series evaluation set includes the following process: For each time step arranged chronologically in the navigation reliability time-series evaluation set, extract its position variance sequence and velocity variance sequence respectively. Perform exponentially weighted moving average calculations on the position variance sequence and velocity variance sequence to obtain a comprehensive position uncertainty index and a comprehensive velocity uncertainty index. In the exponentially weighted moving average calculation, variance data closer to the current time are given higher weights.

[0033] Based on the aforementioned position uncertainty index and velocity uncertainty index, and combined with the maximum controllable acceleration constraint of the UAV, worst-case displacement deviation envelope prediction is performed. Through kinematic envelope calculation, the maximum displacement deviation values ​​of the UAV in the X, Y, and Z coordinate axes at the end of the future preset time window are obtained respectively. The unified kinematic envelope calculation formula for the coordinate axis directions is as follows:

[0034] ;

[0035] in, This represents the maximum positional deviation of the drone along the current coordinate axis from the current moment until the end of a preset future time window. , These represent the components of the position uncertainty index and the velocity uncertainty index along the current coordinate axis, respectively. This indicates the duration of a future preset time window. This represents the maximum controllable acceleration of the drone in the current coordinate axis direction.

[0036] It describes the cumulative effect of velocity uncertainty over time and calculates the maximum additional displacement caused solely by the uncertainty of the initial velocity within a future preset time window.

[0037] This represents the displacement component generated by the maximum controllable acceleration. This term assumes that the UAV will always accelerate in the direction of the current coordinate axis in a direction that deviates from the expected trajectory within a preset time window in the future, resulting in the ultimate displacement increment.

[0038] This calculation formula uses a linear summation form and is a conservative design based on the worst-case assumption. It assumes that the three adverse factors—position error, velocity error, and active acceleration disturbance—occur simultaneously and in the same direction, jointly causing the most severe deviation.

[0039] The maximum displacement deviation along each coordinate axis is taken as twice the length of the three principal axes of the ellipsoid in three-dimensional space. The center of the ellipsoid is set at the current position of the UAV. The coordinates of the center position of the ellipsoid, the lengths of the three principal axes, and the direction vectors of the three principal axes are used as the boundary parameters of the ellipsoid.

[0040] The ellipsoid boundary parameters are transformed into a local coordinate system along the return channel, aligning the three principal axes of the ellipsoid with the longitudinal, lateral, and vertical directions of the return channel. The semi-axis lengths of the ellipsoid along each channel axis are extracted to form the spatial boundary width vector, which is used as the confidence spatial boundary. The specific implementation process is as follows:

[0041] First, establish a local coordinate system for the return-to-home channel: take the projection point of the UAV's current position on the center line of the return-to-home channel as the origin, and define the unit tangent vector of the initial return-to-home path at the origin as the longitudinal axis of the local coordinate system; define the projection unit vector of the reference axis perpendicular to the horizontal plane in the global coordinate system onto the plane perpendicular to the longitudinal axis as the vertical axis of the local coordinate system; perform a cross product operation between the longitudinal axis and the vertical axis, and define the resulting unit vector as the transverse axis of the local coordinate system.

[0042] Secondly, construct the coordinate rotation transformation matrix: arrange the unit vectors corresponding to the longitudinal axis, transverse axis and vertical axis of the local coordinate system of the return channel as column vectors in order to form a rotation transformation matrix from the global coordinate system to the local coordinate system of the return channel.

[0043] Next, the principal axis directions of the ellipsoid are transformed: the direction vectors of each principal axis of the ellipsoid are multiplied by the transpose of the rotation transformation matrix to obtain the direction representation of each principal axis in the local coordinate system of the return channel, thereby obtaining the geometric parameters of the transformed ellipsoid.

[0044] Finally, calculate the semi-axis length of the channel axis: calculate the coordinates of the intersection points of the ellipsoid with the longitudinal axis, transverse axis and vertical axis of the local coordinate system of the return channel respectively. The absolute value of the coordinates of the intersection points is the semi-axis length of the ellipsoid in the corresponding channel axis. Combine the semi-axis lengths of the longitudinal, transverse and vertical axes in sequence to form the spatial boundary width vector.

[0045] S3. Based on the return-to-home target location, real-time environmental obstacle information, and the confidence space boundary, construct the UAV's return-to-home channel through adaptive rules.

[0046] Specifically, constructing a return path for a drone using adaptive rules includes the following steps: constructing a 3D grid map based on real-time environmental obstacle information, and marking impassable areas in the 3D grid map.

[0047] Using the return target point as the endpoint and the current drone position as the starting point, a predetermined spatial range is defined for the path search in this period. The axis of this range is based on the direction from the starting point to the endpoint, and its length is theoretically determined by the product of the drone's current speed and the preset future time window, in order to match the dynamic programming cycle.

[0048] Subsequently, this calculated length is compared with the perception range of the interference intensity signal from the UAV's onboard interference monitoring equipment: if the calculated length exceeds the perception range, the path length is corrected to the limit of the perception range to ensure that the cost assessment of all path points is based on actually perceptible data. Otherwise, the original calculated length is retained. This step ensures that the path planning matches the perception capability on a spatial scale.

[0049] It should be explained that the interference intensity signal perception range refers to the maximum effective detection distance specified in the technical specifications of the UAV airborne interference monitoring equipment, which is a fixed value that is pre-stored in a cloud database and can be directly accessed.

[0050] Reference Figure 2 As shown, within the predetermined spatial range, an initial return path corresponding to a future preset time window is generated through heuristic search. The specific implementation process includes: assigning a basic passage cost to each passable voxel within the predetermined spatial range in a 3D raster map.

[0051] The interference intensity signals that the UAV can perceive in all directions within the predetermined space range are used as the measured intensity values ​​of the corresponding directional rays at the current position of the UAV. The interference intensity signals come from the airborne electromagnetic environment monitoring sensor, and its output is the quantified value of the electromagnetic field intensity in each direction.

[0052] For each voxel within the predetermined spatial range, the following assignment process is performed: determine the orientation of the voxel, associate it with the ray closest to that orientation, and obtain the measured intensity value of that ray at the current position of the UAV as a reference benchmark.

[0053] Starting from the current position of the UAV, extend the ray outwards. Based on the distance of a voxel from the starting point along this extension line, spatial extrapolation is performed on a reference benchmark according to a preset extrapolation rule to estimate the interference intensity at that voxel. The preset extrapolation rule is: along this extension line, the farther the voxel is from the starting point, the stronger its estimated interference intensity; the closer the voxel is to the starting point, the weaker its estimated interference intensity, and at the starting point, it equals the reference benchmark.

[0054] The pre-defined extrapolation rule is based on a reasonable engineering assumption: the interference source is located at the far end of the ray, and the signal strength increases with proximity to the interference source. As a specific implementation, the sum of the voxel's distance from the ray to the starting point and the reference distance can be divided by the reference distance. This ratio is then multiplied by a reference distance to obtain the estimated interference intensity for the corresponding voxel. The reference distance is a preset constant used to control the extrapolation slope.

[0055] By performing linear least squares fitting on the position variance and velocity variance of the navigation reliability time series evaluation set, the slope of the position variance change and the slope of the velocity variance change are obtained. The sum of the two is used as a global trend factor to characterize the overall deterioration trend of navigation uncertainty.

[0056] The estimated interference intensity for each voxel is min-max normalized with the global trend factor to eliminate dimensional differences. The normalized interference intensity and the global trend factor are then weighted and multiplied according to preset weights to obtain the comprehensive navigation risk coefficient for each voxel within a predetermined spatial range. This comprehensive navigation risk coefficient is used as navigation reliability prediction information. The preset weights can be set as follows: the weight of the interference intensity is 0.7, and the weight of the trend factor is 0.3.

[0057] Add the comprehensive navigation risk coefficient to 1, then multiply by the basic passage cost to obtain the passage cost after dynamic adjustment of each voxel.

[0058] A path search algorithm is run on a 3D raster map after the toll cost is adjusted. The path search algorithm is an existing technology and can be exemplarily adopted as the A algorithm or the D Lite algorithm to find the path with the minimum total toll cost within a predetermined spatial range, which is used as the initial return path corresponding to a future preset time window.

[0059] The generated initial return path is enhanced with channelization for safety. A series of discrete path nodes are extracted along the path, and the safe passage radius at each path node is determined based on the confidence space boundary.

[0060] The safe passage radius of adjacent path nodes can be smoothed by interpolation, for example, by using linear interpolation to form a continuously changing channel cross section.

[0061] All channel sections are spliced ​​together along the initial return path to generate the return channel.

[0062] This invention constructs the state estimation covariance within the navigation system into a navigation reliability time series evaluation set, and quantifies it into a confidence space boundary characterizing the potential deviation range of future position and velocity. This determines the return-to-home channel for the UAV, including the safe passage radius of each path node, and realizes the direct linkage between control strategy and navigation uncertainty. This solves the problem that existing safe channels are fixed and cannot adapt to dynamic changes in interference, and helps to improve the UAV's adaptive capability to interference during return-to-home.

[0063] S4. Based on the spatial boundary defined by the UAV return channel, and combined with the UAV motion law, establish an objective function with dynamically adjusted weights based on the uncertainty quantification results. The objective function includes at least the UAV channel centerline deviation term and the remaining flight distance term, and solves the return control command sequence with a future preset time window as the optimization time domain.

[0064] Reference Figure 3 As shown, the specific execution process of solving the return control command sequence is as follows: the return control command sequence is defined as a control quantity covering the future preset time window and a series of time nodes, and the control quantity includes at least thrust command and attitude angle command.

[0065] The current speed of the drone is used as the expected forward speed maintained by the initial return path. The product of the expected forward speed and the span of each time node within the future preset time window relative to the current time is used as the expected return displacement. The current position of the drone is superimposed to locate the expected flight position of each time node within the future preset time window on the initial return path. The tangent direction of the initial return path at the expected flight position is used as the expected flight direction.

[0066] The attitude angle command at each time point is initialized to an attitude that makes the UAV nose point to the expected flight direction, and the aerodynamic drag coefficient and reference area calibrated under the initialized attitude are read from the configuration database pre-stored by the UAV.

[0067] The aerodynamic drag required for the UAV to maintain its expected forward speed in its initial attitude is calculated using the aerodynamic drag formula. Simultaneously, the gravitational component of the UAV in the expected flight direction is obtained through multiplication based on the UAV's total mass, gravitational acceleration, and the sine value corresponding to the pitch angle in the initial attitude. The aerodynamic drag formula is as follows:

[0068] ;

[0069] In the formula, To calculate the aerodynamic drag, To preset atmospheric density, This is the aerodynamic drag coefficient. For reference area, The expected forward speed.

[0070] The aerodynamic drag formula is a fundamental and well-known formula in the fields of fluid mechanics and aerodynamics. It belongs to the existing technology in this field, and its specific meaning will not be elaborated here.

[0071] The sum of aerodynamic drag and gravity components is used as the baseline thrust required for the drone to maintain its expected forward speed.

[0072] Multiply the reference thrust value by the unit vector of the expected flight direction to obtain the initial guess of the thrust vector at the corresponding time node.

[0073] Arrange the initial guesses of thrust vectors and attitude angles at all time points in chronological order to form the initial guesses of the return control command sequence.

[0074] The simulation process of command execution is used to obtain the Euclidean distances from the arrival position of the UAV at each time point to the center line of the return channel and the end point of the initial return path, and these distances are marked as the deviation from the UAV channel center line and the remaining flight distance, respectively.

[0075] Based on the current position and velocity uncertainty indicators of the UAV, weights are assigned to the UAV channel centerline deviation and remaining range. Specifically, the determination process involves setting upper limits for the normal operating range of each position and velocity uncertainty indicator. These upper limits are determined through statistical analysis based on long-term historical operational data of the UAV navigation system under a calibration environment without strong interference. Specifically, navigation data generated by the UAV during long-term operation in a known good signal environment is collected. The statistical distributions of position and velocity estimation errors are extracted from this data. Error boundary values ​​that meet a pre-set confidence level (e.g., 95%), or three times the standard deviation of this distribution, are set as the upper limits for the normal operating range of the corresponding indicators. This value reflects the accuracy limit achievable by the UAV navigation system under ideal or typical operating conditions, providing a quantitative benchmark for determining whether abnormal interference has occurred.

[0076] The position uncertainty index and velocity uncertainty index obtained in the current period are divided by their corresponding upper limits of normal operating range to obtain the normalized position uncertainty coefficient and velocity uncertainty coefficient.

[0077] Set basic weights for the channel centerline deviation and remaining range items, and the sum of the two basic weights is 1.

[0078] Based on the normalized uncertainty coefficient, the basic weights are dynamically adjusted: the preset position weight adjustment gain is multiplied by the position uncertainty coefficient and then added to the basic weights corresponding to the channel centerline deviation term to obtain the initial allocation weights of the channel centerline deviation term.

[0079] The initial weighting of the remaining range term is obtained by subtracting the preset speed weight adjustment gain and the speed uncertainty coefficient from the base weight corresponding to the remaining range term.

[0080] The preset position weight adjustment gain and preset speed weight adjustment gain determine the sensitivity and strength of the normalized uncertainty index's influence on the weights. Their values ​​are obtained through offline simulation debugging. In the UAV dynamics model and the interference environment simulation calibrated based on historical data of the target operating area, a normalized comprehensive evaluation function is established with the system's safety boundary and return-home time performance under interference as the optimization objective. This function is the sum of the return-home time and the safety boundary violation risk. The safety boundary risk can be quantified by the ratio of the distance depth of the probability region of trajectory intrusion into obstacles in the statistical simulation to the preset safety distance. When the return-home time is substituted into the function calculation, it needs to be normalized by first calculating its ratio with the ideal return-home time without interference. Specifically, the ideal return-home time without interference is obtained by dividing the simulated return-home path length by the UAV's design standard return-home speed.

[0081] The parameter sweep finds a set of position weight adjustment gain values ​​and velocity weight adjustment gain values ​​that minimize the output value of the comprehensive evaluation function. These are then used as preset constants for direct use. The parameter sweep range is the gain value, which can be exemplarily from 0.1 to 2.0, with a step size of 0.1.

[0082] The initial allocation weights of the channel centerline deviation term and the remaining range term are normalized to obtain the final allocation weights.

[0083] The logic behind assigning weights to the drone's deviation from the channel centerline and remaining range is as follows: a high positional uncertainty index indicates a blurred understanding of the drone's instantaneous geometric position, with collision risk primarily stemming from positioning drift. Therefore, the weight of the deviation from the channel centerline needs to be increased to generate enhanced space-keeping control commands, bringing the drone as close as possible to the channel centerline.

[0084] A high speed uncertainty index indicates a vague understanding of the UAV's motion state, making aggressive speed control commands prone to dynamic instability. Therefore, it is necessary to reduce the weight of the remaining range term to make the control strategy more gradual, avoiding the risk of oscillations or runaway due to a mismatch between control commands and actual dynamic responses.

[0085] An objective function for evaluating the quality of the return-to-home control command sequence is constructed by linear weighted fusion, wherein the smaller the output value of the objective function, the better the return-to-home control command sequence.

[0086] Based on the initial guess value, through iterative sampling, a solution set containing multiple candidate control variables is generated for each time node within a future preset time window.

[0087] For each time point, calculate the objective function output value corresponding to each candidate control variable in its solution set, and sort them accordingly.

[0088] From the solution set at each time point, select the candidate control variable with the smallest objective function output value, and combine them in chronological order to form the first candidate control instruction sequence.

[0089] Verify whether the first candidate sequence satisfies the multiple constraints consisting of the rate of change of the control variable and the channel boundary.

[0090] The multiple constraints include: A. Thrust change rate constraint: limiting the maximum allowable change in thrust command within a unit control cycle, which is mainly based on the maximum climb / descent thrust response rate calibration of the UAV power system.

[0091] B. Attitude angle change rate constraint: limits the maximum allowable change in attitude angle command within a unit control cycle, which is mainly based on the bandwidth of the UAV attitude control loop and the angular acceleration calibration that the airframe structure can withstand.

[0092] C. It requires that when the UAV executes a sequence of control commands, every point on its entire predicted trajectory must be located within the return-to-home corridor space.

[0093] If the verification is successful, the result will be output as a sequence of return control commands for the corresponding future preset time window.

[0094] Otherwise, identify the key time points that violate the constraints, and in the solution set of that point, select the candidate control quantity whose objective function output value is suboptimal to replace it, recombine and verify, until a return control command sequence that satisfies all constraints is obtained.

[0095] It should be further added that if a return-to-home control command sequence that meets the constraints is not obtained within the preset maximum number of iterations, a degraded control strategy is triggered. The degraded control strategy is a preset two-stage emergency procedure: first, the UAV is controlled to enter a hovering state and continuously attempts to resolve within a first preset duration (e.g., 10 seconds); if the first preset duration expires without success, the UAV is controlled to descend vertically at a constant safe speed.

[0096] This invention establishes an objective function that dynamically adjusts weights based on uncertainty quantification results. By acquiring position uncertainty and velocity uncertainty indices in real time, the weights of the channel centerline deviation term and the remaining range term are dynamically determined, effectively balancing the contradiction between channel maintenance and range advancement. Furthermore, by combining the objective function with the constraint optimization solution process, it helps ensure the physical feasibility and spatial safety of the return control command sequence.

[0097] S5. Execute the immediate control commands in the return control command sequence, update all input information in the next control cycle, and repeat the S1-S4 process until the return target point is reached.

[0098] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, and all such modifications and additions should fall within the protection scope of the present invention.

Claims

1. A method for autonomous return-to-home control of unmanned aerial vehicles (UAVs) under interference environments based on multi-source information, characterized in that, include: S1. Acquire raw observation data from multiple sensors and the UAV's self-check motion status in real time to construct a navigation reliability time series evaluation set; S2. Quantify the uncertainty of the navigation confidence time series evaluation set to generate a confidence space boundary that characterizes the potential deviation range of the UAV's position and speed within a future preset time window; S3. Based on the return-to-home target location, real-time environmental obstacle information, and the confidence space boundary, construct the UAV's return-to-home channel using adaptive rules; S4. Based on the spatial boundary defined by the UAV return-to-home channel, and combined with the motion law of the UAV, establish an objective function that dynamically adjusts the weights by the uncertainty quantification result. The objective function includes at least the UAV channel centerline deviation term and the remaining flight distance term. And with a future preset time window as the optimization time domain, solve the return-to-home control command sequence. S5. Execute the immediate control commands in the return control command sequence, update all input information in the next control cycle, and repeat the S1-S4 process until the return target point is reached. Uncertainty quantification is performed on the navigation confidence time series evaluation set, including: The position variance and velocity variance of each time step in the navigation reliability time series evaluation set are weighted and averaged separately. The weights decrease exponentially with the distance of the time step from the current time to obtain the current position uncertainty index and velocity uncertainty index of the UAV. Based on position uncertainty and velocity uncertainty indices, and combined with the maximum controllable acceleration constraint of the UAV, the maximum displacement deviation of the UAV in each coordinate axis direction within a future preset time window is predicted. Define the boundary parameters of the ellipsoid in three-dimensional space based on the maximum displacement deviation in each coordinate axis direction; The ellipsoid boundary parameters are converted into a spatial boundary width vector in the local coordinate system along the return channel to obtain the confidence spatial boundary. The process of constructing the UAV's return-to-home route using adaptive rules includes: A 3D grid map is constructed based on real-time environmental obstacle information, and impassable areas are marked in the 3D grid map; Using the return target point as the endpoint and the current drone position as the starting point, an initial return path corresponding to a future preset time window is generated through heuristic search within a predetermined spatial range based on the direction from the starting point to the endpoint. For each path node along the initial return path, determine the safe passage radius at each path node based on the confidence space boundary; The safe passage radius of adjacent path nodes is interpolated and smoothed to form a continuously changing channel cross section; All channel sections are spliced ​​together along the initial return path to generate the return channel; The process of generating an initial return path corresponding to a future preset time window through heuristic search includes: Assign a basic passage cost to each accessible voxel within a predetermined spatial range in a 3D raster map; The system acquires the interference intensity signals in all directions that the UAV can perceive within the predetermined space range, and combines the changes in the position variance and velocity variance of the navigation reliability time series evaluation to generate navigation reliability prediction information for each voxel within the predetermined space range. Based on the navigation reliability prediction information, the passage cost of each voxel is dynamically adjusted; A path search algorithm is run on the 3D raster map after the toll cost adjustment to find the path with the minimum total toll cost within a predetermined spatial range, which is used as the initial return path corresponding to a future preset time window.

2. The method for autonomous return-to-home control of a UAV under interference environment based on multi-source information as described in claim 1, characterized in that, The construction of the navigation reliability time series evaluation set includes: Spatiotemporal alignment of raw observation data from multiple sensors is performed to generate the current state estimation vector and its covariance matrix of the UAV. Based on the monitoring results of the UAV's self-check motion status, the main diagonal elements in the covariance matrix that represent the uncertainty of the three-dimensional position and three-dimensional velocity estimation are identified as being associated with the current abnormal state type, and the values ​​of the main diagonal elements are amplified according to the degree of abnormality. The diagonal elements corresponding to the three-dimensional position and three-dimensional velocity in the amplified covariance matrix are extracted as the confidence feature vector at the current moment. The current credibility feature vector is arranged in chronological order together with the credibility feature vectors of several consecutive historical moments to form a navigation credibility time series evaluation set.

3. The method for autonomous return-to-home control of UAVs under interference environment based on multi-source information as described in claim 2, characterized in that, The monitoring results of the UAV's self-check motion status are obtained through the following process: Real-time monitoring of a set of parameters characterizing the motion state of the UAV body, including at least the body angular rate, linear acceleration amplitude, and flight altitude change rate; The real-time monitoring values ​​of each parameter are compared with the calibration range determined by the UAV's operational technical specifications. If the real-time monitoring value of any parameter continuously deviates from its calibration range in multiple consecutive sampling periods, or if its deviation from the calibration range exceeds the preset tolerance limit in a single sampling period, the parameter is determined to be in an abnormal state, and an abnormal flag bit corresponding to the parameter is generated.

4. The method for autonomous return-to-home control of a UAV under interference environment based on multi-source information as described in claim 1, characterized in that, The predetermined spatial range is determined by its length along the route in the following manner: The initial path length is calculated by multiplying the current speed of the UAV by a preset future time window. The initial path length is then constrained within the range of the UAV's interference intensity signal perception to determine the final path length.

5. The method for autonomous return-to-home control of a UAV under interference environment based on multi-source information as described in claim 1, characterized in that, The process of solving the return control command sequence includes: The return control command sequence is defined as a series of control quantities covering the future preset time window, and the control quantities include at least thrust commands and attitude angle commands. Based on the initial return path, generate initial guesses for the return control command sequence; The simulation command execution process is used to obtain the Euclidean distances from the arrival position of the UAV at each time point to the center line of the return channel and the end point of the initial return path, and these distances are marked as the UAV channel center line deviation term and the remaining flight distance term, respectively. Based on the current position uncertainty index and velocity uncertainty index of the UAV, weights are assigned to the UAV channel centerline deviation term and remaining range term, and an objective function is constructed to evaluate the quality of the return control command sequence. Based on the initial guess and the objective function, a sequence of return control commands that satisfy preset constraints is obtained through optimization.

6. The method for autonomous return-to-home control of a UAV under interference environment based on multi-source information as described in claim 5, characterized in that, The initial guess values ​​for generating the return control command sequence include: The current speed of the drone is used as the expected forward speed maintained by the initial return path. This is used to locate the expected flight position at each time node within the future preset time window, and the tangent direction of the initial return path at the expected flight position is used as the expected flight direction. The attitude angle command at each time point is initialized to the attitude that makes the drone's nose point in the expected flight direction; Based on the aerodynamic characteristics of the UAV in its initial attitude, the baseline thrust required to maintain the expected forward speed is calculated. The baseline thrust value is decomposed along the expected flight direction to obtain the initial guess of the thrust vector at the corresponding time node; Arrange the initial guesses of thrust vectors and attitude angles at all time points in chronological order to form the initial guesses of the return control command sequence.

7. The method for autonomous return-to-home control of a UAV under interference environment based on multi-source information as described in claim 5, characterized in that, The process of obtaining the return control command sequence that satisfies preset constraints through optimization includes: Based on the initial guess value, through iterative sampling, a solution set containing multiple candidate control variables is generated for each time node within a future preset time window; For each time point, calculate the objective function output value corresponding to each candidate control variable in its solution set, and sort them accordingly; From the solution set at each time point, select the candidate control variable with the smallest objective function output value, and combine them in time order to form the first candidate control instruction sequence; Verify whether the first candidate sequence satisfies the multiple constraints consisting of the rate of change of the control variable and the channel boundary; If the verification is satisfied, it is output as the return control command sequence for the current cycle; Otherwise, identify the key time points that violate the constraints, and in the solution set of that point, select the candidate control quantity whose objective function output value is suboptimal to replace it, recombine and verify, until a return control command sequence that satisfies all constraints is obtained.

Images