Flexible connection unmanned aerial detection method, device and system
By employing grid division, static compensation, and dynamic obstacle avoidance algorithms, the obstacle avoidance adaptability problem of UAV-borne detection systems in complex terrains was solved, enabling safe and efficient execution of detection missions and ensuring the safety of both the UAV and the payload.
Patent Information
- Application Number
- CN202611602257.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-16
- Publication Date
- 2026-11-17
AI Technical Summary
Existing UAV-borne detection systems cannot effectively adapt to sudden terrain changes in dynamic real-time obstacle avoidance technology, leading to safety hazards such as collisions between flexible suspension systems and terrain obstacles. This is especially true in mountainous and rugged terrain scenarios, where the closer the payload is to the ground, the greater the risk.
By dividing and marking the operational area of the detection mission into grids, determining the safe expansion radius, constructing a grid model, performing path planning and static compensation, pre-calculating the sling swing and flight path offset, updating the flight plan, and combining Kalman filtering and multi-weighted fusion cost functions, offline pre-calibration and dynamic emergency obstacle avoidance are achieved.
It enables ground-following flight closer to the ground in complex terrain, improving detection capabilities, ensuring flight safety, reducing the computing load and response latency risks of airborne chips, and enhancing the stability and adaptability of the flexible sling system.
Smart Images

Figure CN123111297A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of unmanned aerial vehicle (UAV) aerial detection technology, and in particular to a flexible connection UAV-borne detection method, device, and system. Background Technology
[0002] Multi-rotor UAV sling systems are widely used in cargo transportation, emergency rescue, and outdoor operations due to their strong compatibility with load shapes and high operational flexibility. Obstacle avoidance safety and trajectory tracking accuracy have become the core technology research directions for this system.
[0003] Unmanned aerial vehicle (UAV)-borne detection systems used for scientific research are limited by current and signal strength, requiring the detection payload to be very close to the ground. However, this close proximity greatly increases flight risks. To address the obstacle avoidance challenges of these UAV-borne detection systems, various control methods have been proposed in the existing technology. One typical solution is a flexible obstacle avoidance control method for UAV-borne systems based on artificial potential fields. This method overcomes the limitations of traditional obstacle avoidance designs that only target the individual UAV by establishing a nonlinear coupled dynamic model of the quadcopter UAV-borne system that considers external interference. It constructs the UAV, rigging, and payload as a whole into a "capsule-shaped" safety zone. The repulsive force of the artificial potential field designed based on this zone can accurately calculate the effective collision avoidance distance between the system and obstacles. At the same time, by constructing a self-adjusting auxiliary system to dynamically relax the preset performance function boundary, and integrating the repulsive force of the artificial potential field with unconstrained variables to synthesize a flexible obstacle avoidance control signal, the system can safely and smoothly avoid obstacles during dynamic obstacle avoidance, while also taking into account high-precision trajectory tracking in the obstacle-free phase. This effectively improves the autonomous obstacle avoidance capability and mission reliability of the borne system in general complex environments.
[0004] The aforementioned dynamic obstacle avoidance control methods still have limitations in applicable scenarios. Their core reliance on real-time perception, computation, and control during UAV flight places high demands on the UAV's environmental perception speed, data processing efficiency, and control command response speed. Furthermore, the flexible suspension connection between the UAV and payload causes the system's "capsule-shaped" geometry to change dynamically in real time due to the swinging of the suspension rope. In scenarios with sudden terrain changes (such as sudden steep slopes, cliffs, dense jungle areas, building protrusions, and other discontinuous terrain), the spatial characteristics of environmental obstacles change rapidly and significantly. This can lead to problems such as delayed perception of terrain changes by the UAV and untimely calculation of the dynamically coupled system's shape by the computing unit. In some cases, the dynamic coupling characteristics of the flexible suspension can even cause a mismatch between obstacle avoidance control commands and the actual motion state of the system, ultimately resulting in the system being unable to perform effective obstacle avoidance actions and posing a safety hazard of collision between the suspension system and terrain obstacles. In many UAV application scenarios, the closer the detection payload is to the ground, the better the effect. However, in mountainous and rugged terrain scenarios, if the payload is too close to the ground, the risk factor increases significantly. Relying solely on the UAV's real-time environmental perception and response cannot guarantee the safety of both the payload and the UAV. Summary of the Invention
[0005] In view of this, embodiments of the present invention provide a flexible connection UAV-borne detection method, device and system to solve the problem of insufficient adaptability of the dynamic real-time obstacle avoidance technology of existing UAV-borne detection systems in scenarios with sudden terrain changes.
[0006] A first aspect of the present invention provides a flexible-connected unmanned aerial vehicle (UAV) onboard detection method, comprising:
[0007] The mission area is acquired, divided into grids and marked, and the safe expansion radius of each grid is determined to obtain a grid model representing the global terrain and known obstacles; each grid is marked as either flyable or no-fly.
[0008] Path planning is performed based on a grid model to obtain an initial flight plan; the initial flight plan includes at least an initial route.
[0009] Static compensation points are determined based on global terrain and known obstacles. The sling swing of each static compensation point is pre-calculated. Based on the pre-calculated sling swing, the flight path offset of the static compensation points is pre-calculated.
[0010] The initial route is statically compensated and updated using the pre-calculated route offset to obtain the statically compensated and updated flight plan.
[0011] The control system on the unmanned aerial vehicle (UAV) performs reconnaissance missions based on the flight plan updated by static compensation.
[0012] A second aspect of the present invention provides a flexible-connected unmanned aerial vehicle (UAV) onboard detection device, comprising:
[0013] The grid model construction module is configured to acquire the operational area of the detection mission, divide and mark the operational area into grids, and determine the safe expansion radius of each grid to obtain a grid model representing the global terrain and known obstacles; wherein each grid is marked as either winged or no-fly.
[0014] The path planning module is configured to perform path planning based on a grid model to obtain an initial flight plan; the initial flight plan includes at least an initial route.
[0015] The static compensation module is configured to determine static compensation points based on global terrain and known obstacles, pre-calculate the sling swing of each static compensation point, and pre-calculate the flight path offset of the static compensation points based on the pre-calculated sling swing.
[0016] The static compensation module is also configured to use the pre-calculated route offset to perform static compensation update on the initial route, resulting in a statically compensated updated flight plan;
[0017] The detection module is configured to control the unmanned aerial vehicle (UAV) system to perform detection tasks based on the flight plan updated by static compensation.
[0018] A third aspect of the present invention provides a flexible connection unmanned aerial vehicle (UAV)-borne detection system, comprising: a UAV, a flexible sling, a detection payload, and a control unit;
[0019] The drone is equipped with an inertial measurement module, a positioning module, a lidar and a vision camera to collect its own status data and information about the surrounding environment in real time;
[0020] The flexible sling is an integrated composite cable structure that integrates load power supply cables, radio frequency signal transmission cables, and structural support cables. Furthermore, gyroscope modules and tension sensors are installed on the flexible sling to acquire real-time status data of the detected load.
[0021] The control unit is configured to control the UAV-borne detection system to perform a detection task using a flexible connection method as described in the first aspect; or, the control unit includes a flexible connection UAV-borne detection device as described in the second aspect.
[0022] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows:
[0023] This invention first divides and marks the detection mission area into grids and determines the safe expansion radius of each grid, thereby obtaining a grid model representing the global terrain and known obstacles. Then, an initial flight plan is planned based on the grid model. Further, static compensation is performed on known obstacles in the mission area, and the initial flight plan is updated based on the static compensation results. Finally, the unmanned aerial vehicle (UAV) system is controlled to execute the detection mission according to the updated flight plan. This method, through grid model construction and static compensation for known obstacles, achieves closer-to-ground, terrain-following flight, improves detection effectiveness, and ensures flight safety. Attached Figure Description
[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art 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.
[0025] Figure 1 This is a flowchart illustrating a flexible connection UAV-borne detection method provided in an embodiment of the present invention.
[0026] Figure 2 This is a schematic diagram of the structure of a flexible connection UAV-borne detection device provided in an embodiment of the present invention.
[0027] Figure 3 This is a schematic diagram of the structure of a flexible connection UAV-borne detection system provided in an embodiment of the present invention. Detailed Implementation
[0028] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of the invention with unnecessary detail.
[0029] A flexible-connection unmanned aerial vehicle (UAV) detection method and apparatus according to an embodiment of the present invention will be described in detail below with reference to the accompanying drawings.
[0030] As mentioned above, in many drone application scenarios, the closer the detection payload is to the ground, the better the effect. However, in mountainous and rugged terrain scenarios, if the payload is too close to the ground, the risk factor will increase significantly. Relying solely on the drone to respond by sensing the environment in real time cannot guarantee the safety of the payload and the drone.
[0031] In view of this, embodiments of the present invention provide a flexible connection UAV-borne detection method, which improves the operational safety of UAV sling systems in complex and abrupt terrain by global terrain and obstacle grid modeling, constrained global optimal route solution, pre-calculation of filtered sling swing, pre-compensation mechanism for minor route offset, and dynamic emergency obstacle avoidance processing for unknown obstacles.
[0032] Figure 1 This is a flowchart illustrating a flexible connection-based unmanned aerial vehicle (UAV) detection method provided in an embodiment of the present invention. Figure 1 As shown, the method includes the following steps:
[0033] In step S101, the detection task operation area is obtained, the operation area is divided into grids and marked, and the safe expansion radius of each grid is determined to obtain a grid model representing the global terrain and known obstacles.
[0034] Each grid cell is marked as either flyable or no-fly.
[0035] In step S102, path planning is performed based on the grid model to obtain the initial flight plan.
[0036] The initial flight plan must include at least the initial flight route.
[0037] In step S103, static compensation points are determined based on the global terrain and known obstacles. The sling swing amount of each static compensation point is pre-calculated, and the route offset of the static compensation points is pre-calculated based on the pre-calculated sling swing amount.
[0038] In step S104, the initial route is statically compensated and updated using the pre-calculated route offset to obtain the statically compensated and updated flight plan.
[0039] In step S105, the unmanned aerial vehicle (UAV) system is controlled to perform the detection mission according to the flight plan updated by static compensation.
[0040] In some embodiments of the present invention, a detection task can be acquired, the task area can be divided and marked into grids, and the safe expansion radius of each grid can be determined to obtain a grid model representing the global terrain and known obstacles.
[0041] The grid division and labeling can be determined through on-site surveys. For example, a high-precision digital surface model (DSM) of the task operation area can be generated through methods such as LiDAR scanning. The task operation area can then be divided into grids using a rasterization algorithm, and the grid attributes can be labeled based on the terrain elevation gradient, obstacle size, and hazard level.
[0042] In one example, areas with fixed known obstacles such as steep slopes, cliffs, raised terraces, dense forests, high-voltage lines, and buildings can be marked as no-fly grids, while flat and safe work areas can be marked as flyable grids.
[0043] In some embodiments of the present invention, path planning can be performed at least based on the grid model to obtain an initial flight plan. In some examples, the starting and ending points of the UAV operation, and a preset detection flight path can be used as basic constraints, integrating multi-dimensional parameters such as detection operation accuracy requirements, maximum sling swing amplitude, UAV flight speed threshold, and near-ground safety altitude, to complete the global flight path planning using a path planning algorithm. The path planning algorithm can be selected according to actual needs, for example, an improved one can be chosen. algorithm.
[0044] In some embodiments of the present invention, static compensation points can be determined based on global terrain and known obstacles. The sling swing amount is pre-calculated for each static compensation point, and the flight path offset is pre-calculated based on the pre-calculated sling swing amount. Then, the initial flight path is updated with static compensation using the pre-calculated flight path offset to obtain the updated flight plan.
[0045] In other words, the terrain slope change characteristics of different flight segments can be determined based on the results of pre-survey markings, and the flight attitude threshold and speed adjustment range of the UAV in the corresponding flight segment can be pre-calibrated based on the slope change characteristics. This allows for the pre-setting of flight deceleration and attitude fine-tuning pre-commands for the UAV in the transition area of abrupt terrain changes.
[0046] Finally, the unmanned aerial vehicle (UAV) system can be controlled to perform reconnaissance missions based on the flight plan updated by static compensation.
[0047] According to the technical solution provided in this embodiment of the invention, the detection task operation area is first divided and marked into grids, and the safe expansion radius of each grid is determined, thereby obtaining a grid model representing the global terrain and known obstacles. Then, an initial flight plan is planned based on the grid model, and further static compensation is performed on the known obstacles in the operation area. The initial flight plan is updated based on the static compensation results, and finally, the unmanned aerial vehicle (UAV) system is controlled to execute the detection task according to the updated flight plan. This method, through grid model construction and static compensation for known obstacles, achieves terrain-following flight closer to the ground, improves detection effectiveness, and ensures flight safety.
[0048] In some embodiments of the present invention, the safe expansion radius of the target grid can be determined in the following manner:
[0049] Determine the inherent safety radius of the UAV fuselage and the maximum outer contour radius of the detection payload carried by the UAV-borne system;
[0050] Determine the maximum limit angle of the flexible sling relative to the vertical direction under UAV flight conditions;
[0051] Predict the real-time effective suspension length of the UAV's flexible sling on the target grid;
[0052] The product of the real-time effective overhang length and the sine of the maximum limit swing angle is determined as the swing safety radius;
[0053] The sum of the inherent safe radius of the UAV fuselage, the maximum outline radius of the detection payload, and the swing safe radius is determined as the safe expansion radius of the target grid.
[0054] The target grid is any grid.
[0055] In other words, unlike traditional drones that use a fixed grid expansion method based solely on fuselage size, this invention proposes a dynamic safety expansion model adapted to flexible suspension systems. By combining the flexible sling length, load swing extremes, and equipment dimensions, a dynamic adaptive safety expansion radius is calculated. This model accurately covers the dynamic swing envelope of the flexible system, completely eliminating the hidden collision risks caused by sling swing and load offset. The formula for calculating the dynamic safety expansion radius is as follows:
[0056] ;
[0057] in, The safe expansion radius of the target grid; The inherent safety radius of the drone fuselage is used to avoid the risk of collision with the drone itself; The real-time effective suspension length of the UAV's flexible sling over the target grid; The maximum limit angle of the flexible sling relative to the vertical direction under UAV flight conditions is obtained from the system's initial calibration and real-time status calculation. The maximum outer radius of the detection payload carried by the unmanned aerial vehicle (UAV) system.
[0058] This invention achieves adaptive outward expansion of the obstacle grid by updating the expansion radius in real time according to the sling length and swing state of different flight segments. This accurately matches the dynamic motion characteristics of the flexible suspended underactuated twin-body system and avoids the collision blind spot caused by the flexible structure from the modeling source.
[0059] In some embodiments of the present invention, path planning based on a grid model includes:
[0060] Basic constraints are constructed based on the start point, end point, operation area, and marker values of each grid cell of the drone operation;
[0061] Planning constraints are constructed based on the operational accuracy requirements of the reconnaissance mission and the flight parameters of the UAV;
[0062] A multi-weighted fusion cost function is constructed; the multi-weighted fusion cost function is a weighted sum of terrain risk cost, suspension cable swing coupling cost, and detection efficiency cost; the terrain risk cost is determined based on the grid model;
[0063] A path planning model is constructed based on basic constraints, planning constraints, and a multi-weighted fusion cost function. The path planning model is then used to perform path planning, resulting in an initial flight plan.
[0064] In other words, embodiments of the present invention can construct a multi-weighted fusion cost function adapted to a flexible suspended underactuated dual-body system, comprehensively considering three core indicators: terrain safety, sling swing stability, and detection operation efficiency. The multi-weighted fusion cost function is as follows:
[0065] ;
[0066] For the current grid node The total value of the agency is the highest, and the higher the agency value, the higher the flight risk and the worse the operational adaptability of the corresponding flight segment. As for the terrain risk cost, the calculation is normalized based on the terrain elevation gradient of the flight segment, the severity of terrain abruptness, and the safe distance from obstacles. The terrain risk cost is significantly increased in steep slopes, cliffs, areas of terrain abruptness, and areas near obstacles. To account for the cost of sling swing coupling, and considering the flight path curvature, flight speed, and terrain disturbance characteristics, the maximum swing angle, tension fluctuation amplitude, and load offset of the sling are pre-calculated to characterize the dynamic instability risk of the flexible suspension system. To reduce the cost of detection efficiency, the system is designed based on factors such as the integrity of flight strip coverage, the smoothness of flight routes, and the proportion of invalid flight mileage, ensuring the overall efficiency of detection operations and the continuity of data.
[0067] , and As adaptive weighting coefficients, the three satisfy the normalization constraint conditions. The path planning algorithm can adaptively and dynamically adjust the weight ratio according to the task scenario: in scenarios with abrupt and complex terrain such as mountainous areas, cliffs, and steep slopes, it can adaptively increase the weight ratio. and Prioritize ensuring flight terrain safety and the stability of the hoisting system; in conventional reconnaissance areas with flat terrain, the accuracy can be appropriately increased. This takes into account both the efficiency of the detection operation.
[0068] This invention employs multi-dimensional cost-weighted optimization to iteratively solve for the globally optimal flight path, prioritizing the selection of flight segments with gentle terrain, minimal obstacles, and optimal suspension sway suppression, thereby avoiding all known obstacles and abrupt terrain risk areas in advance from the route planning level.
[0069] In some embodiments of the present invention, the sling swing includes the predicted maximum swing angle of the flexible sling at each static compensation point; the flight path offset includes the maximum lateral offset and the maximum vertical offset of the detection load at each static compensation point.
[0070] Pre-calculation of sling sway and route offset for the target static compensation point may include:
[0071] Determine the flight attitude threshold and speed adjustment range for the segment corresponding to the target static compensation point in the initial flight path; the target static compensation point can be any static compensation point.
[0072] Construct the state equations of the suspension system containing the flexible sling: And the observation equations for the suspension system: ;in, For the suspension system at all times The state vector, , The swing angle of the flexible sling relative to the vertical direction. The angular velocity of the flexible sling. For the tension fluctuation deviation of the flexible sling, It is the transpose symbol. For the suspension system at all times The state vector, For the observation vector, For the suspension system at all times The disturbance input quantity. For the suspension system at all times Process noise, For the suspension system at all times Observed Gaussian white noise, Here is the system state transition matrix. To control the input matrix, The observation matrix;
[0073] The predicted maximum swing angle of the target static compensation point is obtained by Kalman filtering. ;
[0074] Calculate the maximum lateral offset of the detection load at the target static compensation point based on the predicted maximum swing angle. and the maximum vertical height deviation of the detection load ;in, ;and ; is the terrain slope correction factor; the target grid is the grid where the target static compensation point is located.
[0075] Furthermore, using the pre-calculated route offset to perform static compensation updates on the initial route can include:
[0076] Determine the lateral coordinates of the probe load after static compensation for: Vertical coordinates of the detection load after static compensation for: ;in, and The horizontal and vertical coordinates of the target static compensation point in the initial flight path are given. This is the flexible oscillation compensation coefficient;
[0077] The horizontal and vertical coordinates of the probe payload after static compensation are fitted using a smoothing algorithm to obtain the updated flight plan after static compensation. The smoothing algorithm includes extreme constraints on the flight path curvature and threshold constraints on the speed change.
[0078] In other words, based on the terrain slope variation characteristics of different flight segments, the flight attitude threshold and speed adjustment range of the UAV in the corresponding flight segment are pre-calibrated. For areas connecting sudden terrain changes, pre-commands for flight deceleration and attitude fine-tuning are set in advance. To solve the problem of instantaneous impact swaying of the sling caused by sudden terrain changes, this embodiment of the invention introduces a Kalman filter sling sway pre-calculation + flight path micro-offset pre-compensation mechanism. Through a standardized state iteration algorithm, the sling sway trend is predicted in advance and the flight path is actively pre-cancelled. The complete algorithm model is as follows:
[0079] State-space modeling of sling swing system: Using the swing angle, swing angular velocity, and tension fluctuation deviation of flexible sling as core state variables, construct the state equation and observation equation of the sling system suitable for terrain change disturbance scenarios.
[0080] Kalman filter iterative optimal estimation: Through a dual iterative process of prediction and correction, sensor noise and vibration interference are filtered out, enabling high-precision prediction of the maximum sway state of the sling in the next leg of the journey. Specifically, this includes:
[0081] State prediction: ;
[0082] Covariance prediction: ;
[0083] Kalman Gain Update: ;
[0084] State correction: ;
[0085] Covariance update: ;
[0086] in, The process noise covariance matrix is... To observe the noise covariance matrix, It is the identity matrix; This is the predicted state value. and The corrected state prediction value; This is the predicted value of covariance. This is the updated covariance value; This is the Kalman gain. Through iterative calculation, the predicted maximum sway angle for abrupt terrain segments can be accurately output. The lateral offset of the load provides precise input for flight path compensation.
[0087] Two-dimensional dynamic pre-compensation model for flight routes: Based on the filtered and predicted sling swing offset, a two-dimensional offset compensation algorithm for the flight route in both lateral and longitudinal dimensions is established to pre-correct the nodes of the original planned flight route and offset the passive swing caused by abrupt terrain changes.
[0088] ;
[0089] in, It is a flexible swing compensation coefficient that is adaptively adjusted according to the sling material and load weight.
[0090] Flight path smoothing: A smoothing algorithm is used to fit the nodes of the compensated discrete flight path, constraining the extreme values of flight path curvature and the threshold for sudden velocity changes. This eliminates trajectory jitter caused by single-point compensation, ensuring continuous UAV attitude and stable speed, and thoroughly suppressing instantaneous dynamic coupling disturbances in the flexible system at the trajectory level. The smoothing algorithm can be selected according to actual needs, such as a third-order spline smoothing algorithm.
[0091] Based on the aforementioned complete algorithm, the entire process of offline pre-calibration, including flight segment state prediction, coordinate compensation, and trajectory smoothing, is achieved. This ensures that the UAV, sling, and payload remain within a safe flight range at all times, enabling zero-delay obstacle avoidance across the entire flight segment for known obstacles. This strategy operates entirely in an offline pre-calculation and pre-calibration mode before operation. During actual flight, the UAV does not require high-frequency environmental perception, real-time model solving, or trajectory iteration calculations, significantly reducing the computational load and response latency risks of the onboard chip. Overall obstacle avoidance stability and operational adaptability are far superior to traditional fully online real-time obstacle avoidance algorithms, perfectly adapting to the dynamic coupling characteristics of flexible sling systems.
[0092] In some embodiments of the present invention, when controlling the unmanned aerial vehicle system to perform a reconnaissance mission according to a statically compensated flight plan, the method further includes the following steps:
[0093] In response to the detection of an unknown obstacle on the flight path during flight, the real-time dynamic hazard envelope radius of the detection payload is calculated; wherein, the real-time dynamic hazard envelope radius is the sum of the current swing safety radius of the UAV and the maximum outer contour radius of the detection payload carried by the UAV-borne system;
[0094] The collision risk level is determined based on the real-time dynamic danger envelope radius;
[0095] In response to determining the collision risk level as a warning level, local trajectory replanning is performed, and the unmanned aerial vehicle system is controlled to perform detection tasks at least according to the replanned local trajectory;
[0096] In response to determining that the collision risk level is dangerous, emergency obstacle avoidance and attitude sway suppression linkage control are implemented for the UAV-borne system.
[0097] In response to determining that the collision risk level is safe, the control system of the unmanned aerial vehicle maintains the flight plan after static compensation to carry out the detection mission.
[0098] Among these methods, unknown obstacles are detected collaboratively by a lidar and a visual camera mounted on the drone; the lidar calculates the spatial coordinates of the unknown obstacles in real time using a point cloud clustering algorithm. ;
[0099] The visual camera uses feature matching and target detection algorithms to identify the features of unknown obstacles and outputs the pixel coordinates and detection confidence of the unknown obstacles. The confidence level of this detection It can be displayed on the human-machine interface to assist human judgment;
[0100] The UAV-borne detection system converts pixel coordinates into spatial coordinates of unknown obstacles calculated by the visual camera. And for spatial coordinates and Weighted fusion is performed to obtain the three-dimensional coordinates of the unknown obstacles after fusion correction. ;in, , Dynamic confidence weights are used to adapt to different scenarios.
[0101] In some implementations, emergency obstacle avoidance and attitude sway suppression are linked and controlled for the unmanned aerial vehicle (UAV) system, including:
[0102] An improved dynamic window algorithm is constructed; the flight speed window of the improved dynamic window algorithm includes the linear velocity and angular velocity of the UAV; the evaluation function of the improved dynamic window algorithm includes obstacle distance cost, flight speed smoothing cost, and sling swing constraint.
[0103] The obstacle avoidance trajectory is calculated based on the dynamic window algorithm, and the flight plan updated by static compensation is compensated in real time based on the obstacle avoidance trajectory to obtain the flight plan after real-time compensation.
[0104] The control system on the unmanned aerial vehicle (UAV) executes the detection mission according to the flight plan after real-time compensation.
[0105] In other words, based on the aforementioned offline obstacle avoidance of known obstacles, to overcome the technical shortcomings of single pre-planned obstacle avoidance in adapting to sudden unknown obstacles at the work site, and to compensate for the scenario limitations of static terrain models, this invention constructs a dynamic emergency obstacle avoidance algorithm system for unknown obstacles adapted to a flexible suspended underactuated dual-body system. This system integrates multi-source real-time sensing, flexible system dynamic coupling criteria, anti-sway local trajectory replanning, and smooth homing steady-state control to form a complete, highly robust online emergency obstacle avoidance mechanism. It can achieve millisecond-level precise avoidance of unknown obstacles such as suddenly fallen vegetation, temporary structures, and moving obstacles not recorded in the DSM terrain model. Simultaneously, it takes into account the sway suppression of the sling and detection load throughout the process, avoiding system instability, load swaying, and secondary collision risks caused by emergency obstacle avoidance maneuvers. The specific layered algorithm implementation scheme is as follows:
[0106] To address the issues of missed or false detections of obstacles and data jumps caused by single sensors in complex outdoor lighting, dust, and airflow disturbance scenarios, and to improve the accuracy and stability of unknown obstacle perception, this invention employs a time-weighted fusion algorithm that combines LiDAR and a visual camera to achieve high-precision and robust calculation of the three-dimensional state of obstacles.
[0107] Specifically, the lidar uses point cloud clustering algorithms to calculate the three-dimensional spatial position, contour size, and elevation distribution information of obstacles in real time, and outputs the spatial coordinates of the obstacles. The visual camera identifies obstacles using feature matching and target detection algorithms, outputting the pixel coordinates of the obstacles and the detection confidence level. And solve for the corresponding spatial coordinates. .
[0108] To integrate the detection advantages of the two types of sensors, a dynamic weighted fusion observation model is constructed to achieve complementary correction and anomaly filtering of multi-source data:
[0109] ;
[0110] in, To fuse the corrected 3D accurate coordinates of the obstacles; The system employs scene-adaptive dynamic confidence weights, which can adjust the weight ratio in real time according to the operating environment: increasing the weight of LiDAR in complex scenes with strong light, dust, and low texture to weaken visual sensing interference; and increasing the weight of the visual camera in low light and high texture scenes to compensate for the sparse detection defects of LiDAR. Simultaneously, through a multi-frame temporal iterative verification mechanism, it eliminates anomalous observations with single-frame jumps, stably outputting core parameters such as the relative distance, azimuth, and altitude difference between obstacles and the UAV, and constructing a local dynamic obstacle grid model in real time, providing accurate environmental input for subsequent risk assessment and trajectory replanning.
[0111] It is understood that the technical solution provided by this invention aims to ensure the safety of the payload and the UAV to the greatest extent possible. Based on this, all possible obstacles need to be avoided. Therefore, for obstacles that cannot be detected by radar and cameras, such as small objects like power lines, on-site reconnaissance needs to be conducted before flight to identify them as non-flyable grids. For obstacles that can only be detected by either lidar or camera, they can be identified as obstacles by default in the path planning and displayed on the human-machine interface. Manual assistance can be introduced when necessary.
[0112] Traditional UAV obstacle avoidance algorithms rely solely on the UAV's dimensions for collision detection, neglecting the large-scale danger envelope caused by flexible cable swaying and dynamic load shifts. This leads to insufficient risk prediction and delayed obstacle avoidance. To address this deficiency, this invention constructs a three-body coupled collision risk quantification model for the UAV, cable, and load. By combining real-time suspended motion status to calculate the real-time dynamic danger envelope radius, a multi-level collision risk criterion is established, enabling accurate prediction of collision risks for flexible systems.
[0113] The real-time dynamic swing radius can be calculated using the following formula:
[0114] ;
[0115] in, The real-time dynamic danger envelope radius of the load characterizes the real-time maximum danger coverage of the flexible system. This represents the actual swing angle of the sling relative to the vertical direction at the current moment. It is based on the real-time dynamic hazard envelope radius and the relative distance to the obstacle. Combined with preset emergency safety margin The risk threshold is divided into three levels to achieve tiered and graded response control:
[0116] 1) Security level: There is no risk of collision; continue flying along the original route.
[0117] 2) Warning Level: This triggers a pre-response, prompting the drone to prepare for local trajectory replanning.
[0118] 3) Hazard level: This immediately triggers the emergency obstacle avoidance and attitude sway suppression linkage control.
[0119] This hierarchical judgment mechanism quantifies and integrates the dynamic swing characteristics of the flexible suspension system into the collision risk assessment, completely abandoning the single distance judgment logic of traditional rigid UAV obstacle avoidance. It can accurately cover the hidden collision risks caused by sling swing and load offset, realize the early prediction and hierarchical response of unknown obstacle risks, and greatly improve obstacle avoidance safety and fault tolerance in complex field scenarios.
[0120] To achieve optimal control that combines rapid obstacle avoidance with flexible system sway suppression, this invention further introduces a sling sway constraint cost term on the basis of the traditional Dynamic Window Approach (DWA), and constructs an improved DWA algorithm adapted to the flexible underactuated system. This solves the problem that the traditional DWA algorithm only considers obstacle avoidance distance and flight speed, ignoring the instability of the system after obstacle avoidance caused by dynamic disturbances of the flexible structure.
[0121] The algorithm constructs a flight speed window that includes linear velocity and angular velocity. The optimal obstacle avoidance trajectory is selected by iterating through a multi-dimensional weighted evaluation function. The comprehensive evaluation function is as follows:
[0122] ;
[0123] in, The obstacle distance cost term represents the safe distance between the trajectory and the obstacle. The greater the distance, the higher the cost, ensuring obstacle avoidance safety. This is a flight speed smoothing cost term, used to constrain sudden changes in flight speed and ensure smooth and continuous flight attitude. The newly added cable swing constraint cost term is used to characterize the cable swing angle increment and tension fluctuation amplitude corresponding to the candidate trajectory, effectively screening the optimal trajectory with low disturbance and small swing. , and These are normalized weighting coefficients, and the weighting ratio of the three can be adaptively adjusted according to terrain complexity and load status.
[0124] All of the above-mentioned optional technical solutions can be combined in any way to form optional embodiments of the present invention, and will not be described in detail here.
[0125] The following are embodiments of the apparatus of the present invention, which can be used to execute embodiments of the method of the present invention. For details not disclosed in the embodiments of the apparatus of the present invention, please refer to the embodiments of the method of the present invention.
[0126] Figure 2 This is a structural schematic diagram of a flexible-connected unmanned aerial vehicle (UAV) detection device provided in an embodiment of the present invention. Figure 2 As shown, the device includes:
[0127] The grid model construction module 201 is configured to acquire the operational area of the exploration mission, divide and mark the operational area into grids, and determine the safe expansion radius of each grid to obtain a grid model representing the global terrain and known obstacles; wherein each grid is marked as either winged or no-fly.
[0128] The path planning module 202 is configured to perform path planning based on a grid model to obtain an initial flight plan; the initial flight plan includes at least an initial route.
[0129] The static compensation module 203 is configured to determine static compensation points based on global terrain and known obstacles, pre-calculate the sling swing of each static compensation point, and pre-calculate the route offset of the static compensation points based on the pre-calculated sling swing.
[0130] The static compensation module 203 is also configured to use the pre-calculated route offset to perform static compensation update on the initial route, so as to obtain the statically compensated updated flight plan.
[0131] The detection module 204 is configured to control the unmanned aerial vehicle system to perform detection tasks according to the flight plan updated by static compensation.
[0132] According to the technical solution provided in the embodiments of the present invention, the detection task operation area is first divided and marked into grids, and the safe expansion radius of each grid is determined, thereby obtaining a grid model representing the global terrain and known obstacles. Then, an initial flight plan is planned based on the grid model, and further static compensation is performed on the known obstacles in the operation area. The initial flight plan is updated based on the static compensation results, and finally, the unmanned aerial vehicle (UAV) system is controlled to execute the detection task according to the updated flight plan. This device, through grid model construction and static compensation for known obstacles, achieves terrain-following flight closer to the ground, improves detection effectiveness, and ensures flight safety.
[0133] In some implementations, the safe expansion radius of the target grid is determined as follows: the inherent safe radius of the UAV fuselage and the maximum outline radius of the detection payload carried by the UAV-borne system are determined; the maximum limit swing angle of the flexible sling in the vertical direction under UAV flight conditions is determined; the real-time effective suspension length of the UAV flexible sling in the target grid is predicted; the product of the real-time effective suspension length and the sine of the maximum limit swing angle is determined as the swing safety radius; the sum of the inherent safe radius of the UAV fuselage, the maximum outline radius of the detection payload, and the swing safety radius is determined as the safe expansion radius of the target grid; wherein, the target grid is any grid.
[0134] In some implementations, path planning is based on a grid model, including: constructing basic constraints based on the UAV's start point, end point, operating area, and the marker values of each grid; constructing planning constraints based on the operational accuracy requirements of the reconnaissance mission and the UAV's flight parameters; constructing a multi-weighted fusion cost function; the multi-weighted fusion cost function is a weighted sum of terrain risk cost, sling swing coupling cost, and reconnaissance efficiency cost; the terrain risk cost is determined based on the grid model; constructing a path planning model based on the basic constraints, planning constraints, and multi-weighted fusion cost function; and using the path planning model to perform path planning to obtain an initial flight plan.
[0135] In some implementations, the static compensation point includes at least abrupt terrain change points; the sling swing includes the predicted maximum swing angle of the flexible sling at each static compensation point; the flight path offset includes the maximum lateral and maximum vertical offset of the detection load at each static compensation point; pre-calculation of the sling swing and flight path offset for the target static compensation point includes: determining the flight attitude threshold and speed adjustment range of the segment corresponding to the target static compensation point in the initial flight path; the target static compensation point is any static compensation point; and constructing the state equation of the suspension system containing the flexible sling. And the observation equations for the suspension system: ;in, For the suspension system at all times The state vector, , The swing angle of the flexible sling relative to the vertical direction. The angular velocity of the flexible sling. For the tension fluctuation deviation of the flexible sling, It is the transpose symbol. For the suspension system at all times The state vector, For the observation vector, For the suspension system at all times The disturbance input quantity. For the suspension system at all times Process noise, For the suspension system at all times Observed Gaussian white noise, Here is the system state transition matrix. To control the input matrix, The observation matrix is used; the predicted maximum swing angle of the target static compensation point is calculated using Kalman filtering. ; Calculate the maximum lateral offset of the detection load at the target static compensation point based on the predicted maximum swing angle. and the maximum vertical height deviation of the detection load ;in, ;and ; This represents the predicted real-time effective suspension length of the flexible sling on the target grid. is the terrain slope correction factor; the target grid is the grid where the target static compensation point is located.
[0136] In some implementations, the initial flight path is statically compensated and updated using the pre-calculated flight path offset, including: determining the lateral coordinates of the statically compensated detection payload. for: Vertical coordinates of the detection load after static compensation for: ;in, and The horizontal and vertical coordinates of the target static compensation point in the initial flight path are given. The coefficient is the flexible oscillation compensation coefficient. The horizontal and vertical coordinates of the probe load after static compensation are fitted using a smoothing algorithm to obtain the updated flight plan after static compensation. The smoothing algorithm includes extreme value constraints on the flight path curvature and threshold constraints on speed mutation.
[0137] In some implementations, the detection module is further configured to, in response to detecting an unknown obstacle on the flight path during flight, calculate the real-time dynamic hazard envelope radius of the detection payload; wherein the real-time dynamic hazard envelope radius is the sum of the current swing safety radius of the UAV and the maximum outline radius of the detection payload carried by the UAV-borne system; determine the collision risk level based on the real-time dynamic hazard envelope radius; in response to determining that the collision risk level is a warning level, perform local trajectory replanning and control the UAV-borne system to perform the detection task at least according to the replanned local trajectory; in response to determining that the collision risk level is a dangerous level, perform emergency obstacle avoidance and attitude swing suppression linkage control on the UAV-borne system; in response to determining that the collision risk level is a safe level, control the UAV-borne system to maintain the statically compensated flight plan to perform the detection task.
[0138] In some implementations, emergency obstacle avoidance and attitude sway suppression are linked for the UAV-borne system, including: constructing an improved dynamic window algorithm; the flight speed window of the improved dynamic window algorithm includes the linear velocity and angular velocity of the UAV; the evaluation function of the improved dynamic window algorithm includes an obstacle distance cost term, a flight speed smoothing cost term, and a sling sway constraint term; calculating the obstacle avoidance trajectory based on the dynamic window algorithm, and performing real-time compensation on the statically compensated updated flight plan based on the obstacle avoidance trajectory to obtain a real-time compensated flight plan; and controlling the UAV-borne system to execute the detection task according to the real-time compensated flight plan.
[0139] In some implementations, unknown obstacles are detected collaboratively by a lidar and a visual camera mounted on the drone; wherein, the lidar calculates the spatial coordinates of the unknown obstacles in real time using a point cloud clustering algorithm. The visual camera uses feature matching and target detection algorithms to identify the features of unknown obstacles and outputs the pixel coordinates and detection confidence level of the unknown obstacles. The UAV-borne detection system converts pixel coordinates into spatial coordinates of unknown obstacles calculated by the visual camera. And for spatial coordinates and Weighted fusion is performed to obtain the three-dimensional coordinates of the unknown obstacles after fusion correction. ;in, , Dynamic confidence weights are used to adapt to different scenarios.
[0140] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.
[0141] like Figure 3 As shown, this embodiment of the invention also provides a flexible connection UAV-borne detection system. The UAV-borne detection system includes a UAV, a flexible sling, a detection payload, and a control unit. The whole system is an underactuated two-body coupled system, which can effectively adapt to low-altitude detection scenarios such as near-ground detection and ice detection, and at the same time has precise payload swing suppression capability.
[0142] The control unit can be located on the ground or elsewhere, and can be connected to the drone and the detection payload wirelessly or otherwise. It can receive real-time detection signals from the detection payload and send control commands to the drone.
[0143] The drone is equipped with an inertial measurement module, a positioning module, a lidar, and a vision camera to collect its own status data and information about the surrounding environment in real time.
[0144] The flexible sling is an integrated composite cable structure, simultaneously providing power cables for the detection load, radio frequency signal transmission cables, and structural support cables. Furthermore, gyroscope modules and tension sensors are installed on the flexible sling to acquire real-time status data of the detection load. The detection load can be a near-ground detection load.
[0145] The control unit is configured to use Figure 1 The embodiment shown provides a flexible connection UAV-borne detection method that controls the UAV-borne detection system to perform detection tasks; or, the control unit includes... Figure 2 The embodiment shown provides a flexible-connected unmanned aerial vehicle (UAV) detection device.
[0146] The unmanned aerial vehicle (UAV) platform can be a multi-rotor UAV or an unmanned helicopter, with a core load sway suppression system providing key support for system attitude control and sway compensation. The UAV itself uses an Inertial Measurement Unit (IMU), a Real-time Kinematic Survey (RTK) module, and a Global Positioning System (GPS) to collect real-time data on its flight attitude, position, velocity, acceleration, and other status information, enabling high-precision perception of the UAV's own state.
[0147] The perception of the load's motion status relies on the collaborative acquisition of multiple modules. A gyroscope module deployed at the upper end of the sling collects load attitude deflection data, a tension sensor built into the sling monitors real-time tension changes and deformation, and a visual sensor mounted on the UAV captures the load's spatial position and swing trajectory. To eliminate acceleration noise and data jumps caused by flight vibration, airflow interference, and sensor acquisition errors, this embodiment of the invention introduces a Kalman filter algorithm to reduce noise and optimize the fusion of multi-source perception data, outputting accurate dynamic state parameters of the UAV-sling-load system.
[0148] Based on the optimized system state data, this embodiment of the invention uses a cascaded proportional-integral-derivative (PID) control algorithm to perform dynamic closed-loop control of the UAV's flight attitude and trajectory, compensate for load swing deviations in real time, suppress disordered swing of the flexible sling, and solve the problem of dynamic coupling interference in the flexible suspended twin-body system.
[0149] The flexible sling is an integrated composite cable structure that integrates load power cables, radio frequency signal transmission cables, and structural support cables, thus fulfilling the triple functions of structural load-bearing, power supply, and signal transmission. To adapt to different load requirements, the sling adopts a differentiated material design: nylon is used for light load conditions (load weight below 30kg), balancing lightweight and flexibility; Kevlar is used for medium load conditions (load weight 30kg~100kg), balancing strength and toughness; and polyethylene is used for heavy load conditions (load weight above 100kg), possessing ultra-high structural strength and tensile strength.
[0150] The UAV-borne detection system provided in this embodiment of the invention carries a detection payload that is a near-ground dedicated detection device, preferably a high-precision low-altitude detection payload such as a ground-penetrating radar or an ice-penetrating radar. It is flexibly connected to the UAV through a flexible sling, making it suitable for low-altitude near-ground detection operation scenarios in complex terrain.
[0151] The flight test operation procedure for this system is as follows:
[0152] (1) Determine the flight area according to the test requirements, delineate the operation boundary and no-fly zone, and sort out the basic working condition information of the operation area.
[0153] (2) The target test area is scanned in all directions by lidar to generate a high-precision DSM digital surface model, accurately identify abrupt terrain such as steep slopes, cliffs, and dense obstacles, and simultaneously mark high-risk obstacle areas such as high-voltage lines and building facilities through on-site surveys to complete the basic data modeling of the working environment.
[0154] (3) Based on the high-precision DSM terrain model, combined with core parameters such as the detection payload operation parameters, flexible sling length, UAV flight performance, and near-ground detection safety altitude, the optimal flight path is solved by the intelligent route planning algorithm to generate a refined flight path adapted to the flexible sling system, avoiding known high-risk areas and abrupt terrain risk areas.
[0155] (4) For high-value, easily damaged, and fault-tolerant detection payloads, simulated payloads with the same shape, weight, and aerodynamic characteristics as the real payloads are used for preliminary test flights. This effectively avoids problems such as DSM model errors, sudden terrain deviations, and insufficient flight path adaptability, and verifies the rationality of the flight path and the stability of the system in advance, eliminating the risk of damage to the real payloads during operation. This step is optional.
[0156] (5) After completing the route verification, equipment debugging and status self-check, carry out formal flight detection operations with real detection payloads according to the planned route, and monitor the UAV attitude, sling status, payload position and terrain environment throughout the process to achieve safe and stable operation.
[0157] The technical solution provided by this invention achieves closer-to-ground, terrain-following flight by scanning existing terrain and detecting rope pose in real time, thus improving detection effectiveness. Flight safety is ensured through flight testing procedures and flight algorithms.
[0158] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.
Claims
1. A flexible connected unmanned airborne detection method, characterized in that, include: The operational area for the reconnaissance mission is acquired, the operational area is divided into grids and marked, and the safe expansion radius of each grid is determined to obtain a grid model representing the global terrain and known obstacles; wherein, each grid is marked as either winged or no-fly zone; Path planning is performed at least based on the grid model to obtain an initial flight plan; the initial flight plan includes at least an initial route. Static compensation points are determined based on the global terrain and known obstacles. The sling swing amount of each static compensation point is pre-calculated. Based on the pre-calculated sling swing amount, the flight path offset of the static compensation points is pre-calculated. The initial route is statically compensated and updated using the pre-calculated route offset to obtain the statically compensated and updated flight plan. The control system of the unmanned aerial vehicle (UAV) performs the detection mission according to the flight plan updated by the static compensation.
2. The method of claim 1, wherein, The safe expansion radius of the target grid is determined as follows: Determine the inherent safety radius of the UAV fuselage and the maximum outer contour radius of the detection payload carried by the UAV-borne system; Determine the maximum limit angle of the flexible sling relative to the vertical direction under UAV flight conditions; Predict the real-time effective suspension length of the UAV's flexible sling on the target grid; The product of the real-time effective overhang length and the sine of the maximum limit swing angle is determined as the swing safety radius; The sum of the inherent safety radius of the UAV fuselage, the maximum outline radius of the detection payload, and the swing safety radius is determined as the safety expansion radius of the target grid. The target grid is any grid.
3. The UAV-borne detection method with flexible connection according to claim 1, characterized in that, Path planning based on the grid model includes: Basic constraints are constructed based on the drone operation start point, operation end point, the operation area, and the marker values of each grid. Planning constraints are constructed based on the operational accuracy requirements of the reconnaissance mission and the flight parameters of the UAV; A multi-weighted fusion cost function is constructed; the multi-weighted fusion cost function is a weighted sum of terrain risk cost, sling swing coupling cost, and detection efficiency cost; the terrain risk cost is determined at least based on the grid model; A path planning model is constructed based on the basic constraints, the planning constraints, and the multi-weighted fusion cost function. The path planning model is then used to perform path planning to obtain an initial flight plan.
4. The method of claim 1, wherein, The static compensation points include at least the terrain abrupt change points; the sling swing includes the predicted maximum swing angle of the flexible sling at each static compensation point; the flight path offset includes the maximum lateral offset and the maximum vertical offset of the detection load at each static compensation point; Pre-calculation of sling sway and route offset is performed on the target static compensation point, including: Determine the flight attitude threshold and speed adjustment range for the segment corresponding to the target static compensation point in the initial flight path; the target static compensation point can be any static compensation point. Construct the state equations of the suspension system containing the flexible sling: And the observation equations for the suspension system: ;in, For the suspension system at all times The state vector, , The swing angle of the flexible sling relative to the vertical direction. The angular velocity of the flexible sling. For the tension fluctuation deviation of the flexible sling, It is the transpose symbol. For the suspension system at all times The state vector, For the observation vector, For the suspension system at all times The disturbance input quantity. For the suspension system at all times Process noise, For the suspension system at all times Observed Gaussian white noise, Here is the system state transition matrix. To control the input matrix, The observation matrix; The pre-judgment maximum swing angle of the target static compensation point is calculated by Kalman filtering ; Calculate the maximum lateral offset of the detection load at the target static compensation point based on the predicted maximum swing angle. and the maximum vertical height deviation of the detection load ;in, ;and ; This represents the predicted real-time effective suspension length of the flexible sling on the target grid. is the terrain slope correction factor; the target grid is the grid where the target static compensation point is located.
5. The method of claim 4, wherein, The initial route is statically compensated and updated using the pre-calculated route offset, including: Determine the lateral coordinates of the probe load after static compensation for: Vertical coordinates of the detection load after static compensation for: ;in, and The horizontal and vertical coordinates of the target static compensation point in the initial flight path are given. This is the flexible oscillation compensation coefficient; The horizontal and vertical coordinates of the probe payload after static compensation are fitted using a smoothing algorithm to obtain the updated flight plan after static compensation. The smoothing algorithm includes extreme constraints on the flight path curvature and threshold constraints on the speed change.
6. The method of claim 2, wherein, The method further includes controlling the unmanned aerial vehicle (UAV) system to execute the reconnaissance mission according to the statically compensated flight plan. In response to the detection of an unknown obstacle on the flight path during flight, the real-time dynamic hazard envelope radius of the detection payload is calculated; wherein, the real-time dynamic hazard envelope radius is the sum of the current swing safety radius of the UAV and the maximum outer contour radius of the detection payload carried by the UAV-borne system; The collision risk level is determined based on the real-time dynamic danger envelope radius. In response to determining the collision risk level as a warning level, local trajectory replanning is performed, and the unmanned aerial vehicle system is controlled to perform detection tasks at least according to the replanned local trajectory; In response to determining that the collision risk level is a dangerous level, emergency obstacle avoidance and attitude sway suppression linkage control are performed on the UAV-borne system. In response to determining that the collision risk level is a safe level, the unmanned aerial vehicle (UAV) system is controlled to maintain the detection mission according to the statically compensated flight plan.
7. The flexibly connected unmanned airborne detection method of claim 6, wherein, Emergency obstacle avoidance and attitude sway suppression linkage control for UAV-borne systems, including: An improved dynamic window algorithm is constructed; the flight speed window of the improved dynamic window algorithm includes the linear velocity and angular velocity of the UAV; the evaluation function of the improved dynamic window algorithm includes an obstacle distance cost term, a flight speed smoothing cost term, and a sling swing constraint term. The obstacle avoidance trajectory is calculated based on the dynamic window algorithm, and the flight plan updated by static compensation is compensated in real time based on the obstacle avoidance trajectory to obtain the flight plan after real-time compensation. The control system of the unmanned aerial vehicle (UAV) executes the detection mission according to the flight plan after real-time compensation.
8. The method of claim 6, wherein, The unknown obstacles are detected by the combined use of lidar and visual cameras mounted on the drone; The laser radar resolves the spatial coordinates of unknown obstacles in real time through a point cloud clustering algorithm ; The visual camera completes feature recognition of the unknown obstacle through feature matching and target detection algorithm, and outputs pixel coordinates and detection confidence of the unknown obstacle ; The unmanned aerial detection system converts the pixel coordinates into spatial coordinates of unknown obstacles calculated by a visual camera ; and the spatial coordinates and are weighted and fused to obtain the three-dimensional coordinates of the unknown obstacles after fusion correction ; wherein , is a scene-adaptive dynamic confidence weight.
9. A flexible connected unmanned airborne detection device, characterized in that, include: The grid model construction module is configured to acquire the operational area of the detection mission, divide and mark the operational area into grids, and determine the safe expansion radius of each grid to obtain a grid model representing the global terrain and known obstacles; wherein each grid is marked as either winged or no-fly. The path planning module is configured to perform path planning based at least on the grid model to obtain an initial flight plan; the initial flight plan includes at least an initial route. The static compensation module is configured to determine static compensation points based on the global terrain and known obstacles, pre-calculate the sling swing amount for each static compensation point, and pre-calculate the flight path offset based on the pre-calculated sling swing amount for the static compensation points. The static compensation module is also configured to use the pre-calculated route offset to perform static compensation update on the initial route, so as to obtain the statically compensated updated flight plan. The detection module is configured to control the unmanned aerial vehicle system to perform detection tasks according to the flight plan updated by the static compensation.
10. A flexible connected unmanned airborne detection system characterized by, Includes drones, flexible slings, detection payloads, and control units; The drone is equipped with an inertial measurement module, a positioning module, a lidar and a vision camera to collect its own status data and information about the surrounding environment in real time. The flexible sling is an integrated composite cable structure that integrates load power supply cables, radio frequency signal transmission cables, and structural support cables. Furthermore, the flexible sling is equipped with gyroscope modules and tension sensors to acquire real-time status data of the detected load. The control unit is configured to control the UAV-borne detection system to perform a detection task using the flexible connection UAV-borne detection method as described in any one of claims 1 to 8; or, the control unit includes the flexible connection UAV-borne detection device as described in claim 9.
