Timely cooperative control method and system of airborne lidar unmanned aerial vehicle for under-forest target searching and vegetation sheltering scene

Abstract

The application belongs to the technical field of unmanned aerial vehicle flight control, and discloses a kind of time sequence coordination control method and system of airborne laser radar unmanned aerial vehicle for under-forest target search and vegetation sheltered scene, solve the problem of time sequence mismatch between non-real-time perception calculation result and strong real-time flight control system, time delay uncertainty. Including multi-sensor acquisition module, intelligent perception and calculation module, flight control instruction generation module, flight control execution module and communication and state feedback module;Flight control instruction generation module is used for time-effectiveness judgment, state constraint and control mapping processing to the perception result, and the pose increment of flight state at the time stamp moment and current time flight state is generated using the perception result, the coordinate system compensation or prediction correction is carried out to the point cloud perception result, and the quasi-real-time mapping of non-real-time perception at real-time control moment is realized.The application realizes the safe decoupling of laser radar perception and flight control.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicle (UAV) flight control technology, and particularly relates to a time-series coordinated control method and system for airborne lidar UAVs in forest target search and vegetation cover scenarios. Background Technology

[0002] With the widespread application of drones in scenarios such as ecological monitoring, wildlife surveys, forestry inspections, emergency search and rescue, and complex terrain exploration, environmental perception technology based on airborne lidar is gradually becoming an important component of drone autonomous flight systems. Especially in forest areas, bamboo forests, and dense vegetation environments, airborne lidar has stronger spatial penetration capabilities than ordinary visual sensors, enabling it to acquire three-dimensional point cloud information of target areas under the forest canopy, providing effective data support for searching for understory animals, identifying activity areas, and analyzing terrain structure.

[0003] In giant panda habitat surveys and forest target searches, drones typically need to be equipped with lidar to continuously scan the forest floor and use edge computing modules to filter, cluster, segment, identify targets, extract obstacles, and locate suspected targets from point cloud data. However, the lidar point cloud perception computation in existing drone systems usually has nondeterministic latency. The computation process is affected by factors such as point cloud size, vegetation density, algorithm complexity, edge computing load, and system task scheduling, making it difficult to strictly predict the output time of the perception results.

[0004] The flight control system of UAVs is a strong real-time control system, and its attitude control, speed control and position control all rely on deterministic and periodic control command inputs. In the existing technical solutions, the perception computing module and the flight control module often adopt a direct or semi-direct data interaction method, and the perception results are directly used in flight decision or control logic. This method has the following problems in engineering practice: (1) The point cloud perception computing of airborne lidar has non-deterministic time delay, which makes it difficult to keep in line with the real-time control cycle of the flight controller. (2) There is no clear time constraint mechanism between the point cloud perception results and the flight control commands, which may lead to the use of outdated target position or obstacle information in control decisions. (3) In the complex environment of forest, if the delayed perception results are used for target tracking or obstacle avoidance, it is easy to lead to misjudgment, rescanning deviation or even safety collision risk. (4) Existing solutions mostly focus on lidar mapping, obstacle detection or animal target recognition itself, and lack specific technical means for "how non-real-time point cloud perception results can be safely used in the real-time flight control of UAVs".

[0005] Therefore, how to enable airborne lidar perception results with nondeterministic time delay to participate in UAV flight control safely and controllably, and serve forest target search and point cloud acquisition tasks, without compromising the real-time performance and stability of the flight control system, has become an urgent technical problem to be solved in the existing technology. Summary of the Invention

[0006] One objective of this invention is to provide an airborne lidar UAV timing-coordinated control system for forest target search and vegetation-covered scenarios, effectively solving the problems of timing mismatch, latency uncertainty, and safety hazards between non-real-time perception calculation results and strong real-time flight control systems.

[0007] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: an airborne lidar UAV time-series collaborative control system for forest target search and vegetation occlusion scenarios, including a multi-sensor acquisition module, an intelligent perception and computing module, a flight control command generation module, a flight control execution module, and a communication and status feedback module.

[0008] The flight control command generation module is used to perform timeliness judgment, state constraint and control mapping processing on the perception results output by the intelligent perception and computing module, and use the perception results to generate the pose increment of the flight state at the timestamp and the current flight state to perform coordinate system compensation or prediction correction on the point cloud perception results, so as to realize the pseudo-real-time mapping of non-real-time perception at the real-time control moment.

[0009] Furthermore, the multi-sensor acquisition module is used to collect forest environment perception data and UAV flight status data; the intelligent perception and computing module is used to process the laser point cloud data collected by the multi-sensor acquisition module, performing point cloud denoising, ground point separation, clustering and segmentation, obstacle detection and suspected forest target identification, and outputting perception results and corresponding timestamp information; the flight control execution module is used to receive and execute flight control commands output by the flight control command generation module, and complete attitude stabilization, speed control, position control and motor drive; the communication and status feedback module is used to establish a bidirectional communication link between the flight control execution module and the flight control command generation module, sending flight control commands to the flight control execution module and transmitting real-time flight status back to the flight control command generation module.

[0010] Furthermore, the flight control command generation module simultaneously subscribes to the nondeterministic point cloud perception stream from the intelligent perception and computing module, as well as the real-time state stream from the flight control execution module.

[0011] Furthermore, the flight control command generation module includes a sliding window state buffer and a state predictor; the sliding window state buffer is used to store the UAV's pose, velocity, and acceleration information for the most recent multiple cycles in real time; the state predictor is used to extrapolate and predict temporary flight control commands during the control cycle gap when the perception results have not yet been output, based on the perception correction value of the previous cycle and the current motion trend, to ensure the continuous output of flight control commands.

[0012] Another objective of this invention is to provide a time-series collaborative control method for airborne lidar UAVs in forest target search and vegetation occlusion scenarios. Based on the airborne lidar UAV time-series collaborative control system described in the above embodiments, the method includes the following steps: S1, collecting forest environment point cloud data and flight status data by a multi-sensor acquisition module.

[0013] S2. The intelligent sensing and computing module processes the laser point cloud data and outputs information on suspected forest targets, obstacles, or local spatial structures, along with corresponding timestamps.

[0014] S3. The flight control command generation module judges the timeliness of the sensing results and calculates the total sensing delay time. : ; like If so, the safety protection logic is triggered, executing actions such as maintaining the original course, decelerating cruise, or hovering control; According to Extrapolation compensation is performed using the current drone motion vector to correct the predicted position of suspected forest targets or obstacles at the current moment; where, Indicates the current moment. Indicates the time when the perception result is generated. This indicates the delay threshold.

[0015] S4. The compensated or predicted perception information is mapped into flight control commands, and the flight control commands are transmitted to the flight control execution module through the communication and status feedback module.

[0016] S5. The flight control execution module executes flight actions according to flight control commands, and at the same time transmits the current pose state stream back to the flight control command generation module at a fixed frequency through the communication and status feedback module.

[0017] Furthermore, during the control cycle gap before the perception results are output, the state predictor of the flight control command generation module is used to perform extrapolation calculations to generate temporary flight control commands, ensuring the continuous output of flight control commands.

[0018] Furthermore, in step S3, if When the perception result is output, the flight control command generation module extracts the timestamp carried by the perception result and retrieves the historical pose corresponding to the timestamp from the sliding window state buffer of the flight control command generation module. By comparing the historical pose with the real-time pose at the current moment, the cumulative displacement vector and attitude deviation generated during the perception calculation are calculated. Using the differential compensation algorithm, the perception result is projected and compensated from the timestamp to the current real-time moment to generate the corrected pseudo-real-time spatial features.

[0019] Furthermore, in step S5, the execution and feedback process performed by the flight control execution module after receiving the flight control command is as follows: S61, command reception and command reconstruction: The flight control execution module receives the flight control command in real time and performs secondary interpolation processing on the received flight control command.

[0020] S62. Real-time closed-loop execution: The flight control execution module takes the flight control command as the desired target, calculates the desired speed of each motor through the cascade control logic of the attitude loop, speed loop and position loop, and finally drives the motor through the ESC to complete the predetermined flight action.

[0021] S63. Real-time status feedback: While performing flight maneuvers, the flight control execution module transmits the current pose status stream back to the flight control command generation module at a fixed frequency through the communication and status feedback module.

[0022] S64. Closed-loop collaborative update: The returned state stream data is pushed into the sliding window state buffer maintained by the flight control command generation module in real time. The state stream data serves as the reference data for time delay backtracking and pose compensation in the next control cycle.

[0023] Compared with the prior art, the beneficial technical effects of the present invention are as follows: The present invention introduces a flight control command generation module with timing alignment mechanism, timeliness constraint and safety control logic between laser point cloud perception calculation and flight control, aiming to achieve the following technical effects.

[0024] (1) Improve the spatial positioning accuracy of forest target search: By introducing a pose increment compensation mechanism based on dual-rate sampling, the real-time pose with high-frequency feedback is used to spatially remap the low-frequency non-real-time point cloud perception results with timestamps. Even if the UAV moves continuously in the complex forest environment, it can still dynamically correct the position of suspected forest targets and obstacles, reduce the spatial deviation caused by the hundreds of milliseconds of perception delay, and improve the target area supplementation and positioning accuracy.

[0025] (2) Enhance the robustness of the system to fluctuations in edge computing power: By setting a timeliness judgment threshold mechanism for the flight control command generation module, the non-deterministic computing delays caused by sudden increases in point cloud size, complex vegetation obstruction, hardware overheating and frequency reduction, or fluctuations in system task scheduling are effectively isolated between the intelligent sensing and computing units. When the sensing delay exceeds the safety threshold, the system automatically triggers the safety protection logic to prevent expired point cloud sensing results from interfering with flight control and improve the system's safety in complex forest environments.

[0026] (3) Ensuring the continuity and smoothness of flight control: By utilizing the state predictor built into the flight control command generation module, predictive control commands are generated through first-order or higher-order extrapolation during the control cycle gap before the point cloud perception results are output. This mechanism can avoid the interruption of control commands caused by perception frame drops or point cloud processing delays, and ensure the trajectory smoothness and flight stability of the UAV during forest cruising, hovering observation, low-speed supplementary scanning, and obstacle avoidance flight.

[0027] (4) Achieving safe decoupling between lidar perception and flight control: Without changing the underlying strong real-time closed-loop control logic of the flight controller, a safe collaboration between the high-load, nondeterministic lidar point cloud perception algorithm and the high-frequency deterministic flight control logic is achieved. This scheme lowers the threshold for lidar target search, forest obstacle perception, and animal point cloud acquisition tasks to be integrated into the flight control system, and improves the engineering feasibility of UAVs in ecological surveys and autonomous operations in complex environments. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the connection structure of the airborne lidar UAV timing cooperative control system of the present invention.

[0029] Figure 2 This is a flowchart of the flight control command generation module in Example 1.

[0030] Figure 3 This is a schematic diagram of the hardware system connection structure in Example 1. Detailed Implementation

[0031] Example 1: This example provides a time-series collaborative control system for airborne lidar UAVs in forest target search and vegetation obscuring scenarios, such as... Figure 1 As shown, it includes a multi-sensor acquisition module, an intelligent sensing and computing module, a flight control command generation module, a flight control execution module, and a communication and status feedback module. This embodiment achieves secure access to the non-real-time sensing results of the airborne lidar by setting up a flight control command generation module between the airborne lidar point cloud perception calculation and the UAV flight control, without changing the underlying attitude calculation and strong real-time closed-loop control logic of the flight controller.

[0032] The flight control command generation module performs timeliness judgment, state constraint, and control mapping processing on the perception results output by the intelligent perception and computing module. It generates high-level flight control commands that can be accepted by the flight control execution module only when preset conditions are met, thereby avoiding the direct impact of non-real-time or abnormal perception results on the flight control system. The flight control command generation module not only performs timeliness judgment, but also uses the pose increment of the flight state at the time stamp generated by the perception results and the current time to perform coordinate system compensation or prediction correction on the point cloud perception results, realizing the near real-time mapping of non-real-time perception at the real-time control time.

[0033] In this embodiment, the UAV, equipped with an airborne LiDAR, performs understory target search and point cloud acquisition tasks in a forest area. The LiDAR continuously scans the understory area, and the intelligent perception and computing module filters, clusters, and identifies targets in the acquired point cloud, outputting the suspected understory target locations, spatial distribution of obstacles, and corresponding timestamps. The flight control command generation module generates high-level control commands such as deceleration, hovering, turning, re-scanning around, or maintaining the original heading based on the time delay and pose compensation results, to support safe flight and high-quality point cloud acquisition in the complex understory environment.

[0034] In this embodiment, the flight control command generation module maintains a sliding window state buffer and a state predictor. The sliding window state buffer is used to store the UAV's pose, velocity, and acceleration information for the most recent multiple cycles in real time. The state predictor is used to perform extrapolation calculations based on the perception correction value of the previous cycle and the current motion trend during the control cycle gap when the perception results have not yet been output. The extrapolation calculations include: using first-order extrapolation calculations containing only the velocity term under uniform linear motion conditions; and using second-order or higher-order extrapolation containing the acceleration term in combination with the acceleration information stored in the sliding window state buffer under conditions with strong nonlinear motion such as acceleration, deceleration, turning, or circling, in order to suppress the accumulation of prediction errors under nonlinear motion scenarios and generate temporary continuous flight control commands to ensure the continuous output of flight control commands.

[0035] like Figure 1As shown, the multi-sensor acquisition module is used to collect forest environment perception data and UAV flight status data, including LiDAR point cloud data, IMU data, and GPS data. The flight status data is mainly used by the flight control command generation module for pose compensation and timing alignment calculations. The intelligent perception and computing module is used to process the LiDAR point cloud data collected by the multi-sensor acquisition module, performing point cloud denoising, ground point separation, clustering and segmentation, obstacle detection, and suspected forest target identification, and outputting perception results and corresponding timestamp information. That is, the intelligent perception and computing module performs point cloud processing, target detection, Panda target analysis, and timestamp output. The flight control execution module is used to receive flight control commands from the flight control command generation module and complete attitude stabilization, speed control, position control, and motor drive. That is, the flight control execution module performs timeliness evaluation, attitude compensation, state constraints, and control mapping, while the flight control command generation module performs attitude control, speed control, and position control. The communication and status feedback module is used to establish a two-way communication link between the flight control execution module and the flight control command generation module, sending flight control commands to the flight control execution module and transmitting real-time flight status back to the flight control command generation module. That is, the communication and status feedback module performs data transmission and status feedback.

[0036] The UAV timing cooperative control method based on the airborne lidar UAV timing cooperative control system for forest target search and vegetation occlusion scenarios described in this embodiment includes the following steps: S1, collecting forest environment point cloud perception data and flight status data by a multi-sensor acquisition module. The perception data includes at least three-dimensional point cloud data collected by the airborne lidar, and the flight status data includes at least the UAV's pose, velocity, acceleration, and heading information.

[0037] S2. The intelligent sensing and computing module processes the laser point cloud data, performing point cloud denoising, ground point separation, clustering and segmentation, obstacle detection and suspected forest under-spot identification, and outputs the sensing results (suspected forest under-spot information, obstacle information or local spatial structure information) and the corresponding timestamp information.

[0038] During the control cycle gap before the perception results are output, the state predictor of the flight control command generation module is used to perform first-order extrapolation calculation, or second-order or higher-order extrapolation is performed in combination with acceleration information under conditions of strong motion nonlinearity to generate temporary continuous flight control commands, ensuring the continuous output of flight control commands.

[0039] S3. The flight control command generation module judges the timeliness of the sensing results and calculates the total sensing delay time. : ; like If the flight path is not cleared, the safety protection logic is triggered, and the flight path is maintained, cruise is decelerated, or hovering is executed. These three options are selected by the flight control command generation module based on the environmental state reflected by the most recent valid perception result: when the most recent valid perception result indicates that the area ahead is open, the flight path is maintained or cruise is decelerated; when the most recent valid perception result indicates that there are obstacles, narrow passages, or spatial constraints ahead, hovering is selected first to prevent outdated perception results from affecting flight control.

[0040] like According to Extrapolation compensation is performed by combining the current drone motion vector to correct the predicted position of suspected targets or obstacles in the forest at the current moment.

[0041] In the formula, This indicates the current time, which is the system time when the flight control command generation module makes judgments and calculations, and is used to determine the delay reference. The timestamp indicates the moment when the sensing result is generated, that is, when the intelligent sensing and computing module outputs the sensing result, it identifies the actual collection time corresponding to the sensing result; This represents the delay threshold, used to determine whether the sensing result has expired or timed out.

[0042] S4. The compensated or predicted perception information is mapped into high-level flight control commands (such as desired velocity vector, desired position point, desired heading angle, or rescan path command) acceptable to the flight control execution module, and the flight control commands are transmitted to the flight control execution module through the communication and status feedback module.

[0043] When the compensated perception results indicate the presence of a suspected forest understory target ahead, the flight control command generation module can generate deceleration, hovering, yaw adjustment, low-speed detour, or target area supplementation control commands to improve the point cloud sampling density of the target area; when the compensated perception results indicate the presence of an obstacle ahead, the flight control command generation module can generate steering obstacle avoidance, lateral drift, ascent and descent obstacle avoidance, or deceleration and hovering control commands.

[0044] S5. The flight control execution module executes flight actions according to flight control commands, and at the same time transmits the current pose state stream back to the flight control command generation module at a fixed frequency through the communication and status feedback module.

[0045] The invention will now be described in further detail from the aspects of hardware system, software system and operating environment, intelligent perception, flight control command generation and timing alignment, and flight control execution and feedback.

[0046] (a) Hardware system.

[0047] In this embodiment, at the hardware level, such as Figure 3 As shown, the UAV platform integrates a Jetson Orin Nano 8GB edge AI computing unit or other equivalent edge AI computing units as the core unit for non-real-time perception computing and flight control command generation. The flight controller serves as the real-time flight control execution unit, and the two work collaboratively through a physical interface. The Jetson Orin Nano 8GB is primarily responsible for laser point cloud preprocessing, target discrimination, control intent generation, and communication with the flight controller, while the flight controller is only responsible for highly real-time tasks such as attitude stabilization, speed control, and motor drive.

[0048] like Figure 3 As shown, the UAV's front end is equipped with an onboard LiDAR (including a LiDAR sensor), and can optionally be equipped with an RGB camera, thermal imaging camera, depth camera, or binocular camera as auxiliary sensors to acquire 3D structural information and image-aiding information of the forest environment. The LiDAR and auxiliary sensors are connected to the Jetson Orin Nano 8GB via USB, Ethernet, serial port, or CSI interface, and sensor data is continuously input to the edge computing platform at a fixed frequency. The flight controller establishes a communication channel with the Jetson Orin Nano 8GB via serial port, CAN bus, or USB interface to receive flight control commands from the edge computing platform and provide real-time flight status information.

[0049] The power management module provides power to the Jetson Orin Nano 8GB, the airborne LiDAR, and the flight controller, preventing power interference from affecting the stability of the intelligent sensing and computing modules and ensuring system safety from a hardware perspective.

[0050] (ii) Software system.

[0051] In this embodiment, at the software level, the Jetson Orin Nano 8GB runs the Ubuntu 22.04 operating system and deploys ROS2 Humble as system middleware to achieve decoupled communication between functional nodes. The LiDAR driver node is responsible for collecting point cloud data and publishing it to the ROS2 communication bus, while the perception computing node subscribes to the point cloud data and performs point cloud preprocessing, target detection, clustering and segmentation, and obstacle recognition tasks.

[0052] The flight controller runs PX4 or ArduPilot firmware, maintaining its original attitude calculation and flight control logic without modification. The Jetson Orin Nano 8GB establishes a communication connection with the flight controller via MAVROS or MAVLink communication nodes, enabling bidirectional transmission of control commands and flight status information.

[0053] Upon system startup, the Jetson Orin Nano 8GB first initializes the operating system and ROS2 environment, then starts the lidar driver node, status feedback node, and flight control communication node, establishing a communication connection with the flight controller. After the communication link stabilizes, the perception computing node and control command generation node start sequentially, and the system enters normal operation.

[0054] (III) Intelligent perception.

[0055] During flight, the Jetson Orin Nano 8GB continuously receives point cloud data from the airborne LiDAR. The perception computing node first preprocesses the point cloud, including time synchronization, coordinate transformation, noise reduction filtering, outlier removal, and point cloud downsampling. Subsequently, it performs ground point separation, vegetation structure extraction, obstacle recognition, and clustering and discrimination of suspected understory targets on the processed point cloud.

[0056] The perception model can be a rule-based point cloud clustering algorithm, a deep learning-based 3D object detection model, a semantic segmentation model, or a multi-sensor fusion recognition model. Its inference process runs on a graphics processing unit (GPU) and can be optimized using TensorRT or other acceleration frameworks. Due to the unpredictable computational latency in the point cloud perception calculation process, the system adds a unique timestamp to each frame of perception result when it is generated to identify the actual acquisition time corresponding to that perception result.

[0057] When the perception results indicate the presence of suspected forest targets or high-risk obstacles in the forest area, the system will not send the results directly to the flight controller. Instead, it will use them as input data for the flight control command generation module, which will then perform timeliness judgment, pose compensation, and control mapping.

[0058] (iv) Flight control command generation and timing alignment.

[0059] In this embodiment, the Jetson Orin Nano 8GB has an independent flight control command generation module, such as... Figure 2 As shown, the flight control command generation module adopts a dual-rate sampling collaborative mechanism to solve the timing mismatch problem between non-real-time point cloud perception results (low frequency, non-deterministic delay) and flight control execution (high frequency, deterministic cycle).

[0060] Dual-rate data input: such as Figure 2As shown, the flight control command generation module simultaneously subscribes to a low-frequency nondeterministic point cloud perception stream (low-frequency point cloud data) from the intelligent sensing and computing module, and a high-frequency (200Hz) real-time state stream (high-frequency state data) from the flight control execution module. By introducing a pose increment compensation mechanism based on dual-rate sampling, the real-time pose feedback from the high-frequency feedback is used to spatially remap the timestamped low-frequency non-real-time perception results. This enables the UAV to eliminate spatial position deviations caused by hundreds of milliseconds of perception delay even at high-speed flight (e.g., above 5 m / s), improving the accuracy of target area rescanning and positioning.

[0061] State Predictor and Buffer Management: The flight control command generation module maintains a sliding window state buffer to store the UAV's attitude, velocity, and acceleration information for the most recent cycles in real time.

[0062] Delay backtracking and differential compensation: When the perception result is output, the flight control command generation module extracts the timestamp it carries and retrieves the historical pose corresponding to the timestamp from the sliding window state buffer. By comparing the current real-time pose with the historical pose, the cumulative displacement vector and attitude deviation generated during the perception calculation are calculated. Using a high-order differential compensation algorithm, the perception result (suspected forest target location, relative coordinates of obstacles, or local point cloud features) is projected and compensated from the timestamp to the current real-time moment, generating corrected pseudo-real-time spatial features.

[0063] Smoothing of commands during idle periods: During the control cycle gap before the point cloud perception results are output, the flight control command generation module activates the state predictor. Based on the perception correction value of the previous cycle and the current motion trend, it performs extrapolation calculations to ensure continuous output of flight control commands. This effectively eliminates control command abrupt changes caused by dropped perception frames, avoids attitude jerking during UAV flight, ensures the smoothness and stability of the flight trajectory, and guarantees the smoothness and flight stability of the UAV during forest cruising, hovering observation, low-speed rescanning, and obstacle avoidance.

[0064] Control mode switching: When the compensated perception results indicate that there is a suspected forest target ahead, the flight control command generation module can generate deceleration, hovering, yaw adjustment, low-speed flyaround, or target area supplementation control commands to improve the point cloud sampling density of the target area; when the compensated perception results indicate that there is an obstacle ahead, the command generation module can generate steering obstacle avoidance, lateral offset, ascent and descent obstacle avoidance, or deceleration and hovering control commands.

[0065] Security closed loop: If the perceived latency (timestamp delay) exceeds a preset dynamic threshold The dynamic threshold is based on the current flight speed of the drone. Effective range of perception according to The relationship is adjusted online when the drone is hovering at low speed or in a re-sweeping state. When it approaches 0, This can be relaxed when in high-speed cruising conditions. Automatic tightening ensures that spatial position deviations caused by sensing delays do not exceed a safety margin. The flight control command generation module automatically determines that the perception branch has failed and switches to inertial maintenance logic based on real-time status feedback (i.e., switches to safety mode) to ensure the underlying security of the system when point cloud perception computing power fluctuates or target discrimination delay increases. By setting a timeliness judgment threshold mechanism, the non-deterministic computing delay caused by hardware overheating and frequency reduction, instantaneous increase in model load, or system task scheduling in the edge artificial intelligence computing unit (i.e., intelligent perception and computing module) is effectively isolated. When the perception delay exceeds the safety threshold, the system can automatically trigger safety protection logic to prevent expired perception data from interfering with flight control, significantly improving the system's survivability in complex edge computing environments.

[0066] (v) Flight control execution and feedback.

[0067] In this embodiment, the flight control execution module (i.e., the flight controller) and the flight control command generation module form a high-frequency, collaborative closed-loop control system. The specific execution and feedback process is as follows.

[0068] (1) Command reception and command reconstruction: The flight control execution module receives high-level flight control commands (such as desired velocity vector, desired position point, desired heading angle, or target area rescan trajectory points) from the Jetson Orin Nano 8GB in real time. In order to ensure the smoothness of flight, the command parser inside the flight controller will perform secondary interpolation processing on the received asynchronous high-level commands to match the operating frequency of the underlying PID control algorithm (such as 200Hz or higher).

[0069] (2) Real-time closed-loop execution: Based on its own IMU (Inertial Measurement Unit), barometer, GNSS and other navigation sensor data, the flight controller takes the upper-level flight control commands as the desired target, and calculates the desired speed of each motor through the cascade control logic of attitude loop, velocity loop and position loop. Finally, the electronic speed controller (ESC) drives the motor to complete the predetermined flight maneuver. During the execution, the flight controller always maintains its deterministic real-time control priority. If the upper-level command is interrupted, it automatically switches to inertial stabilization mode or safe hovering mode.

[0070] (3) Real-time status feedback (core feedback loop): While performing flight maneuvers, the flight controller transmits the current high-precision pose status stream (including attitude, three-axis velocity, three-axis acceleration and relative position information) to the flight control command generation module of Jetson Orin Nano8GB through the communication and status feedback module at a fixed frequency (e.g. 50Hz-100Hz) via the MAVLink protocol.

[0071] (4) Closed-loop collaborative update: The returned state flow data will be pushed into the sliding window state buffer maintained by the flight control command generation module in real time. This state flow data serves as the reference data for time delay backtracking and pose compensation in the next control cycle. Through this closed-loop mechanism of "command issuance - state return - pose remapping", it is ensured that even if there is nondeterministic time delay in the airborne lidar perception calculation, the system can still dynamically correct the perception results based on the real physical motion trajectory.

[0072] In summary, this embodiment clearly separates the non-real-time point cloud perception calculation of airborne lidar from the strong real-time flight control in its system structure, and achieves safe collaboration between the two through flight control command generation and timing alignment mechanisms. The uncertainty of point cloud processing, target detection, or deep learning inference directly impacts the real-time performance of flight control, improving the stability, safety, and reliability of the UAV in complex forest environments. Furthermore, the hardware platform, communication interface, and software framework used in this embodiment are all mature and widely applied technical solutions. The system structure is clear, the module boundaries are well-defined, and it can be directly deployed on existing UAV platforms, exhibiting good engineering feasibility and scalability. Especially in forest target search and point cloud acquisition tasks, this invention can improve the completeness and data validity of point cloud acquisition in suspected target areas while ensuring flight safety, demonstrating strong application value.

[0073] Of course, the above description is not intended to limit the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A time-series collaborative control system for airborne lidar-based UAVs in forest understory target search and vegetation-covered scenarios, characterized in that, It includes a multi-sensor acquisition module, an intelligent sensing and computing module, a flight control command generation module, a flight control execution module, and a communication and status feedback module; The flight control command generation module is used to perform timeliness judgment, state constraint and control mapping processing on the perception results output by the intelligent perception and computing module, and use the perception results to generate the pose increment of the flight state at the timestamp and the current flight state to perform coordinate system compensation or prediction correction on the point cloud perception results, so as to realize the pseudo-real-time mapping of non-real-time perception at the real-time control moment.

2. The airborne lidar UAV timing-coordinated control system for forest target search and vegetation-covered scenarios as described in claim 1, characterized in that, The multi-sensor acquisition module is used to collect forest environment perception data and UAV flight status data; The intelligent sensing and computing module is used to process the laser point cloud data collected by the multi-sensor acquisition module, perform point cloud denoising, ground point separation, clustering and segmentation, obstacle detection and suspected forest under-spot identification, and output the sensing results and corresponding timestamp information. The flight control execution module is used to receive and execute flight control commands output by the flight control command generation module, and to complete attitude stabilization, speed control, position control and motor drive. The communication and status feedback module is used to establish a two-way communication link between the flight control execution module and the flight control command generation module, send flight control commands to the flight control execution module, and send real-time flight status back to the flight control command generation module.

3. The airborne lidar UAV timing-coordinated control system for forest target search and vegetation-covered scenarios as described in claim 1, characterized in that, The flight control command generation module simultaneously subscribes to the nondeterministic point cloud perception stream from the intelligent perception and computing module, as well as the real-time state stream from the flight control execution module.

4. The airborne lidar UAV timing-coordinated control system for forest target search and vegetation-covered scenarios as described in claim 1, characterized in that, The flight control command generation module includes a sliding window state buffer and a state predictor; The sliding window state buffer is used to store the UAV pose, velocity, and acceleration information of the most recent multiple cycles in real time; The state predictor is used to extrapolate and extrapolate based on the perception correction value of the previous cycle and the current motion trend during the control cycle gap before the perception results are output, to predict temporary flight control commands and ensure the continuous output of flight control commands.

5. A time-series cooperative control method for airborne lidar UAVs in forest understory target search and vegetation-covered scenarios, characterized in that, The airborne lidar UAV timing cooperative control system based on any one of claims 1-4 includes the following steps: S1. Forest understory point cloud data and flight status data are collected by a multi-sensor acquisition module; S2. The intelligent sensing and computing module processes the laser point cloud data and outputs information on suspected forest targets, obstacles, or local spatial structures, along with corresponding timestamps. S3. The flight control command generation module judges the timeliness of the sensing results and calculates the total sensing delay time. : ； like If so, the safety protection logic is triggered, executing actions such as maintaining the original course, decelerating cruise, or hovering control; According to Extrapolation compensation is performed by combining the current drone motion vector to correct the predicted position of suspected targets or obstacles in the forest at the current moment; In the formula, Indicates the current moment. Indicates the time when the perception result is generated. Indicates the delay threshold; S4. Map the compensated or predicted perception information into flight control commands, and transmit the flight control commands to the flight control execution module through the communication and status feedback module. S5. The flight control execution module executes flight actions according to flight control commands, and at the same time transmits the current pose state stream back to the flight control command generation module at a fixed frequency through the communication and status feedback module.

6. The time-series cooperative control method for airborne lidar UAVs in forest target search and vegetation-covered scenarios according to claim 5, characterized in that, During the control cycle gap before the perception results are output, the state predictor of the flight control command generation module is used to perform extrapolation calculations to generate temporary flight control commands, ensuring the continuous output of flight control commands.

7. The time-series cooperative control method for airborne lidar UAVs in forest target search and vegetation-covered scenarios according to claim 5, characterized in that, In step S3, if When the perception result is output, the flight control command generation module extracts the timestamp carried by the perception result and retrieves the historical pose corresponding to the timestamp from the sliding window state buffer of the flight control command generation module. By comparing the historical pose with the real-time pose at the current moment, the cumulative displacement vector and attitude deviation generated during the perception calculation are calculated. Using a differential compensation algorithm, the perception result is projected and compensated from the timestamp to the current real-time time to generate a corrected pseudo-real-time spatial feature.

8. The time-series cooperative control method for airborne lidar UAVs in forest target search and vegetation-covered scenarios according to claim 7, characterized in that, In step S5, the execution and feedback process of the flight control execution module after receiving the flight control command is as follows: S61. Command Reception and Command Reconstruction: The flight control execution module receives flight control commands in real time and performs secondary interpolation processing on the received flight control commands; S62, Real-time Closed-Loop Execution: The flight control execution module takes the flight control command as the desired target, calculates the desired speed of each motor through the cascade control logic of the attitude loop, speed loop and position loop, and finally drives the motor through the ESC to complete the predetermined flight action; S63. Real-time status feedback: While performing flight maneuvers, the flight control execution module transmits the current pose status stream back to the flight control command generation module at a fixed frequency through the communication and status feedback module. S64. Closed-loop collaborative update: The returned state stream data is pushed into the sliding window state buffer maintained by the flight control command generation module in real time. The state stream data serves as the reference data for time delay backtracking and pose compensation in the next control cycle.

Images