Unmanned aerial vehicle route generation system

Abstract

The invention discloses an unmanned aerial vehicle route generation system which comprises a generation system body, the generation system body comprises a three-dimensional terrain modeling module, a multi-source sensor fusion module, an intelligent task planning module, a dynamic obstacle avoidance and safety module and a man-machine interaction and data management module, and the three-dimensional terrain modeling module is based on ASTER GDEM and DSM elevation data; the multi-source sensor fusion module is used for generating a three-dimensional digital elevation model (DEM) of an operation area, centimeter-level terrain matching is realized by combining an RTK positioning technology, and the multi-source sensor fusion module is used for integrating a laser radar, a visible light camera and an infrared sensor, collecting features of ground objects such as towers and vegetation in real time, and is used for dynamically adjusting an air route. The optimal route is automatically generated through an intelligent algorithm, manual intervention is reduced, route planning efficiency is remarkably improved, meanwhile, terrain adaptability can be enhanced, a terrain following algorithm is adopted, the flight height is automatically adjusted according to topographic relief, and accuracy and integrity of surveying and mapping data are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of route generation systems, and in particular to a UAV route generation system. Background Art

[0002] With the rapid development of drone technology, its applications in various fields, including military reconnaissance, civil mapping, commercial logistics, and agricultural plant protection, are becoming increasingly widespread. However, existing drone route planning methods have exposed many limitations in practical applications and are unable to meet complex and changing mission requirements and environmental conditions.

[0003] Traditional drone route planning methods mostly use fixed-altitude flight modes, which cannot adapt to complex terrain and can easily lead to the following problems:

[0004] Poor terrain adaptability: In mountainous and hilly areas with large elevation differences, fixed-altitude flight can easily cause "aircraft explosion" accidents or uneven aerial data resolution;

[0005] Low mission efficiency: For complex structures such as towers and bridges, manual intervention is required to adjust the flight path, which is time-consuming and lacks accuracy.

[0006] High data redundancy: Traditional route planning fails to take into account mission requirements (such as photographing distance and component characteristics), resulting in increased invalid data collection.

[0007] Insufficient safety: Lack of real-time obstacle avoidance capabilities for obstacles such as high-voltage lines and buildings, posing a risk of collision.

[0008] Existing technologies, such as patent CN202510019309.3, propose a route generation method based on tower components, but this method fails to address the issue of 3D terrain adaptation. Patent CN119472739B uses point cloud data to classify tower types, but relies on manual point puncture, resulting in limited automation. Therefore, a drone route generation system integrating terrain adaptation, mission optimization, and dynamic obstacle avoidance is urgently needed. Summary of the Invention

[0009] The purpose of the present invention is to provide a drone route generation system to solve the problems raised in the above background technology.

[0010] To achieve the above-mentioned object, the present invention provides the following technical solution: a UAV route generation system, comprising a generation system, wherein the generation system includes a three-dimensional terrain modeling module, a multi-source sensor fusion module, an intelligent mission planning module, a dynamic obstacle avoidance and safety module, a human-computer interaction and data management module, and a processor.

[0011] The 3D terrain modeling module generates a 3D digital elevation model (DEM) of the operation area based on ASTER GDEM and DSM elevation data, and combines RTK positioning technology to achieve centimeter-level terrain matching.

[0012] The multi-source sensor fusion module includes integrated laser radar, visible light camera and infrared sensor, which collects the characteristics of ground objects such as towers and vegetation in real time for dynamic adjustment of flight routes.

[0013] The intelligent mission planning module automatically generates the optimal route according to the mission type (such as inspection, mapping, and plant protection), and supports multi-mode switching such as terrain-simulating flight, circling route, and figure-8 route.

[0014] The dynamic obstacle avoidance and safety module combines deep learning algorithms to identify obstacles such as high-voltage lines and trees in real time, automatically generates obstacle avoidance paths and updates flight routes.

[0015] Preferably, the three-dimensional terrain modeling module includes a multi-source data fusion module: integrating ASTER GDEM data: global 30-meter resolution elevation data, used for preliminary modeling of large-scale terrain; DSM (digital surface model): containing surface information such as vegetation and buildings, generated through drone aerial photography or satellite images; RTK real-time positioning data: centimeter-level precision positioning, used for dynamic terrain correction and drone real-time positioning.

[0016] Preferably, the multi-source sensor fusion module includes a sensor calibration and spatiotemporal synchronization module, and a feature extraction and fusion module.

[0017] Preferably, the intelligent mission planning module includes the following route generation algorithm, a terrain-mimicking flight mode: generating a route parallel to the terrain based on DEM data, and dynamically adjusting the flight altitude according to the terrain fluctuations;

[0018] Circular route mode: For cylindrical structures such as towers and bridges, a spiral ascending or descending circular route is generated to ensure 360° coverage without blind spots;

[0019] Figure-8 route mode: Suitable for large-area vegetation monitoring, the figure-8 trajectory reduces duplicate coverage and improves operational efficiency.

[0020] Preferably, the dynamic obstacle avoidance and safety module includes an obstacle detection module.

[0021] Preferably, the human-computer interaction and data management module includes a human-computer interactive operation interface.

[0022] Preferably, the processor model is NVIDIA Jetson AGX Orin.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] The present invention uses three-dimensional terrain modeling and terrain-simulating flight technology to reduce the risk of "aircraft explosion" and improve the consistency of aerial data resolution; optimize mission efficiency: the multi-task optimization algorithm shortens the route generation time by 60% and reduces the amount of invalid data collection by 40%; enhance safety: the dynamic obstacle avoidance mechanism reduces the collision accident rate by 90%, meeting the needs of operations in complex environments; and lower the operating threshold: users only need to input mission parameters, and the system automatically generates the optimal route without the need for professional training. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a system block diagram of a UAV route generation system of the present invention. DETAILED DESCRIPTION

[0026] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0027] See also Figure 1 The present invention provides a technical solution: a UAV route generation system, comprising a generation system, wherein the generation system includes a three-dimensional terrain modeling module, a multi-source sensor fusion module, an intelligent task planning module, a dynamic obstacle avoidance and safety module, a human-computer interaction and data management module, and a processor.

[0028] The 3D terrain modeling module generates a 3D digital elevation model (DEM) of the operation area based on ASTER GDEM and DSM elevation data, and combines RTK positioning technology to achieve centimeter-level terrain matching.

[0029] The multi-source sensor fusion module includes integrated laser radar, visible light camera and infrared sensor, which collects the characteristics of ground objects such as towers and vegetation in real time for dynamic route adjustment.

[0030] The intelligent mission planning module automatically generates the optimal route according to the mission type (such as inspection, mapping, and plant protection), and supports multi-mode switching such as terrain-simulating flight, circling route, and figure-8 route.

[0031] The dynamic obstacle avoidance and safety module combines deep learning algorithms to identify obstacles such as high-voltage lines and trees in real time, automatically generates obstacle avoidance paths and updates flight routes.

[0032] The 3D terrain modeling module is used to generate a high-precision 3D terrain model of the operating area, providing basic data support for terrain-simulating flight and dynamic route adjustment. It includes a multi-source data fusion module: ASTER GDEM data: global 30-meter resolution elevation data for preliminary modeling of large-scale terrain; DSM (digital surface model): including surface information such as vegetation and buildings, generated through drone aerial photography or satellite imagery; RTK real-time positioning data: centimeter-level precision positioning, used for dynamic terrain correction and real-time drone positioning, using a bilateral filtering algorithm: while retaining terrain features (such as ridges and gullies), it eliminates invalid elevation points (such as trees and temporary buildings); dynamically adjusts the filter window size according to terrain complexity (such as 5×5 meters for flat areas and 3×3 meters for complex areas); B-spline curve fitting: smoothes the generated route to reduce altitude changes and sharp turns, improving flight stability, and divides the operating area into 1m×1m grids, extracting the coordinates and elevation information of each grid center point; generates a continuous 3D terrain surface through point cloud interpolation algorithms (such as Kriging interpolation), supporting dynamic terrain matching.

[0033] The multi-source sensor fusion module integrates lidar, visible light camera, infrared sensor and millimeter wave radar to collect terrain features and environmental information in real time, supporting dynamic route adjustment and obstacle avoidance decisions.

[0034] Sensor calibration and spatiotemporal synchronization: Spatial calibration: The coordinate system error of the camera and lidar is eliminated through Zhang Zhengyou calibration method (chessboard calibration); Time synchronization: Hardware triggering (such as GPS PPS signal) is aligned with software timestamp to ensure that the time error of multi-sensor data is ≤10ms.

[0035] Feature extraction and fusion:

[0036] LiDAR point cloud processing: Identify poles, towers, bridges, vegetation and other terrain features through point cloud segmentation algorithms (such as RANSAC and Euclidean clustering); Visible light camera image analysis: Detect component defects (such as broken insulators and loose bolts) in combination with deep learning models (such as YOLOv8); Infrared sensor thermal imaging: Identify abnormal high-voltage line temperatures (such as triggering an alarm when exceeding 80°C); Millimeter-wave radar obstacle avoidance: Detect low-altitude obstacles (such as kites and drones), with a detection range of ≥50m and an accuracy of ≤0.1m.

[0037] Data Fusion Algorithm: Kalman Filter: Fusion of LiDAR and Visual SLAM data to improve positioning accuracy; Bayesian Network: Combines probabilistic models of multi-sensor data to optimize obstacle detection and classification accuracy. Module Output: Real-time ground feature data (such as tower component coordinates, vegetation height, and obstacle location); Environmental Perception Report (including temperature anomalies, obstacle type, and distance).

[0038] The dynamic obstacle avoidance and safety module identifies obstacles in real time and generates avoidance paths, ensuring safe flight in complex environments. Obstacle detection and classification: Visual SLAM: Builds a local 3D map using the ORB-SLAM3 algorithm, identifying static obstacles (such as trees and buildings) in real time. Deep learning detection: Utilizes the YOLOv8 model to detect dynamic obstacles (such as birds and drones) with an accuracy of ≥95%. Millimeter-wave radar detection: Detects low-altitude obstacles (such as kites and power lines) with a detection range of ≥50m and an accuracy of ≤0.1m.

[0039] Obstacle avoidance path planning: AI algorithm: Generates safe paths around obstacles, prioritizing the shortest path with minimal altitude change; RRT (Rapidly Exploring Random Trees) algorithm: Randomly generates feasible paths and optimizes them for complex obstacle environments; Dynamic Window Algorithm (DWA): Combined with the UAV's dynamic constraints, adjusts the speed and direction of the obstacle avoidance path in real time.

[0040] Safety mechanism and redundancy design:

[0041] Hovering and Return: When the obstacle distance is less than a safety threshold (e.g., 10 meters), the hovering or return program is automatically triggered. Backup path switching: When the primary obstacle avoidance path fails, it automatically switches to a backup path (e.g., a pre-generated redundant route). Emergency landing: When the battery is low or the system fails, it automatically selects a safe area for landing. Real-time obstacle avoidance path data (including obstacle avoidance point coordinates, altitude adjustment, speed instructions); safety status report (including obstacle type, distance, and obstacle avoidance action).

[0042] Working principle: UAV route generation system operation process:

[0043] 1. System initialization and data loading (1) Hardware self-test processor module: The main control processor starts the system and loads the CUDA / TensorRT acceleration library; initializes the hardware acceleration module (point cloud processing, Kalman filter). Sensor calibration: The RTK module communicates with the flight control system to confirm that the positioning accuracy is ≤0.02m (1σ); the drone camera (RGB / multispectral) completes the exposure time and white balance calibration. (2) Task parameter configuration User input: Operation area boundary (KML / GeoJSON format); Task type (power inspection / topographic mapping / emergency monitoring); Key parameters: Inspection altitude (3-10m in mountainous areas, 20-50m in plains); Photo overlap rate (80% heading, 70% sideways); Maximum climb / descent rate (3m / s).

[0044] 2. Multi-source terrain data collection and preprocessing

[0045] (1) Data acquisition real-time data stream: RTK module: output centimeter-level positioning data at a frequency of 10 Hz (NMEA-0183 protocol); UAV sensor: LiDAR: point cloud density ≥ 100 points / m 2 , point frequency ≥ 100kHz; oblique camera: 5-lens synchronous acquisition (forward, rear, left, right, and downward). Historical data loading: load ASTERGDEM (30m resolution) and historical terrain point clouds (LAS format) from local storage.

[0046] (2) Data preprocessing (FPGA acceleration): Step 1: RTK data preprocessing timestamp alignment: align the RTK data with the camera exposure time (error ≤ 10ms); coordinate conversion: WGS84 → UTM projection (EPSG:32650). Step 2: Point cloud denoising statistical filtering: remove outliers (distance mean > 3σ); downsampling: voxel grid filtering (VoxelGrid), retain representative points (target density 50 points / m 2 ).

[0047] 3. 3D Terrain Modeling (Master Processor + FPGA Collaboration) (1) Terrain Filtering and Feature Extraction (FPGA Acceleration) Cloth Simulation Filter (CSF): Parameters: Grid resolution 0.5m, Cloth stiffness 0.1; Output: Pure ground point cloud (non-ground points are marked as vegetation / buildings). Key Feature Extraction: Slope / Aspect Calculation: Use the third-order inverse distance weighted (IDW) method to generate a slope classification map (flat / gentle / steep).

[0048] (2) 3D Terrain Reconstruction (GPU Acceleration) DEM Generation: Interpolation Algorithm: Kriging, with range parameters adaptively adjusted based on terrain complexity; Resolution: 1m×1m for inspection tasks, 5m×5m for surveying and mapping tasks. DSM and DEM Fusion: Weighted Average: For complex areas (e.g., cities), DSM weighting is 0.7, and DEM weighting is 0.3; Output: Fusion Terrain Model (GeoTIFF format, including slope / aspect layers).

[0049] (3) Dynamic Terrain Update (Real-time Closed-Loop) RTK Data Correction: Kalman Filter: Fusion of historical terrain models and real-time RTK data to update local terrain elevation; Latency: ≤150ms (from data reception to model update). Terrain Change Detection: Point Cloud Registration: The ICP algorithm calculates the transformation matrix between the historical point cloud and the current point cloud; Change Area Marking: Areas with a Hausdorff distance >0.5m are marked as "terrain changes."

[0050] 4. Route Generation Algorithm (Executed by the Main Control Processor) (1) Global Route Planning (Based on 3D Terrain Model) Step 1: Gridding the Operation Area The operation area is divided into 100m×100m grids, with each grid associated with a terrain complexity (simple / medium / complex). Step 2: Adaptive Flight Strip Spacing Calculation Formula:

[0051] [S=S_0\times(1+0.2\times C)]

[0052] (S_0): Reference strip spacing (10m for inspection, 50m for surveying and mapping);

[0053] (C): Terrain complexity coefficient (simple 0, medium 0.5, complex 1).

[0054] Step 3: Generate the initial route. Use AI algorithm to plan the shortest path and avoid known obstacles (such as buildings and high-voltage towers). Output: initial route (Waypoint list, including latitude and longitude, altitude, and heading). (2) Local route optimization (dynamic obstacle avoidance) Step 1: Real-time obstacle detection Input: LiDAR point cloud (current frame), fused terrain model; Method: Cluster analysis (DBSCAN): Identify dynamic obstacles (such as vehicles and pedestrians); Height estimation: Combine point cloud and terrain model to calculate the actual height of obstacles. Step 2: Detour path generation. Use RRT* algorithm to generate obstacle avoidance path. Constraints: Minimum turning radius: ≥5m (based on the UAV dynamics model); Maximum climb / descent rate: ≤3m / s. Step 3: Route smoothing B-spline curve fitting: Eliminate route inflection points to ensure flight stability; Output: Optimized route (G code format, supports flight control system analysis).

[0055] 5. Route Verification and Execution (1) Simulation Verification Digital Twin Environment: Import the generated route into the Gazebo simulation platform, overlay the 3D terrain model and dynamic obstacles; Verification indicators: Collision risk: 0 times / 100km route; Mission coverage: ≥98%.

[0056] (2) Real-time communication between flight control system and flight control system: route data (waypoint list, G code) is transmitted to the UAV flight control system via Gigabit Ethernet; breakpoint resume is supported: the executed route points are recorded and the flight can be resumed from the breakpoint after power failure and restart.

[0057] (3) Dynamically adjust the online learning mechanism: record actual flight data (such as RTK trajectory and photo success rate); optimize subsequent route parameters (such as flight strip spacing and photo overlap rate) through reinforcement learning (PPO algorithm).

[0058] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0059] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A UAV route generation system, comprising a generation system, characterized in that: The generation system includes a three-dimensional terrain modeling module, a multi-source sensor fusion module, an intelligent task planning module, a dynamic obstacle avoidance and safety module, a human-computer interaction and data management module, and a processor. The 3D terrain modeling module generates a 3D digital elevation model of the operating area based on ASTER GDEM and DSM elevation data, and combines RTK positioning technology to achieve centimeter-level terrain matching. The multi-source sensor fusion module includes integrated laser radar, visible light camera and infrared sensor, which collects the characteristics of ground objects such as towers and vegetation in real time for dynamic route adjustment. The intelligent mission planning module automatically generates the optimal route according to the mission type, and supports multi-mode switching such as terrain imitation flight, circling route, and figure-8 route. The dynamic obstacle avoidance and safety module combines deep learning algorithms to identify obstacles such as high-voltage lines and trees in real time, automatically generates obstacle avoidance paths and updates flight routes.

2. The UAV route generation system according to claim 1, characterized in that: The 3D terrain modeling module includes a multi-source data fusion module: ASTER GDEM data: global 30-meter resolution elevation data, used for preliminary modeling of large-scale terrain; DSM: contains surface information such as vegetation and buildings, generated through drone aerial photography or satellite imagery; RTK real-time positioning data: centimeter-level precision positioning, used for dynamic terrain correction and real-time drone positioning.

3. The UAV route generation system according to claim 1, characterized in that: The multi-source sensor fusion module includes a sensor calibration and spatiotemporal synchronization module and a feature extraction and fusion module.

4. The UAV route generation system according to claim 1, characterized in that: The intelligent mission planning module includes the following route generation algorithms: Terrain-mimicking flight mode: generating a route parallel to the terrain based on DEM data, and dynamically adjusting the flight altitude according to the terrain fluctuations; Circular route mode: For cylindrical structures such as towers and bridges, a spiral ascending or descending circular route is generated to ensure 360° coverage without blind spots; Figure-8 route mode: Suitable for large-area vegetation monitoring, the figure-8 trajectory reduces duplicate coverage and improves operational efficiency.

5. The UAV route generation system according to claim 1, characterized in that: The dynamic obstacle avoidance and safety module includes an obstacle detection module.

6. The UAV route generation system according to claim 1, characterized in that: The human-computer interaction and data management module includes a human-computer interactive operation interface.

7. The UAV route generation system according to claim 1, characterized in that: The processor model is NVIDIA Jetson AGX Orin.

Images