Air-ground cooperative control and obstacle avoidance method and system

Abstract

The application belongs to the field of air-ground cooperative control and obstacle avoidance, and discloses an air-ground cooperative control and obstacle avoidance algorithm, according to different environment data, the obstacle avoidance algorithm is divided into two-dimensional and three-dimensional algorithms, the algorithm is suitable for an environment lacking height data, and includes a static ellipse tangent method (SETG-TG) and a dynamic ellipse tangent method (DETG-TG), which are respectively suitable for static and dynamic environments; the three-dimensional obstacle avoidance algorithm adopts a three-dimensional tangent method (3D STG) for path planning and obstacle avoidance in a three-dimensional space. The application aims to ensure that a UAV and a ground vehicle can effectively avoid obstacles in real time in a complex and dynamic environment, and guarantee the safety, stability and path planning accuracy of the UAV and the ground vehicle.

Description

Technical Field

[0001] This invention belongs to, but is not limited to, the field of air-ground cooperative control and obstacle avoidance technology, and particularly relates to an air-ground cooperative control and obstacle avoidance method and system. Background Technology

[0002] Natural resource surveys and monitoring in complex and challenging areas have always been a difficult problem for the industry. These areas often feature alternating peaks and valleys, with complex and varied terrain, making some areas inaccessible and lacking ground communication facilities. This significantly impacts the conduct of natural resource monitoring work in these areas and severely limits the efficiency of monitoring efforts.

[0003] The language model-driven vehicle-to-machine (V2M) cooperative planning system aims to achieve efficient collaboration between vehicles and drones through two phases: task planning and dynamic collaboration. In the task planning phase, the system utilizes a large-model-driven intelligent engine to analyze and decompose tasks, generating specific sub-tasks and allocating and scheduling them. In the dynamic collaboration phase, the system employs various algorithms and technologies to ensure real-time adjustments and optimizations during task execution, addressing various dynamic changes and ultimately achieving successful task execution.

[0004] During the task planning phase, the system first conducts in-depth analysis based on key information such as the task nature, objectives, and scope, using a large-model-driven intelligent engine. Through the instruction fusion and conflict resolution module, the system can integrate instructions from multiple sources and effectively eliminate logical contradictions, ensuring the consistency and feasibility of task instructions. This process lays a solid foundation for subsequent task allocation.

[0005] Subsequently, the task allocation and scheduling module comprehensively considers constraints such as the vehicle and UAV's endurance, payload limitations, and sensor types to scientifically and rationally allocate sub-tasks. These sub-tasks are then passed to the UAV planning module and the vehicle planning module, respectively. The UAV planning module is responsible for developing detailed speed plans, payload usage schemes, flight mission arrangements, and route plans to ensure that the UAV can complete its mission efficiently and safely. Meanwhile, the vehicle planning module focuses on the specific arrangements of vehicle tasks, including path selection and time planning, to ensure that ground vehicles can work collaboratively with the UAV in the air to achieve the mission objectives. Through this series of planning and scheduling, the system completes the initial static allocation of tasks, forming a comprehensive and detailed mission execution plan.

[0006] Once in the dynamic collaboration phase, the system employs a series of advanced technologies to ensure real-time adjustments and optimizations during mission execution. In the aerial platform portion, the system uses formation flight as its core architecture, constructing a basic collaborative framework for the UAV swarm through pre-defined synchronization and path algorithms. Based on this architecture, the air-to-air collaboration module utilizes a distributed communication protocol to achieve real-time status sharing among multiple UAVs, including key information such as location, battery level, and mission progress. Through load balancing strategies, the system dynamically allocates computing resources and mission load, ensuring that each UAV can efficiently complete its assigned tasks.

[0007] On the ground platform, the system utilizes the vehicle's built-in computing power as its core resource pool, providing robust distributed computing support for the aerial cluster through an edge computing architecture. The high-performance computing units onboard the vehicles form a hybrid computing network with the cloud, using a spatiotemporal synchronization protocol to integrate real-time status data of the UAVs in the airspace (such as flight trajectories and mission progress) and ground environmental data (such as road conditions and obstacle information). This real-time data integration not only enhances the system's perception capabilities but also provides a solid foundation for dynamic scheduling. Through the air-ground linkage mechanism, the system can flexibly respond to various environmental changes, enabling real-time adjustments to takeoff and landing points, vehicle routes, etc., ensuring the continuity and efficiency of mission execution.

[0008] In the air-to-ground coordination component, the system effectively responds to environmental changes through a dynamic scheduling mechanism, combined with arc route calculation using RVND-SA and neighborhood search, as well as obstacle avoidance and correction technologies based on 2D and 3D obstacle avoidance. These technologies ensure the safe operation of UAVs and vehicles in complex environments. The task reallocation module fully utilizes edge computing capabilities to dynamically allocate resources based on real-time status (such as vehicle location and task point changes), further enhancing the system's robustness and flexibility. Ultimately, seamless integration across the entire process is achieved, ensuring the smooth execution and high-quality completion of tasks.

[0009] The existing technology is "3-D Low-Altitude Urban Environment UAV Path Planning Algorithm Based on Elliptical Tangent Graph (APPATT)" published by Liu Huan et al. in IEEET-ITS. This method uses the circumscribed ellipse of the obstacle as the geometric primitive, generates a local elliptical tangent graph online, and constructs three candidate sub-paths (left, right, and top) each time an obstacle is encountered. It then selects one path to continue forward using heuristic rules. To improve executability, the authors further smooth the discrete trajectory using B-splines.

[0010] However, APPATT still faces three key technical bottlenecks: ① The complete tangent network branches exponentially in dense urban canyon scenes, making real-time performance sensitive to CPU and memory usage. Literature also points out that traditional tangent-graph methods are "inefficient in generating the global road network, making it difficult to meet real-time planning requirements"; ② The algorithm essentially falls under the category of local navigation. Relying solely on shortening the step size for dynamic obstacles still requires assuming measurable obstacle speeds, easily leading to multiple replanning attempts when encountering suddenly inserted or speed-altering targets; ③ Similar to typical bug / Tangent-Bug families, APPATT's local decision rules are prone to getting lost in "local minima" within concave fences and U-shaped passages, requiring additional strategies (such as virtual sub-targets or dual windows) to escape. These issues are precisely the focus of this application's 3D-STG further addressing. Summary of the Invention

[0011] To address the problems existing in the prior art, this invention provides an air-ground cooperative control and obstacle avoidance algorithm.

[0012] This invention is implemented as follows: an air-ground cooperative control and obstacle avoidance algorithm, characterized in that, depending on the environmental data, the obstacle avoidance algorithm is divided into two-dimensional and three-dimensional algorithms. The two-dimensional algorithms are suitable for environments lacking elevation data and include the static elliptical tangent method (SETG-TG) and the dynamic elliptical tangent method (DETG-TG), which are applicable to static and dynamic environments, respectively. The three-dimensional obstacle avoidance algorithm uses the three-dimensional tangent method (3D STG) for path planning and obstacle avoidance in three-dimensional space.

[0013] Furthermore, the static elliptical tangent method determines the safe path of the UAV by calculating the tangent between the static boundary of the obstacle and the current position of the UAV, specifically including:

[0014] (1) Environmental modeling and obstacle recognition: The system acquires two-dimensional data of the surrounding environment and identifies all static obstacles (such as buildings, fixed obstacles, etc.).

[0015] (2) Geometric modeling of obstacles: Each obstacle is represented by a geometric shape (such as an ellipse or rectangle), and the system calculates the boundary of the obstacle;

[0016] (3) Tangent calculation: Through geometric calculation, the system calculates multiple possible tangents between the UAV's current position and each obstacle. A tangent is a line originating from the current position that intersects the boundary of an obstacle;

[0017] (4) Path planning: The system selects the safest and shortest tangent path and generates an obstacle avoidance path to ensure that the UAV will not collide with obstacles;

[0018] (5) Path execution: Based on the calculated tangent path, the UAV begins to perform obstacle avoidance maneuvers and flies along the tangent.

[0019] Furthermore, the dynamic elliptical tangent method ensures that the UAV avoids moving obstacles in a constantly changing environment by monitoring the dynamic position of obstacles in real time and updating the UAV's path planning. Specifically, it includes:

[0020] (1) Environmental modeling and dynamic obstacle recognition: The system acquires real-time two-dimensional data of the surrounding environment and identifies all dynamic obstacles (such as moving vehicles, pedestrians, etc.).

[0021] (2) Obstacle position update: As time goes by, the position of the obstacle changes continuously, and the system updates the geometric position of the obstacle in real time;

[0022] (3) Dynamic tangent calculation: Based on the current position of the obstacle and the position of the UAV, the system recalculates the tangent path to ensure that the UAV can avoid moving obstacles;

[0023] (4) Path planning and adjustment: The system plans the safest and shortest path for the UAV and adjusts the path in real time each time the obstacle position is updated to avoid collision;

[0024] (5) Path execution and dynamic adjustment: The UAV flies along the calculated path and adjusts its course according to the real-time updated obstacle information to ensure safe obstacle avoidance in a dynamic environment.

[0025] Furthermore, the core principle of the three-dimensional tangent method is to determine the safe path of the drone by calculating the "tangent" of the obstacle, specifically including:

[0026] (1) Environmental perception and data collection: The system first collects three-dimensional environmental data around the UAV through sensors (such as lidar, cameras, etc.). This data includes the three-dimensional position, shape and size of obstacles, as well as the current position of the UAV. This information is used to generate an accurate three-dimensional environmental model.

[0027] (2) Obstacle modeling and data preprocessing: Using the collected 3D data, the system establishes a complete 3D environment model to represent the shape of obstacles (such as buildings, hills, etc.). Each obstacle is represented in detail as a 3D geometry in this model. The position of the UAV and the surrounding environment are also taken into consideration.

[0028] (3) Tangent calculation: Based on the current position of the UAV, the algorithm calculates multiple possible "tangents". These tangents refer to the paths that start from the UAV and are tangent to the surface of the obstacle. When the UAV flies along these tangents, it will not collide with the obstacle, thus ensuring flight safety.

[0029] (4) Path planning and selection: Based on the calculated tangent, generate the safest, shortest and most suitable path;

[0030] (5) Dynamic update and adjustment: Real-time update of the position and shape of obstacles, recalculation of tangents, and adjustment of flight path.

[0031] Based on the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solution to be protected by this invention are as follows:

[0032] This invention aims to ensure that drones and ground vehicles can avoid obstacles in real time and effectively in complex and dynamic environments, thereby ensuring the safety, stability, and accuracy of path planning for drones and ground vehicles.

[0033] First, this invention achieves a closed-loop link from centimeter-level 3D perception to sub-second-level path replanning through a four-level collaborative architecture of "environmental perception - obstacle modeling - dimensional adaptive tangent planning - flight execution," overcoming the gap between the two ends of existing technologies that rely on cumbersome two-dimensional static assumptions or global sampling. Based on obstacle geometry reconstruction coupled with octagonal voxels and α-shapes, this solution can dynamically adjust the safety margin without increasing redundancy expansion, enabling the UAV to maintain a stable safety clearance in various scenarios such as building canyons and hilly terrain.

[0034] Secondly, the path planning module divides the system into three sub-modules: two-dimensional static, two-dimensional dynamic, and three-dimensional tangent. These sub-modules automatically switch based on obstacle movement characteristics and spatial dimensions, significantly reducing the probability of getting lost in local minima and frequent backtracking. Combined with Dubins-Air-3D curvature continuity constraints and A*-ε cost evaluation, the output trajectory achieves a balance between energy consumption, communication link quality, and range redundancy. Experiments show that the average path excess rate is controlled at around 4%, while the safety gap is increased by more than 35%.

[0035] Finally, the flight execution layer employs an LQR-Backstepping-sliding mode composite controller and operates in a closed-loop manner at 400Hz on the SoC-FPGA, ensuring rapid convergence of attitude-velocity commands and a lateral tracking error of <0.15m. Combined with millimeter-wave radar blind spot compensation and multi-sensor fusion positioning, it can maintain a replanning latency of 50ms under dynamic obstacle insertion or sudden turbulence conditions. These advantages collectively endow this invention with high reliability, high real-time performance, and high energy efficiency in industrial scenarios such as urban low-altitude logistics, disaster emergency monitoring, and complex mine inspection, significantly outperforming existing obstacle avoidance systems of the same type. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the static elliptical tangent method provided in an embodiment of the present invention;

[0037] Figure 2 This is a schematic diagram of the dynamic elliptical tangent method provided in an embodiment of the present invention;

[0038] Figure 3 This is a schematic diagram of the three-dimensional tangent method provided in an embodiment of the present invention;

[0039] Figure 4 This is a schematic diagram of the three-dimensional tangent path decision method provided in an embodiment of the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0041] In urban airspace and complex low-altitude operational scenarios, traditional UAV obstacle avoidance solutions mostly rely on single-dimensional path planning—either handling static obstacles in a two-dimensional projection or employing costly global sampling methods in three-dimensional space. The lack of dimensional adaptation and dynamic behavior characterization means that existing systems often experience path backtracking, frequent replanning, and even emergency stops due to failure to capture new obstacles in confined environments like valleys or hilly terrain. This directly impacts the timeliness and safety of industrial-grade tasks (such as emergency delivery, detailed mapping, and inspection).

[0042] To overcome this bottleneck, this system first utilizes multi-source heterogeneous sensor fusion—millimeter-wave radar for blind spot filling, solid-state lidar for providing high-precision depth point clouds, and enhanced visual-inertial navigation links for positional constraints in weakly textured areas—to generate a dense 3D occupancy map in real time. The weighted and sparsified point cloud is mapped onto an octet tree and coupled with the high-frequency attitude data from the inertial station, thereby maintaining centimeter-level positioning and obstacle boundary noise of less than 2cm without sacrificing resolution.

[0043] The obstacle modeling module performs Hausdorff distance-based clustering on the dense voxel map, followed by α-shape surface reconstruction and Minkowski dilation for each connected volume, giving the boundaries redundant aircraft shape. Simultaneously, it constructs the minimum circumscribed ellipsoid parameters for the connected volumes, providing analytical support for subsequent elliptical tangent solving. Thanks to the adaptive scaling of clustering granularity with voxel density, this step has an average latency of only 8ms within a 32m×32m×16m local window.

[0044] The core of path planning automatically triages traffic based on "environmental data dimension + obstacle dynamics": if the perception result is a static planar scene, the two-dimensional static elliptical tangent submodule is scheduled; when the obstacle velocity threshold is detected to exceed 0.1ms... -1The system automatically migrates to the 2D dynamic elliptical tangent submodule, introducing obstacle velocity vectors for time reachability analysis. If obstacles are significantly distributed in the vertical dimension, it switches to the 3D tangent planning submodule. All three submodules uniformly calculate closed candidate paths based on analytical tangents, and then use the improved A*-ε to perform secondary screening in the Dubins-Air-3D curvature continuous grid, comprehensively considering energy consumption, communication link quality, and range redundancy to give the optimal sequence.

[0045] The flight execution layer maps the selected trajectory to an attitude-velocity dual target in the airframe coordinate system and iteratively outputs the desired pitch, roll, and pitch-airspeed commands through an LQR-Backstepping composite controller. This control strategy operates on a 400Hz closed-loop system on the FPGA-SoC, and combines non-integer sliding mode to suppress thrust interference, ensuring a path tracking lateral error of <0.15m and no overshoot during pitch transitions.

[0046] The system has been applied to 25kg-class VTOL logistics drones with a speed of 40km / h. -1 The unmanned vehicle platform successfully replayed industrial scenarios: in urban canyons filled with high-rise buildings and hilly power line inspection corridors, the average safety clearance was improved by 35%, the path overrun rate was controlled at 4.1%, and the total CPU usage was less than 38%. The results demonstrate that, thanks to the coupling of dimension-adaptive hierarchical tangent planning and low-latency control, this air-ground cooperative obstacle avoidance system can provide real-time and reliable trajectory assurance for industrial-grade tasks such as urban low-altitude delivery, disaster monitoring, and mine inspection, solving the core pain point of traditional solutions in balancing three-dimensional dynamic obstacles and real-time performance.

[0047] This invention provides an air-ground cooperative control and obstacle avoidance algorithm. The algorithm aims to ensure that UAVs and ground vehicles can avoid obstacles in real-time and effectively in complex and dynamic environments, guaranteeing their safety, stability, and path planning accuracy. Depending on the environmental data, the obstacle avoidance algorithm is divided into two-dimensional and three-dimensional algorithms. Two-dimensional algorithms are suitable for environments lacking elevation data, while three-dimensional algorithms utilize three-dimensional spatial information to achieve more precise path planning and obstacle avoidance.

[0048] The two-dimensional obstacle avoidance algorithm includes the static elliptic tangent method (SETG-TG) and the dynamic elliptic tangent method (DETG-TG), specifically including:

[0049] (1) Static Ellipse Tangent Method (SETG-TG)

[0050] The static ellipse tangent method is suitable for static environments, where the position of the obstacle remains fixed. For example... Figure 1As shown, this method determines a safe path for the UAV by calculating the tangent between the static boundary of the obstacle and the UAV's current position. The tangent is the path along which the UAV starts from its current position and is tangent to the obstacle boundary; following this path, the UAV can avoid the obstacle and prevent collisions.

[0051] 1) Environmental modeling and obstacle recognition: The system acquires two-dimensional data of the surrounding environment and identifies all static obstacles (such as buildings, fixed obstacles, etc.).

[0052] 2) Geometric modeling of obstacles: Each obstacle is represented by a geometric shape (such as an ellipse or rectangle), and the system calculates the boundary of the obstacle.

[0053] 3) Tangent Calculation: Through geometric calculations, the system calculates multiple possible tangents between the UAV's current position and each obstacle. A tangent is a line originating from the current position that intersects the boundary of an obstacle.

[0054] 4) Path planning: The system selects the safest and shortest tangent path and generates an obstacle avoidance path to ensure that the drone will not collide with obstacles.

[0055] 5) Path execution: Based on the calculated tangent path, the UAV begins to perform obstacle avoidance maneuvers and flies along the tangent.

[0056] (2) Dynamic Ellipse Tangent Method (DETG-TG)

[0057] The dynamic elliptical tangent method is an extension of the static elliptical tangent method, applicable to dynamic environments where the position of obstacles changes over time. For example... Figure 2 As shown, this method ensures that the UAV avoids moving obstacles in a constantly changing environment by monitoring the dynamic position of obstacles in real time and updating the UAV's path planning.

[0058] 1) Environmental modeling and dynamic obstacle recognition: The system acquires real-time two-dimensional data of the surrounding environment and identifies all dynamic obstacles (such as moving vehicles, pedestrians, etc.).

[0059] 2) Obstacle position update: As time goes by, the position of obstacles changes continuously, and the system updates the geometric position of obstacles in real time.

[0060] 3) Dynamic tangent calculation: Based on the current position of the obstacle and the position of the drone, the system recalculates the tangent path to ensure that the drone can avoid moving obstacles.

[0061] 4) Path planning and adjustment: The system plans the safest and shortest path for the drone and adjusts the path in real time each time the obstacle position is updated to avoid collisions.

[0062] 5) Path execution and dynamic adjustment: The UAV flies along the calculated path and adjusts its course according to the real-time updated obstacle information to ensure safe obstacle avoidance in a dynamic environment.

[0063] The described 3D obstacle avoidance algorithm employs the 3D tangent method (3D STG), an algorithm used for path planning and obstacle avoidance in 3D space, particularly suitable for obstacle avoidance by drones or ground vehicles in complex terrain. Figure 3 As shown, its core principle is to determine the safe path for the drone by calculating the "tangent" of the obstacle. This method considers not only the planar position of the obstacle but also its height difference, ensuring that the drone can avoid obstacles in three-dimensional space such as tall buildings and hills.

[0064] (1) Environmental perception and data collection: The system first collects three-dimensional environmental data around the UAV through sensors (such as lidar, cameras, etc.). This data includes the three-dimensional position, shape and size of obstacles, as well as the current position of the UAV. This information is used to generate an accurate three-dimensional environment model.

[0065] (2) Obstacle modeling and data preprocessing: Using the collected 3D data, the system establishes a complete 3D environment model to represent the shape of obstacles (such as buildings, hills, etc.). Each obstacle is represented in detail as a 3D geometry in this model, and the position of the UAV and its surrounding environment are also taken into consideration.

[0066] (3) Tangent calculation: Based on the current position of the UAV, the algorithm calculates multiple possible "tangents". These tangents refer to the paths that start from the UAV and are tangent to the surface of the obstacle. When the UAV flies along these tangents, it will not collide with the obstacle, ensuring flight safety.

[0067] (4) Path planning and selection: Based on the calculated tangent, the system will generate the safest, shortest and most suitable path. Figure 4 This path takes into account factors such as the drone's flight speed, altitude, and heading changes. The system will select a path that avoids obstacles and meets flight requirements.

[0068] (5) Dynamic updates and adjustments: As the drone continues to move forward, the environment may change, and new obstacles may appear. The system will update the position and shape of obstacles in real time, recalculate the tangent, and adjust the flight path. This enables the drone to cope with dynamically changing environments and maintain safe flight.

[0069] Through this series of steps, the 3D tangent method can plan accurate and safe paths for drones or ground vehicles, avoiding collisions with obstacles such as tall buildings and mountains, and ensuring safe and efficient navigation in complex 3D environments.

[0070] To address the bottlenecks of traditional 2D obstacle avoidance algorithms, which are prone to "height blind spots" and "local minima traps" in urban scenes and undulating terrain, this study introduces the 3D Tangent-Guided Planning (3D-STG) method. This method uses dense point clouds generated by heterogeneous fusion of multiple sensors as a data foundation. By constructing a directed voxel graph with topological constraints in real time, it achieves accurate reconstruction of spatial convex hulls such as buildings, vegetation clusters, and terrain steps, thus providing complete geometric boundaries for subsequent tangent construction.

[0071] After the dense point cloud is mapped to an octree voxel mesh, the algorithm employs an adaptive clustering strategy based on Hausdorff distance to segment connected obstacles and transforms each obstacle cell into a polyhedral model supporting Minkowski morphological expansion. This approach ensures that the safety margin dynamically adjusts with the aircraft size while avoiding unnecessary erosion of adjacent passable voxels by traditional expansion occupancy, thereby increasing the effective width of the passageway.

[0072] The tangent generation stage employs a combination of surface normal consistency constraints and extreme value envelope search. First, the tangent plane connecting the surface of each obstacle polyhedron to the aircraft's centroid is calculated. Then, within the tangent plane, the candidate set of tangents with the minimum rotation angle is obtained using Clarke's generalized gradient. The tangent vectors are coupled and mapped with the real-time attitude solver (based on a 3-DoF RTK-INS fusion unit) and immediately projected into a cluster of executable tracks in the aircraft's reference coordinate system.

[0073] To avoid oscillations in the path within the alternating three-dimensional spiral intervals, the algorithm embeds curvature continuity constraints based on the Dubins-Air-3D model into each feasible tangent and evaluates the cost on a three-dimensional grid of velocity, heading, and pitch using an improved A*-ε search. The comprehensive cost function simultaneously considers energy consumption, range redundancy, and communication link quality, ensuring that the output path not only meets the shortest path requirement but also balances energy efficiency and link robustness.

[0074] As the aircraft advances, the onboard FPGA-accelerated point cloud filtering module reconstructs the local environment at a refresh rate of 20Hz. If a new obstacle boundary is detected to cross within five times the aircraft's scale, the scheduler immediately triggers incremental tangent regeneration. Because the tangent search is limited to the local update region, the replanning delay is controlled within 50ms, fully meeting the sub-second response requirements. Even in the event of strong wind disturbances or temporary changes in takeoff and landing points, the flight safety envelope can be maintained.

[0075] This 3D-STG solution has been applied to 25kg-class vertical takeoff and landing fixed-wing aircraft and 40km / h aircraft. -1Scene playback tests were completed on the unmanned ground vehicle platform. The results showed that in complex building and canyon environments, the average safety clearance was increased by 37%, the path length excess rate was controlled within 4.2%, and the CPU usage throughout the process was less than 35%. This verifies the feasibility and efficiency of the 3D tangent method in multi-scenario industrial applications, providing a reliable obstacle avoidance solution for urban low-altitude logistics and automated inspection in complex mining areas.

[0076] It should be noted that embodiments of the present invention can be implemented in hardware, software, or a combination of both. The hardware portion can be implemented using dedicated logic; the software portion can be stored in memory and executed by a suitable instruction execution system, such as a microprocessor or dedicated-design hardware. Those skilled in the art will understand that the above-described devices and methods can be implemented using computer-executable instructions and / or included in processor control code, for example, such code provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The devices and modules of the present invention can be implemented by hardware circuitry such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field-programmable gate arrays, programmable logic devices, etc., or by software executed by various types of processors, or by a combination of the above-described hardware circuitry and software, such as firmware.

[0077] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions, and improvements made by those skilled in the art within the scope of the technology disclosed in the present invention, and within the spirit and principles of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An air-ground cooperative control and obstacle avoidance system, characterized in that, It includes an environmental perception module, an obstacle modeling module, a path planning module, and a flight execution module, among which... The environmental perception module is used to acquire two-dimensional or three-dimensional environmental data; The obstacle modeling module constructs the geometric boundaries of obstacles based on environmental data; The path planning module calls the two-dimensional static ellipse tangent planning submodule, the two-dimensional dynamic ellipse tangent planning submodule, or the three-dimensional tangent planning submodule respectively to generate obstacle avoidance paths based on the environmental data dimensions. The flight execution module receives the obstacle avoidance path and outputs attitude and speed commands to control the UAV to complete obstacle avoidance flight.

2. The system according to claim 1, characterized in that, The two-dimensional static ellipse tangent planning submodule performs the following steps: Step 1: Acquire two-dimensional static environmental data and calibrate the geometric outline of the obstacle; Step 2: Calculate the left and right tangents based on the current position of the UAV and the geometric outline of the obstacle; Step 3: Based on the preset cost function, select the tangent with the smaller cost from the left and right tangents as the obstacle avoidance path.

3. The system according to claim 1, characterized in that, The two-dimensional dynamic ellipse tangent planning submodule performs the following steps: Step 1: Periodically acquire the two-dimensional position and velocity of dynamic obstacles; Step 2: Update the geometric outline of the obstacle; Step 3: Calculate the left and right tangents based on the current position of the drone and the updated geometric outline of the obstacle; Step 4: When the position of the obstacle changes, repeat steps 2 and 3 and refresh the obstacle avoidance path in real time.

4. The system according to claim 1, characterized in that, The 3D tangent planning submodule performs the following steps: Step 1: Collect 3D environmental data and reconstruct 3D obstacle models; Step 2: Calculate the set of tangent directions based on the current 3D coordinates of the UAV and the 3D obstacle model; Step 3: Construct a cost function based on path length, safety distance, and attitude change, and select the path with the minimum cost from the set of tangent directions as the obstacle avoidance path.

5. A method for air-ground cooperative control and obstacle avoidance, characterized in that, Includes the following steps: Step 1: Obtain environmental data and determine whether it contains valid elevation data; Step 2: When no elevation data exists, call the 2D obstacle avoidance algorithm; when elevation data exists, call the 3D obstacle avoidance algorithm. Step 3: Output the obstacle avoidance path based on the invoked algorithm; Step four: Output attitude and speed commands and control the UAV to fly along the obstacle avoidance path.

6. The method according to claim 5, characterized in that, In step two, both the two-dimensional obstacle avoidance algorithm and the three-dimensional obstacle avoidance algorithm evaluate candidate paths based on the following cost function: The cost is equal to the path length weight multiplied by the path length plus the safety distance weight multiplied by the nearest obstacle distance plus the attitude change weight multiplied by the heading change angle. The path with the minimum replacement value is selected as the output path from the candidate paths.

7. The method according to claim 5, characterized in that, Step three further includes reacquiring environmental data and recalculating the obstacle avoidance path at each preset update cycle during the execution of the obstacle avoidance path, and replacing the currently executing path when the heading difference between the new path and the currently executing path exceeds a preset threshold.

8. A computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform the method of claim 5.

9. An environmental dimension discrimination and algorithm switching device, characterized in that, It includes a data analysis unit and an algorithm calling unit. The data analysis unit is used to determine whether there is elevation information based on environmental sensor data, and the algorithm calling unit calls a two-dimensional obstacle avoidance algorithm or a three-dimensional obstacle avoidance algorithm based on the judgment result of the data analysis unit.

10. A method for modeling obstacle boundaries, characterized in that, Includes the following steps: Step 1: Receive point cloud data output by the environmental perception module; Step 2: Perform density clustering on the point cloud data to separate each obstacle; Step 3: Perform minimum envelope ellipsoid fitting for each obstacle to generate the geometric boundary; Step 4: Write the geometric boundary parameters into the obstacle avoidance path planning module for use in tangent calculation.

Images