Autonomous flight control system for drones without external navigation

Abstract

The present invention discloses an autonomous flight control system for unmanned aerial vehicles without external navigation, and the present invention relates to the technical field of unmanned aerial vehicle control. The autonomous flight control system for unmanned aerial vehicles without external navigation, the integrated autonomous obstacle avoidance system can accurately sense and quickly respond to changes in the surrounding environment through multi-sensor fusion, identify and classify obstacles, automatically calculate and select the best obstacle avoidance path, and effectively ensure flight safety; in particular, the system can dynamically adjust the flight path and navigation strategy according to real-time environmental changes, and can maintain efficient and accurate navigation capabilities even in complex and changeable environments; this intelligent dynamic path planning capability not only improves the autonomy and flexibility of the unmanned aerial vehicle, but also significantly enhances its execution efficiency and success rate in complex mission scenarios.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle control, and in particular to an unmanned aerial vehicle autonomous flight control system without external navigation. Background Art

[0002] Autonomous UAV flight means that the UAV can take off, fly, perform tasks (such as filming, monitoring, inspection, etc.) and land autonomously without or with little human intervention. The realization of this technology relies on the support of advanced flight control systems, navigation technology, sensor technology, data processing and decision-making algorithms.

[0003] For example, the cloud control system for autonomous flight of unmanned aerial vehicles disclosed in publication number CN104932529B includes a ground control station, a cloud server, unmanned aerial vehicle equipment, and aerial photography equipment. The real-time autonomous flight of unmanned aerial vehicles is to establish a route planning control analysis module based on real-time information. The ground control station uses the flight coordinate range information to perform preliminary automatic route planning. During low-altitude flight, the aerial photography equipment is used to collect aerial data of terrain spatial information, which is transmitted to the cloud server through the ground control station using the 4G network. Cloud processing technology is used to perform three-dimensional real-time visualization spatial modeling of aerial data, and the initial route of the unmanned aerial vehicle mission flight is corrected in real time in combination with DEM and spatial topological relationships. The flight control signal is returned to the autopilot to realize autonomous flight of the unmanned aerial vehicle.

[0004] Existing autonomous flight of drones only flies autonomously along routes planned in advance. However, in actual use, the drone is easily deviated by unexpected strong winds or obstacles. Simply relying on the positioning system to return to the course is not accurate enough and is still prone to deviation, which in turn affects the flight of normal routes. Moreover, if the drone flies completely according to the set route, it cannot avoid obstacles well. The current means mainly rely on radar and other means to conduct on-site detection and avoidance. It cannot make predictions and plans in advance, and the obstacle avoidance efficiency is low, and it may even be impossible to avoid the obstacle. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides an autonomous flight control system for unmanned aerial vehicles that does not require external navigation, thereby solving the problems existing in the existing autonomous flight control systems for unmanned aerial vehicles.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: an autonomous flight control system for a drone without external navigation, comprising:

[0007] The sensor unit integrates LiDAR, IMU, GPS, visual sensor and ultrasonic sensor;

[0008] Flight control system, used to output control signals to control the autonomous flight of the UAV;

[0009] The attitude and coordinate detection unit reads the IMU data in real time, calculates the attitude and spatial coordinates of the drone through algorithms, and feeds back to the flight control system for attitude adjustment and spatial positioning;

[0010] The displacement detection module obtains the speed and displacement of the drone through integral calculation based on the acceleration data of the IMU, and corrects the integral result in combination with the visual odometer or GPS data. The displacement detection result is compared with the expected target, and closed-loop control is performed through the flight control system to make the drone fly along the predetermined trajectory;

[0011] Autonomous obstacle avoidance system, which is used to scan the surrounding environment to obtain obstacle distance and location information. The obstacle avoidance algorithm processes the collected data to identify and classify obstacles, and calculates the best obstacle avoidance path among multiple obstacle avoidance paths. At the same time, it generates control instructions to adjust the drone's attitude and speed to avoid obstacles;

[0012] The internal navigation model builds a 3D map of the surrounding environment and generates a flight path from the starting point to the end point. It also dynamically adjusts the flight path and navigation strategy based on real-time environmental changes and obstacle avoidance requirements.

[0013] The multi-dimensional fusion optimization module uses a data fusion algorithm to fuse data from different sensors and optimizes the obstacle avoidance algorithm and path planning algorithm.

[0014] Preferably, the sensor unit further comprises:

[0015] Processor unit for real-time data processing and algorithm execution;

[0016] Power management module, providing stable power supply to support flight and sensor operation;

[0017] Storage device for recording flight data and algorithm models.

[0018] Preferably, the posture and coordinate detection unit includes:

[0019] A1. IMU data reading: real-time reading of IMU data, including: acceleration data a=(a X , a Y , a Z ), angular velocity data ω=(ω X ,ω Y ,ω Z ), magnetic field strength data m=(m X , m Y , m Z ), X, Y, and Z represent the values ​​obtained based on the XYZ three-axis coordinates;

[0020] A2. Attitude solution: Use Kalman filter or extended Kalman filter algorithm, combined with IMU and GPS data F GPS =(X GPS , Y GPS , Z GPS ), calculate the estimated posture q=(q0, q1, q2, q3) and spatial coordinates p=(X, Y, Z) of the drone;

[0021] A3. Data feedback: Feedback the estimated attitude q and spatial coordinates p to the flight control system for attitude adjustment and spatial positioning.

[0022] Preferably, the operation steps of the displacement detection module include:

[0023] B1. Integral calculation: Use numerical integration to calculate velocity and displacement p from acceleration data a t ;

[0024] ;

[0025] p t =p t-1 + V t Δt;

[0026] Among them, V t and V t−1 denote the velocity vector at time t and t−1 respectively, a t and a t−1 denote the acceleration vectors at time t and t−1 respectively, Δt is the time interval, and p t and p t−1 Respectively represent the position vector at time t and t−1;

[0027] B2. Error correction: Combine visual odometer or GPS data for error correction and adjust p t ;

[0028] B3. Closed-loop control: Let the predetermined trajectory be p target , the corrected displacement is compared with the predetermined trajectory p target In contrast, the flight state is adjusted by PID controller.

[0029] Preferably, the steps of controlling and adjusting the flight state by the PID controller are:

[0030] B3.1. Calculation error: The error e is the current position p t With the target position p target The vector difference between them is:

[0031] e=ptarget − p t ;

[0032] B3.2, PID controller: Use the PID control algorithm to calculate the control signal u. The output of the PID controller is the weighted sum of the proportional P, integral I and differential D:

[0033] ;

[0034] Among them, Kp, Ki and Kd are the gains corresponding to proportional P, integral I and differential D respectively;

[0035] B3.3. Adjust the flight status: According to the output u of the PID controller, adjust the control input of the aircraft to reduce the error and move the aircraft to the target position.

[0036] Preferably, the operating steps of the autonomous obstacle avoidance system include:

[0037] C1. Data collection: Use LiDAR to scan the environment and obtain point cloud data , where N is the number of points in the point cloud, and r is any integer from 1 to N;

[0038] The visual sensor captures the image T and obtains the obstacle information O through the image processing algorithm;

[0039] C2. Data processing: Input PL and O into the obstacle avoidance algorithm, use the deep learning model or rule-based classifier C, output the obstacle type T and position (X0, Y0, Z0), and the output expression is:

[0040] T, (X0, Y0, Z0)=C(PL, O);

[0041] C3, Decision generation: Based on the obstacle information, use the A* algorithm or RRT algorithm to calculate the obstacle avoidance path ; Where M is the number of points in the path, j is any integer from 1 to M;

[0042] C4, Execution Control: P a Converted into flight control instructions U=(uX, uY, uZ, uθ, uϕ, uψ), where (uX, uY, uZ) corresponds to the linear velocity control instructions on the XYZ three axes, and (uθ, uϕ, uψ) corresponds to the angular velocity control instructions on the XYZ three axes, and sent to the flight control system.

[0043] Preferably, the operation steps of the internal navigation model include:

[0044] D1. Environmental map construction: Use SLAM technology to combine PL and T to build a three-dimensional map S;

[0045] D2, path planning: Use A* or RRT algorithm to plan the path L on the three-dimensional map S;

[0046] D3, Dynamic Adjustment: Re-evaluate and adjust L based on new obstacle information and environmental changes.

[0047] Preferably, the operation steps of the multivariate fusion optimization module include:

[0048] D1. Multi-sensor data fusion:

[0049] Use Bayesian filtering or particle filtering algorithms to fuse LiDAR, visual sensors, IMU and GPS multi-source data to obtain more accurate environmental perception and drone state estimation formulas:

[0050] Belief Update=Fuse(PL,O,a,ω,m,F GPS );

[0051] Among them, Fuse represents the fusion function, which integrates the information of each sensor through probability theory or statistical methods according to the characteristics and uncertainty of sensor data, and outputs a state estimation or environmental model;

[0052] D2. System optimization:

[0053] Optimize parameters of obstacle avoidance algorithm and path planning algorithm.

[0054] Preferably, the obstacle avoidance algorithm optimization includes:

[0055] Parameter tuning: According to the actual flight environment and mission requirements of the UAV, adjust the key parameters in the obstacle avoidance algorithm, including obstacle detection threshold, optimization target of obstacle avoidance path planning; optimize the target packet time and minimize energy consumption;

[0056] Algorithm selection: For complex dynamic environments, use a real-time obstacle avoidance algorithm based on deep learning; for static or simple environments, use a geometric or grid search algorithm;

[0057] Fusion of multiple algorithms: Combine multiple obstacle avoidance algorithms and switch between different obstacle avoidance algorithms for different detection environments.

[0058] Preferably, the path planning algorithm optimization includes:

[0059] Improved algorithm efficiency: Optimized the calculation process of the path planning algorithm and used priority queues to store and update candidate paths;

[0060] Multi-objective optimization: multiple optimization objectives are considered during path planning, including the shortest path, the lowest energy consumption, and the highest safety. Weights are set for multiple optimization objectives, and the path with the smallest weighted value is taken as the best path.

[0061] Real-time improvement: For dynamic environments, incremental path planning algorithms or rolling horizon planning algorithms are used to quickly respond to environmental changes.

[0062] The present invention provides an autonomous flight control system for unmanned aerial vehicles that does not require external navigation. Compared with the prior art, it has the following beneficial effects:

[0063] 1. The autonomous flight control system of the UAV does not require external navigation. The integrated autonomous obstacle avoidance system can accurately sense and quickly respond to changes in the surrounding environment through multi-sensor fusion, identify and classify obstacles, automatically calculate and select the best obstacle avoidance path, and effectively ensure flight safety. In particular, the system can dynamically adjust the flight path and navigation strategy according to real-time environmental changes, and maintain efficient and accurate navigation capabilities even in complex and changing environments. This intelligent dynamic path planning capability not only improves the autonomy and flexibility of the UAV, but also significantly enhances its execution efficiency and success rate in complex mission scenarios.

[0064] 2. The attitude and coordinate detection unit of the autonomous flight control system of the UAV does not require external navigation and realizes high-precision and high-stability detection of the attitude and spatial coordinates of the UAV. By reading the IMU multi-source data in real time and combining it with GPS information, the attitude solution is performed using the Kalman filter or extended Kalman filter algorithm, which effectively reduces noise interference and improves the accuracy and reliability of attitude estimation and spatial positioning. This technical effect not only enhances the autonomous flight capability of the UAV, but also ensures its stable flight in complex environments, providing solid technical support for the UAV to perform high-precision tasks.

[0065] 3. This autonomous flight control system for UAVs that does not require external navigation calculates speed and displacement through numerical integration, and combines visual odometer or GPS for error correction, which significantly improves the accuracy of displacement detection. At the same time, it uses PID control algorithm for closed-loop control, calculates errors in real time and dynamically adjusts the flight state, so that the UAV can accurately track the predetermined trajectory, enhancing the stability and reliability of the flight, which is of great significance for improving the autonomous flight capability and mission execution accuracy of UAVs in complex environments.

[0066] 4. The autonomous flight control system of the UAV does not require external navigation, and the autonomous obstacle avoidance system realizes high precision and dynamic adaptability of autonomous obstacle avoidance and navigation of the UAV; by integrating LiDAR and visual sensors and combining deep learning models, it realizes accurate perception and obstacle recognition of complex environments; at the same time, by using advanced path planning algorithms and dynamic adjustment mechanisms, the UAV can flexibly plan and adjust the flight path in the constructed three-dimensional map, effectively respond to environmental changes and sudden obstacles, and significantly improve the intelligence level and safety of autonomous obstacle avoidance and navigation.

[0067] 5. This autonomous flight control system for drones does not require external navigation. It builds an accurate three-dimensional environmental map through SLAM technology. Combined with an efficient path planning algorithm, the drone can autonomously plan the optimal flight path. At the same time, the model has dynamic adjustment capabilities, can respond to environmental changes and obstacle information in real time, and flexibly adjust the flight path to ensure that the drone completes its mission safely and efficiently, significantly improving the autonomy and adaptability of the drone in complex scenarios.

[0068] 6. The autonomous flight control system for drones that does not require external navigation has significantly improved the accuracy of environmental perception and state estimation through multi-sensor data fusion; at the same time, system optimization ensures the best performance of obstacle avoidance and path planning algorithms in different environments, including parameter tuning, algorithm selection and fusion, efficiency improvement, multi-objective optimization and real-time improvement, enabling drones to respond to complex and changeable flight missions more intelligently and flexibly, and improving the autonomy and reliability of the overall system. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 It is a system principle block diagram of the present invention;

[0070] Figure 2 It is a schematic diagram of the overall process of the present invention;

[0071] Figure 3 Schematic diagram of the working process of the posture and coordinate detection unit of the present invention;

[0072] Figure 4 Schematic diagram of the working process of the displacement detection module of the present invention;

[0073] Figure 5 Schematic diagram of the working process of the autonomous obstacle avoidance system of the present invention. DETAILED DESCRIPTION

[0074] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.

[0075] See also Figure 1-Figure 5 , the present invention provides the following six technical solutions:

[0076] The first implementation method: an autonomous flight control system for a UAV without external navigation, comprising:

[0077] The sensor unit integrates LiDAR, IMU, GPS, visual sensor and ultrasonic sensor;

[0078] Flight control system, used to output control signals to control the autonomous flight of the UAV;

[0079] The attitude and coordinate detection unit reads the IMU data in real time, calculates the attitude and spatial coordinates of the drone through algorithms, and feeds back to the flight control system for attitude adjustment and spatial positioning;

[0080] The displacement detection module obtains the speed and displacement of the drone through integral calculation based on the acceleration data of the IMU, and corrects the integral result in combination with the visual odometer or GPS data. The displacement detection result is compared with the expected target, and closed-loop control is performed through the flight control system to make the drone fly along the predetermined trajectory;

[0081] Autonomous obstacle avoidance system, which is used to scan the surrounding environment to obtain obstacle distance and location information. The obstacle avoidance algorithm processes the collected data to identify and classify obstacles, and calculates the best obstacle avoidance path among multiple obstacle avoidance paths. At the same time, it generates control instructions to adjust the drone's attitude and speed to avoid obstacles;

[0082] The internal navigation model builds a 3D map of the surrounding environment and generates a flight path from the starting point to the end point. It also dynamically adjusts the flight path and navigation strategy based on real-time environmental changes and obstacle avoidance requirements.

[0083] The multi-dimensional fusion optimization module uses a data fusion algorithm to fuse data from different sensors and optimizes the obstacle avoidance algorithm and path planning algorithm.

[0084] The sensor unit further comprises:

[0085] Processor unit for real-time data processing and algorithm execution.

[0086] Power management module, providing stable power supply to support flight and sensor operation;

[0087] Storage device for recording flight data and algorithm models.

[0088] The integrated autonomous obstacle avoidance system can accurately sense and quickly respond to changes in the surrounding environment through multi-sensor fusion, identify and classify obstacles, automatically calculate and select the best obstacle avoidance path, and effectively ensure flight safety; what is particularly outstanding is that the system can dynamically adjust the flight path and navigation strategy according to real-time environmental changes, and maintain efficient and accurate navigation capabilities even in complex and changing environments; this intelligent dynamic path planning capability not only improves the autonomy and flexibility of the UAV, but also significantly enhances its execution efficiency and success rate in complex mission scenarios.

[0089] The second implementation mode is mainly different from the first implementation mode in that the posture and coordinate detection unit includes:

[0090] A1. IMU data reading: real-time reading of IMU data, including: acceleration data a=(a X , a Y , a Z ), angular velocity data ω=(ω X ,ω Y ,ω Z ), magnetic field strength data m=(m X , m Y , m Z ), X, Y, and Z represent the values ​​obtained based on the XYZ three-axis coordinates;

[0091] A2. Attitude solution: Use Kalman filter or extended Kalman filter algorithm, combined with IMU and GPS data F GPS =(X GPS , Y GPS , Z GPS ), calculate the estimated posture q=(q0, q1, q2, q3) and spatial coordinates p=(X, Y, Z) of the drone;

[0092] Kalman filtering is an algorithm that uses the linear system state equation to optimally estimate the system state through the system input and output observation data; the extended Kalman filter is an extended form of Kalman filtering in nonlinear systems. Its basic idea is to linearize the nonlinear system through Taylor series expansion, and then apply the Kalman filter framework to filter the signal. Specifically, in the prediction and update steps, EKF calculates the Jacobian matrix (i.e., the first-order derivative matrix) of the nonlinear function to approximate the linearization characteristics of the system.

[0093] A3. Data feedback: Feedback the estimated attitude q and spatial coordinates p to the flight control system for attitude adjustment and spatial positioning.

[0094] The attitude and coordinate detection unit realizes high-precision and high-stability detection of the UAV attitude and spatial coordinates; by reading the IMU multi-source data in real time and combining it with GPS information, the Kalman filter or extended Kalman filter algorithm is used for attitude solution, which effectively reduces noise interference and improves the accuracy and reliability of attitude estimation and spatial positioning; this technical effect not only enhances the autonomous flight capability of the UAV, but also ensures its stable flight in complex environments, providing solid technical support for the UAV to perform high-precision tasks.

[0095] The third implementation mode is mainly different from the first implementation mode in that the operation steps of the displacement detection module include:

[0096] B1. Integral calculation: Use numerical integration (such as trapezoidal integration, also known as trapezoidal rule or trapezoidal numerical integration, which is a numerical integration method. Its basic idea is to divide the integrand into several small trapezoids on the integration interval, and then calculate the sum of the areas of these small trapezoids to obtain the approximate integral value of the integrand) to calculate the velocity and displacement p from the acceleration data a t ;

[0097] ;

[0098] p t =p t-1 + V t Δt;

[0099] Among them, V t and V t−1 denote the velocity vector at time t and t−1 respectively, a t and a t−1 denote the acceleration vectors at time t and t−1 respectively, Δt is the time interval, and p t and p t−1 Respectively represent the position vector at time t and t−1;

[0100] B2. Error correction: Combine visual odometer or GPS data for error correction and adjust p t ;

[0101] B3. Closed-loop control: Let the predetermined trajectory be p target , the corrected displacement is compared with the predetermined trajectory p target In contrast, the flight state is adjusted by PID controller.

[0102] The steps to adjust the flight state through PID controller are:

[0103] B3.1. Calculation error: The error e is the current position p t With the target position p target The vector difference between them is:

[0104] e=p target −p t ;

[0105] B3.2, PID controller: Use the PID control algorithm to calculate the control signal u. The output of the PID controller is the weighted sum of the proportional P, integral I and differential D:

[0106] ;

[0107] Among them, Kp, Ki and Kd are the gains corresponding to proportional P, integral I and differential D respectively;

[0108] B3.3. Adjust the flight status: According to the output u of the PID controller, adjust the control input of the aircraft (such as thrust, rudder angle, etc.) to reduce the error and move the aircraft to the target position.

[0109] The speed and displacement are calculated through numerical integration, and error correction is performed in combination with visual odometer or GPS, which significantly improves the accuracy of displacement detection. At the same time, the PID control algorithm is used for closed-loop control, and the error is calculated in real time and the flight state is dynamically adjusted, so that the UAV can accurately track the predetermined trajectory, enhancing the stability and reliability of the flight. This is of great significance for improving the autonomous flight capability and mission execution accuracy of the UAV in complex environments.

[0110] The fourth implementation mode is mainly different from the first implementation mode in that the operation steps of the autonomous obstacle avoidance system include:

[0111] C1. Data collection: Use LiDAR to scan the environment and obtain point cloud data , where N is the number of points in the point cloud, and r is any integer from 1 to N;

[0112] The visual sensor captures the image T and obtains the obstacle information O through image processing algorithms (such as edge detection and feature extraction);

[0113] C2. Data processing: Input PL and O into the obstacle avoidance algorithm, use a deep learning model (such as CNN) or a rule-based classifier C, and output the obstacle type T and position (X0, Y0, Z0). The output expression is:

[0114] T, (X0, Y0, Z0)=C(PL, O);

[0115] C3, Decision generation: Based on the obstacle information, use the A* algorithm (A* algorithm is a heuristic search algorithm that finds the shortest path from the starting node to the target node in the graph) or the RRT algorithm (Rapidly-exploring Random Tree algorithm, which is a sampling-based global path planning algorithm. This algorithm searches for the path from the starting point to the end point in multi-dimensional space through random sampling and tree structure expansion. It is particularly suitable for high-dimensional and multi-degree-of-freedom systems) to calculate the obstacle avoidance path ; Where M is the number of points in the path, j is any integer from 1 to M;

[0116] C4, Execution Control: P aConverted into flight control instructions U=(uX, uY, uZ, uθ, uϕ, uψ), where (uX, uY, uZ) corresponds to the linear velocity control instructions on the XYZ three axes, and (uθ, uϕ, uψ) corresponds to the angular velocity control instructions on the XYZ three axes, and sent to the flight control system.

[0117] The autonomous obstacle avoidance system achieves high precision and dynamic adaptability in the drone's autonomous obstacle avoidance and navigation; by integrating LiDAR and visual sensors and combining deep learning models, it achieves accurate perception of complex environments and obstacle recognition; at the same time, using advanced path planning algorithms and dynamic adjustment mechanisms, the drone can flexibly plan and adjust the flight path in the constructed three-dimensional map, effectively respond to environmental changes and sudden obstacles, and significantly improve the intelligence level and safety of autonomous obstacle avoidance and navigation.

[0118] The fifth implementation mode is mainly different from the first implementation mode in that the operation steps of the internal navigation model include:

[0119] D1. Environmental map construction: Use SLAM technology to combine PL and T to build a three-dimensional map S;

[0120] D2, path planning: Use A* or RRT algorithm to plan the path L on the three-dimensional map S;

[0121] D3, Dynamic Adjustment: Re-evaluate and adjust L based on new obstacle information and environmental changes.

[0122] By building an accurate three-dimensional environmental map through SLAM technology and combining it with an efficient path planning algorithm, the drone can autonomously plan the optimal flight path; at the same time, the model has dynamic adjustment capabilities, can respond to environmental changes and obstacle information in real time, and flexibly adjust the flight path to ensure that the drone completes its mission safely and efficiently, significantly improving the drone's autonomy and adaptability in complex scenarios.

[0123] The sixth implementation mode is mainly different from the first implementation mode in that the operation steps of the multi-element fusion optimization module include:

[0124] D1. Multi-sensor data fusion:

[0125] Using Bayesian filtering or particle filtering algorithm (Bayesian filtering is a probabilistic inference method based on Bayesian theorem and recursive Bayesian estimation, mainly used to estimate the system state over time. It is widely used in target tracking, robot positioning and navigation, speech recognition and other fields to solve the problem of system state estimation under noise or incomplete observation; Particle Filter (PF) is a filtering algorithm based on Monte Carlo methods. It approximates the probability density function by finding a set of random samples propagating in the state space, and then obtains the minimum variance estimate of the system state. These samples are figuratively called "particles", hence the name particle filtering), LiDAR, visual sensor, IMU and GPS multi-source data are integrated to obtain a more accurate environmental perception and drone state estimation formula:

[0126] Belief Update=Fuse(PL,O,a,ω,m,F GPS );

[0127] Among them, Fuse represents the fusion function, which integrates the information of each sensor through probability theory or statistical methods according to the characteristics and uncertainty of sensor data, and outputs a state estimation or environmental model;

[0128] D2. System optimization:

[0129] Optimize parameters of obstacle avoidance algorithm and path planning algorithm.

[0130] Obstacle avoidance algorithm optimization includes:

[0131] Parameter tuning: According to the actual flight environment and mission requirements of the UAV, adjust the key parameters in the obstacle avoidance algorithm, including obstacle detection threshold, optimization target of obstacle avoidance path planning; optimize the target packet time and minimize energy consumption;

[0132] Algorithm selection: For complex dynamic environments, use a real-time obstacle avoidance algorithm based on deep learning; for static or simple environments, use a geometric or grid search algorithm;

[0133] Fusion of multiple algorithms: Combine multiple obstacle avoidance algorithms and switch between different obstacle avoidance algorithms for different detection environments.

[0134] Path planning algorithm optimization includes:

[0135] Improved algorithm efficiency: Optimized the calculation process of the path planning algorithm and used priority queues to store and update candidate paths;

[0136] Multi-objective optimization: multiple optimization objectives are considered during path planning, including the shortest path, the lowest energy consumption, and the highest safety. Weights are set for multiple optimization objectives, and the path with the smallest weighted value is taken as the best path.

[0137] Real-time improvement: For dynamic environments, incremental path planning algorithms or rolling horizon planning algorithms are used to quickly respond to environmental changes.

[0138] Through multi-sensor data fusion, the accuracy of environmental perception and state prediction has been significantly improved; at the same time, system optimization ensures the best performance of obstacle avoidance and path planning algorithms in different environments, including parameter tuning, algorithm selection and fusion, efficiency improvement, multi-objective optimization and real-time improvement, so that drones can respond to complex and changeable flight missions more intelligently and flexibly, and improve the autonomy and reliability of the overall system.

[0139] Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0140] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0141] Although 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 the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An autonomous flight control system for unmanned aerial vehicles without external navigation, characterized in that: include: The sensor unit integrates LiDAR, IMU, GPS, vision sensor and ultrasonic sensor; Flight control system, used to output control signals to control the autonomous flight of the UAV; The attitude and coordinate detection unit reads the IMU data in real time, calculates the attitude and spatial coordinates of the drone through algorithms, and feeds back to the flight control system for attitude adjustment and spatial positioning; The posture and coordinate detection unit comprises: A1. IMU data reading: real-time reading of IMU data, including: acceleration data a=(a X , a Y , a Z ), angular velocity data ω=(ω X ,ω Y ,ω Z ), magnetic field strength data m=(m X , m Y , m Z ), X, Y, and Z represent the values ​​obtained based on the XYZ three-axis coordinates; A2. Attitude solution: Use Kalman filter or extended Kalman filter algorithm, combined with IMU and GPS data F GPS =(X GPS , Y GPS , Z GPS ), calculate the estimated posture q=(q0, q1, q2, q3) and spatial coordinates p=(X, Y, Z) of the drone; A3, data feedback: Feedback the estimated attitude q and spatial coordinates p to the flight control system for attitude adjustment and spatial positioning; The displacement detection module obtains the speed and displacement of the drone through integral calculation based on the acceleration data of the IMU, and corrects the integral result in combination with the visual odometer or GPS data. The displacement detection result is compared with the expected target, and closed-loop control is performed through the flight control system to make the drone fly along the predetermined trajectory; Autonomous obstacle avoidance system, which is used to scan the surrounding environment to obtain obstacle distance and location information. The obstacle avoidance algorithm processes the collected data to identify and classify obstacles, and calculates the best obstacle avoidance path among multiple obstacle avoidance paths. At the same time, it generates control instructions to adjust the drone's attitude and speed to avoid obstacles; The operation steps of the autonomous obstacle avoidance system include: C1. Data collection: Use LiDAR to scan the environment and obtain point cloud data , where N is the number of points in the point cloud, and r is any integer from 1 to N; The visual sensor captures the image T and obtains the obstacle information O through the image processing algorithm; C2. Data processing: Input PL and O into the obstacle avoidance algorithm, use the deep learning model or rule-based classifier C, output the obstacle type T and position (X0, Y0, Z0), and the output expression is: T, (X0, Y0, Z0)=C(PL, O); C3, Decision generation: Based on the obstacle information, use the A* algorithm or RRT algorithm to calculate the obstacle avoidance path ; Where M is the number of points in the path, j is any integer from 1 to M; C4, Execution Control: P a Converted into flight control instructions U=(uX, uY, uZ, uθ, uϕ, uψ), where (uX, uY, uZ) corresponds to the linear velocity control instructions on the XYZ three axes, and (uθ, uϕ, uψ) corresponds to the angular velocity control instructions on the XYZ three axes, and sent to the flight control system; The internal navigation model builds a 3D map of the surrounding environment and generates a flight path from the starting point to the end point. It also dynamically adjusts the flight path and navigation strategy based on real-time environmental changes and obstacle avoidance requirements. Multi-dimensional fusion optimization module, which uses data fusion algorithm to fuse data from different sensors and optimizes obstacle avoidance algorithm and path planning algorithm; The operation steps of the multivariate fusion optimization module include: D1. Multi-sensor data fusion: Use Bayesian filtering or particle filtering algorithms to fuse LiDAR, visual sensors, IMU and GPS multi-source data to obtain more accurate environmental perception and drone state estimation formulas: Belief Update=Fuse(PL,O,a,ω,m,F GPS ); Among them, Fuse represents the fusion function, which integrates the information of each sensor through probability theory or statistical methods according to the characteristics and uncertainty of sensor data, and outputs a state estimation or environmental model; D2. System optimization: Optimize parameters of obstacle avoidance algorithm and path planning algorithm.

2. The autonomous flight control system for unmanned aerial vehicles without external navigation according to claim 1, characterized in that: The sensor unit further comprises: Processor unit for real-time data processing and algorithm execution; Power management module, providing stable power supply to support flight and sensor operation; Storage device for recording flight data and algorithm models.

3. The autonomous flight control system for unmanned aerial vehicles without external navigation according to claim 1, characterized in that: The operation steps of the displacement detection module include: B1. Integral calculation: Use numerical integration to calculate velocity and displacement p from acceleration data a t ; ； p t =p t-1 + 5 t Δt; Among them, V t and V t−1 denote the velocity vector at time t and t−1 respectively, a t and a t−1 denote the acceleration vectors at time t and t−1 respectively, Δt is the time interval, and p t and p t−1 Represent the position vectors at time t and t−1 respectively; B2. Error correction: Combine visual odometer or GPS data for error correction and adjust p t ; B3. Closed-loop control: Let the predetermined trajectory be p target , the corrected displacement is compared with the predetermined trajectory p target In contrast, the flight state is adjusted through PID controller control.

4. The autonomous flight control system for unmanned aerial vehicles without external navigation according to claim 1, characterized in that: The steps to adjust the flight state through PID controller are: B3.

1. Calculation error: The error e is the current position p t With the target position p target The vector difference between them is: e=p target − p t ; B3.2, PID controller: Use the PID control algorithm to calculate the control signal u. The output of the PID controller is the weighted sum of the proportional P, integral I and differential D: ； Among them, Kp, Ki and Kd are the gains corresponding to proportional P, integral I and differential D respectively; B3.

3. Adjust the flight status: According to the output u of the PID controller, adjust the control input of the aircraft to reduce the error and move the aircraft to the target position.

5. The autonomous flight control system for unmanned aerial vehicles without external navigation according to claim 1, characterized in that: The operation steps of the internal navigation model include: D1. Environmental map construction: Use SLAM technology to combine PL and T to build a three-dimensional map S; D2, path planning: Use A* or RRT algorithm to plan the path L on the three-dimensional map S; D3, Dynamic Adjustment: Re-evaluate and adjust L based on new obstacle information and environmental changes.

6. The autonomous flight control system for unmanned aerial vehicles without external navigation according to claim 1, characterized in that: Obstacle avoidance algorithm optimization includes: Parameter tuning: According to the actual flight environment and mission requirements of the UAV, adjust the key parameters in the obstacle avoidance algorithm, including obstacle detection threshold, optimization target of obstacle avoidance path planning; optimize the target packet time and minimize energy consumption; Algorithm selection: For complex dynamic environments, use a real-time obstacle avoidance algorithm based on deep learning; for static or simple environments, use a geometric or grid search algorithm; Fusion of multiple algorithms: Combine multiple obstacle avoidance algorithms and switch between different obstacle avoidance algorithms for different detection environments.

7. The autonomous flight control system for unmanned aerial vehicles without external navigation according to claim 1, characterized in that: Path planning algorithm optimization includes: Improved algorithm efficiency: Optimized the calculation process of the path planning algorithm and used priority queues to store and update candidate paths; Multi-objective optimization: multiple optimization objectives are considered during path planning, including the shortest path, the lowest energy consumption, and the highest safety. Weights are set for multiple optimization objectives, and the path with the smallest weighted value is taken as the best path. Real-time improvement: For dynamic environments, incremental path planning algorithms or rolling horizon planning algorithms are used to quickly respond to environmental changes.

Images