Unmanned aerial vehicle low-altitude precision navigation and obstacle avoidance method based on multi-source fusion perception

Abstract

The present application relates to the field of unmanned aerial vehicle navigation and obstacle avoidance processing, and specifically discloses a method for precise low-altitude navigation and obstacle avoidance of unmanned aerial vehicles based on multi-source fusion perception, which comprises: when the unmanned aerial vehicle performs a navigation and obstacle avoidance task in a low-altitude environment, starting a multi-source fusion perception system, which is configured with at least four perception sensors with independent data acquisition capabilities, and each perception sensor is provided with an independent switch node for precisely controlling the acquisition start and stop of the corresponding sensor. The present application constructs a phased energy consumption modeling optimization strategy, sets different power coefficients according to the energy consumption characteristics of different sensor acquisition stages, and simultaneously considers the energy consumption recovery of the waiting stage, thereby realizing the fine control of sensor acquisition energy consumption. Through the energy consumption target function, the total energy consumption of the system is matched with the energy consumption threshold of the unmanned aerial vehicle operation, thereby avoiding excessive consumption of on-board energy and effectively improving the endurance of the unmanned aerial vehicle in low-altitude operation.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) navigation and obstacle avoidance technology, and in particular to a method for low-altitude precision navigation and obstacle avoidance of UAVs based on multi-source fusion perception. Background Technology

[0002] Low-altitude navigation and obstacle avoidance for drones is a core technology for both civilian and industrial applications. The low-altitude environment is complex, containing buildings, trees, cables, and obstacles, and is greatly affected by weather, electromagnetic fields, and terrain, demanding extremely high accuracy in environmental perception and obstacle avoidance for drones. Multi-source fusion perception technology, by integrating detection data from multiple sensors such as LiDAR, visual cameras, ultrasonic sensors, and millimeter-wave radar, enables comprehensive and multi-dimensional perception of the low-altitude environment, becoming the mainstream technical solution for low-altitude navigation and obstacle avoidance for drones.

[0003] Current multi-source fusion perception navigation and obstacle avoidance systems for drones generally suffer from the technical pain point of data time misalignment: different sensors have inherent data acquisition delays due to differences in hardware technology and detection principles (such as delay t1 for lidar, delay t2 for visual cameras, delay t3 for ultrasonic sensors, and delay t4 for millimeter-wave radar), and the complex low-altitude environment further amplifies these delay differences; existing systems do not have a targeted acquisition timing control mechanism designed to address these delay differences, and directly fuse the data collected by each sensor, resulting in significant deviations in the fused data due to time misalignment. This makes it impossible to accurately reflect the real-time low-altitude environment of the drone, leading to errors in obstacle avoidance path planning and even drone collision accidents.

[0004] In addition, the existing system has the following problems: First, it does not integrate a data acquisition rhythm adjustment model driven by "sensor delay characteristics and environmental interference", so it cannot dynamically adjust the acquisition duration according to the sensor's own delay and environmental changes, resulting in a mismatch between the acquisition rhythm and environmental requirements; Second, the sensor energy allocation lacks fine-grained control and does not consider the energy consumption differences at different acquisition stages, which can easily lead to excessive consumption of airborne energy and reduce the drone's low-altitude operation endurance; Third, it does not realize the linkage and timing scheduling of acquisition rhythms among multiple sensors, and synchronous acquisition by sensors can easily cause data congestion, further reducing the fusion accuracy.

[0005] To address the aforementioned issues, this invention proposes a method for precise low-altitude navigation and obstacle avoidance of unmanned aerial vehicles (UAVs) based on multi-source fusion perception. By configuring independent switching nodes for sensors and constructing a simple and easy-to-understand delay time model and time coordination strategy, the method achieves peak-shifting of sensor data collection to avoid time misalignment. At the same time, it combines energy consumption optimization strategies to achieve airborne energy management, ultimately improving the accuracy of multi-source data fusion and the precision of UAV low-altitude obstacle avoidance. Furthermore, the mathematical model of this invention is simple and easy to implement, requiring no complex hardware modifications, and is adaptable to various small low-altitude UAVs. Summary of the Invention

[0006] To overcome the shortcomings of existing technologies, this invention proposes a method for precise low-altitude navigation and obstacle avoidance of UAVs based on multi-source fusion perception. This method solves the technical problems of data time misalignment, large fusion deviation, and low low-altitude obstacle avoidance caused by different acquisition delays of each sensor in existing multi-sensor fusion obstacle avoidance systems. At the same time, it enables refined control of sensor acquisition energy consumption, thereby improving the endurance and stability of UAVs in low-altitude operations.

[0007] To solve the above-mentioned technical problems, the basic technical solution proposed by this invention is as follows:

[0008] A method for precise low-altitude navigation and obstacle avoidance of UAVs based on multi-source fusion perception includes: when the UAV performs navigation and obstacle avoidance tasks in a low-altitude environment, activating a multi-source fusion perception system, wherein the multi-source fusion perception system is configured with at least four perception sensors with independent data acquisition capabilities, and each perception sensor is configured with an independent switch node, wherein the switch node is used to precisely control the start and stop of the acquisition of the corresponding sensor.

[0009] The data acquisition process for each sensor is divided into three stages in sequence: sensing start-up, data transmission, and waiting for scheduling.

[0010] The hardware attributes and low-altitude environment sensing parameters of each sensor are acquired, and each sensor is associated with a switch node of the multi-source fusion sensing system. The multi-source fusion sensing system is then controlled to perform data acquisition operations, specifically:

[0011] For any given sensor, obtain its hardware acquisition delay coefficient, environmental interference coefficient, and detection range; set the sensor's basic acquisition frequency; execute a delay time modeling strategy to calculate the duration of each sensor's data acquisition in the three stages of sensing initiation, data transmission, and waiting for scheduling.

[0012] Set the target period for data fusion, which is the total time required for the multi-source fusion sensing system to complete one round of data acquisition from all sensors and achieve precise temporal fusion.

[0013] Set the start and end times for each of the three stages of data acquisition from each sensor.

[0014] The execution time coordination strategy coordinates the start and stop of each switching node based on the delay time of each sensor, calculates the start time of each stage of each sensor, and optimizes the total data acquisition time of the overall sensors.

[0015] Implement energy consumption modeling and optimization strategies, calculate the energy consumption of each sensor at each stage of data acquisition, and optimize the energy consumption objective function to achieve a reasonable allocation of airborne energy.

[0016] Based on time-series precise fusion of multi-source sensor data, the system performs obstacle identification and obstacle avoidance path planning in the low-altitude environment of UAVs, completes precise navigation and obstacle avoidance, and feeds back the obstacle avoidance planning results to the UAV flight control system in real time.

[0017] Preferably, the hardware architecture of the multi-source fusion sensing system includes:

[0018] A number of perception drive modules are set up to match the number of sensors. The perception drive modules are distributed and symmetrically installed on the fuselage of the UAV and are connected to the central fusion control module. Each perception drive module can independently drive the corresponding sensor to perform data acquisition operations.

[0019] The central fusion control module includes an airborne energy storage management unit and a central timing linkage unit;

[0020] The airborne energy storage management unit is used to store the energy of the UAV's airborne power supply and to allocate energy to each perception drive module and sensor.

[0021] Among them, the idle energy generated by the sensor during the waiting scheduling phase is recovered and stored in the airborne energy storage management unit, and the airborne energy storage management unit outputs energy to power the sensor's sensing start-up and data transmission phases;

[0022] The central timing linkage unit is used to couple each sensing drive module and control the start and stop of each switch node according to the calculation result of the time coordination strategy, so as to realize the parallel and time-sequential data acquisition operation of each sensor.

[0023] A signal filtering unit is provided at the signal transmission end of each of the sensing drive modules. The signal filtering unit is used to reduce the interference of complex low-altitude environments on sensor data transmission and improve the accuracy of the original acquired data.

[0024] Preferably, the execution delay time modeling strategy calculates the duration of each stage of data acquisition for each sensor, including:

[0025] Let the hardware acquisition delay coefficient of the nth sensor be denoted as k_{dn}}, the environmental interference coefficient be denoted as k_{hn}}, and the detection distance be denoted as L. n n = 1, 2, 3, 4;

[0026] Obtain the sensor's base acquisition frequency f and set an empirical correction coefficient k (k is a constant between 0.8 and 1.2).

[0027] Calculate the duration of the sensor's sensing initiation phase. ;

[0028] Calculate the duration of the sensor data transmission phase. ;

[0029] Calculate the duration of the sensor waiting for scheduling. ;in, Add the average data transmission time to the start-up of all sensor sensing. Minimum waiting time for the sensor ( (a constant of 0.01-0.05 s).

[0030] The duration of the waiting and scheduling phase is dynamically adjusted based on the difference between the total duration of the first two phases of the sensor and the average duration, thereby achieving a balance in the total acquisition time of each sensor.

[0031] Preferably, the execution time coordination strategy coordinates the start and stop of each switching node, calculates the start time of each stage for each sensor, and optimizes the overall data acquisition time of the sensors, including:

[0032] Calculate the start time of each stage for the nth sensor, where n = 1, 2, 3, 4;

[0033] The start time of the perception initiation phase is denoted as The start time of the data transmission phase is denoted as The start time of the waiting scheduling phase is denoted as ;

[0034] Sort each sensor by its hardware acquisition delay coefficient k_{dn}} from smallest to largest, and prioritize activating the switch nodes of sensors with the smallest k_{dn}} to coordinate the sensing startup phase. ,in, The base duration of the sensor start-up interval ( (A constant of 0.02~0.08s) to achieve staggered triggering of sensor activation;

[0035] Coordinated data transmission phase: That is, after the sensor sensing start-up phase ends, the data transmission phase begins immediately;

[0036] Coordination and waiting for scheduling phase: That is, after the sensor data transmission phase ends, it immediately enters the waiting scheduling phase.

[0037] Preferably, the execution time coordination strategy further includes optimizing the overall total data acquisition time of the sensors, specifically:

[0038] Calculate the total data acquisition time T(n) of the nth sensor, where the total time is the sum of the durations of each stage: Obtain the set data fusion target period duration. ;

[0039] The objective function for optimizing computation time is to minimize the deviation between the total acquisition time of each sensor and the target cycle time. Through dynamic adjustment or The values ​​are set so that the total acquisition time of each sensor is as close as possible to the target cycle time, ensuring that the multi-source data is fused in a timely and accurate manner within a fixed period.

[0040] Preferably, the energy consumption modeling and optimization strategy involves calculating the energy consumption of each sensor data acquisition stage and optimizing the energy consumption objective function, including:

[0041] Calculate the energy consumption of the nth sensor at each stage of the data acquisition process, and set the sensor's basic operating power as [value missing]. (A constant, determined by the sensor hardware parameters);

[0042] Energy consumption during the computational sensing initiation phase ;

[0043] Energy consumption during the data transmission phase of the calculation The power factor during the data transmission stage is 1.2, which is suitable for the energy consumption requirements of signal transmission.

[0044] Calculate the energy recovery during the waiting scheduling phase. During the waiting and scheduling phase, the sensor operates at low power to achieve partial energy recovery.

[0045] Among them, positive energy consumption indicates the consumption of energy by the airborne energy storage management unit, and positive energy recovery indicates the replenishment of energy to the airborne energy storage management unit.

[0046] Preferably, the energy consumption modeling optimization strategy further includes optimizing the energy consumption objective function, specifically:

[0047] Calculate the net energy consumption of the entire data acquisition process for the nth sensor. ;

[0048] Calculate the total net energy consumption of the multi-source fusion sensing system ;

[0049] Set the energy consumption threshold per unit time for low-altitude drone operations The energy consumption optimization objective function is: By adjusting the sensor's basic acquisition frequency f, the total net energy consumption of the system can be matched with the energy consumption threshold, thus avoiding excessive consumption or waste of airborne energy.

[0050] The beneficial effects of this invention are:

[0051] This invention configures an independent switching node for each sensor, divides data acquisition into three stages, and constructs a simple and easy-to-understand delay time modeling strategy. The duration of each stage is calculated using basic parameters such as hardware acquisition delay coefficient and environmental interference coefficient. The model formula has no complex calculations, is easy to implement and debug, and is compatible with embedded systems of various small UAVs. At the same time, the waiting scheduling time is dynamically adjusted according to the total duration of the first two stages of the sensor, so as to achieve the equalization of the acquisition time of each sensor and lay the foundation for time-series fusion.

[0052] This invention employs a time coordination strategy, prioritizing the activation of switch nodes based on the sensor hardware acquisition delay coefficient, and achieving staggered sensor acquisition through fixed intervals. This avoids data time misalignment issues from the source, significantly improving the accuracy of multi-source data fusion. The fused data accurately reflects the real-time low-altitude environment of the UAV, providing reliable data support for precise obstacle avoidance.

[0053] This invention constructs a phased energy consumption modeling and optimization strategy, sets different power coefficients according to the energy consumption characteristics of different sensor acquisition stages, and considers energy recovery during the waiting stage, thereby realizing refined control of sensor acquisition energy consumption; by matching the total system energy consumption with the UAV operation energy consumption threshold through the energy consumption objective function, excessive consumption of airborne energy is avoided, and the endurance of UAV low-altitude operation is effectively improved.

[0054] This invention realizes the linkage of perception rhythm and intelligent scheduling of acquisition timing of multi-source sensors. It coordinates the design of sensor acquisition timing, data fusion accuracy and airborne energy consumption optimization, and the three are linked and dynamically optimized to significantly improve the navigation and obstacle avoidance accuracy and stability of UAVs in low-altitude complex environments. Moreover, this invention does not require complex modification of the existing hardware of UAVs, and can be achieved only through software algorithm optimization, which has low modification cost and strong practicality.

[0055] This invention sets up a signal filtering unit and a central timing linkage unit in a multi-source fusion sensing system. The signal filtering unit reduces the interference of the low-altitude environment on data transmission and improves the accuracy of the raw data. The central timing linkage unit realizes parallel and time-sequential acquisition of each sensor, which not only ensures acquisition efficiency but also avoids data congestion, further improving the overall performance of the system. Attached Figure Description

[0056] Figure 1 This is a schematic diagram of the structure of the present invention; Detailed Implementation

[0057] The following will be combined with the appendix Figure 1The technical solutions in the embodiments of the present invention have been clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0058] A method for precise low-altitude navigation and obstacle avoidance of unmanned aerial vehicles (UAVs) based on multi-source fusion perception includes: when the UAV performs navigation and obstacle avoidance tasks in a low-altitude environment, activating a multi-source fusion perception system. The multi-source fusion perception system is equipped with at least four sensing sensors with independent data acquisition capabilities. Each sensing sensor is assigned an independent switch node, which is used to precisely control the activation and deactivation of the corresponding sensor's data acquisition. The hardware architecture of the multi-source fusion perception system includes: setting a number of sensing drive modules matching the number of sensors. These sensing drive modules are distributed and symmetrically installed on the UAV fuselage and connected to a central fusion control module. Each sensing drive module can independently drive its corresponding sensor to perform data acquisition operations. The central fusion control module includes an onboard storage... The system comprises an energy management unit and a central timing linkage unit. The airborne energy storage management unit stores the energy from the UAV's onboard power supply and allocates energy to each sensing drive module and sensor. Idle energy generated by the sensors during the waiting scheduling phase is recovered and stored in the airborne energy storage management unit, which outputs energy to power the sensor's sensing startup and data transmission phases. The central timing linkage unit couples each sensing drive module and controls the start and stop of each switching node based on the calculation results of the time coordination strategy, enabling parallel and time-sequential data acquisition operations for each sensor. A signal filtering unit is set at the signal transmission end of each sensing drive module to reduce interference from the complex low-altitude environment on sensor data transmission and improve the accuracy of the original acquired data. The system executes an energy consumption modeling optimization strategy, calculates the energy consumption of each sensor at each stage of data acquisition, and optimizes the energy consumption objective function, including: calculating the energy consumption of the nth sensor at each stage of data acquisition, and setting the sensor's basic operating power to be... (A constant, determined by the sensor hardware parameters); calculate the energy consumption during the sensing initiation phase. ;Calculate the energy consumption during the data transmission phase : The power factor during the data transmission phase is 1.2 to meet the energy consumption requirements of signal transmission; the energy recovery during the waiting scheduling phase is calculated. During the waiting and scheduling phase, the sensors operate at low power to achieve partial energy recovery. Here, positive energy consumption indicates the consumption of energy by the onboard energy storage management unit, and positive energy recovery indicates the replenishment of energy to the onboard energy storage management unit. The energy consumption modeling and optimization strategy also includes optimizing the energy consumption objective function, specifically: calculating the net energy consumption of the entire data acquisition process for the nth sensor. ; Calculate the total net energy consumption of the multi-source fusion sensing system Set the energy consumption threshold per unit time for low-altitude drone operations. The energy consumption optimization objective function is: By adjusting the sensor's basic acquisition frequency f, the total net energy consumption of the system can be matched with the energy consumption threshold, thus avoiding excessive consumption or waste of airborne energy.

[0059] The data acquisition process for each sensor is divided into three stages in sequence: sensing start-up, data transmission, and waiting for scheduling.

[0060] The hardware attributes and low-altitude environment sensing parameters of each sensor are acquired, and each sensor is associated with a switch node of the multi-source fusion sensing system. The multi-source fusion sensing system is then controlled to perform data acquisition operations, specifically:

[0061] For any given sensor, obtain its hardware acquisition delay coefficient, environmental interference coefficient, and detection range; set the sensor's basic acquisition frequency; execute a delay time modeling strategy to calculate the duration of each of the three stages of data acquisition for each sensor: sensing start-up, data transmission, and waiting for scheduling; the execution of the delay time modeling strategy to calculate the duration of each stage of data acquisition for each sensor includes: representing the hardware acquisition delay coefficient of the nth sensor as k_{dn}}, the environmental interference coefficient as k_{hn}}, and the detection range as... n=1, 2, 3, 4; obtain the sensor's basic acquisition frequency f, and set an empirical correction coefficient k (k is a constant between 0.8 and 1.2); calculate the duration of the sensor's sensing start-up phase. ; Calculate the duration of the sensor data transmission phase ; Calculate the duration of the sensor waiting for scheduling. ;in, Add the average data transmission time to the start-up of all sensor sensing. Minimum waiting time for the sensor ( (A constant between 0.01 and 0.05 s); the duration of the waiting scheduling phase is dynamically adjusted based on the difference between the total duration of the first two phases of the sensor and the average duration, thereby achieving a balance in the total data acquisition time of each sensor. The execution time coordination strategy coordinates the start and stop of each switching node, calculates the start time of each phase for each sensor, and optimizes the overall total data acquisition time of the sensors, including: calculating the start time of each phase for the nth sensor, n = 1, 2, 3, 4; and recording the start time of the sensing start-up phase as... The start time of the data transmission phase is denoted as The start time of the waiting scheduling phase is denoted as Sort each sensor by its hardware acquisition delay coefficient k_{dn}} from smallest to largest, and prioritize activating the sensor switching nodes with the smallest k_{dn}} to coordinate the sensing startup phase. ,in, The base duration of the sensor start-up interval ( (A constant of 0.02~0.08s) to achieve staggered triggering of sensor activation; coordinating the data transmission phase: That is, after the sensor sensing initiation phase ends, the data transmission phase begins immediately; the coordination and waiting scheduling phase follows. That is, after the sensor data transmission phase ends, the system immediately enters the waiting and scheduling phase. The execution time coordination strategy also includes optimizing the overall total data acquisition time of the sensors, specifically: calculating the total data acquisition time T(n) of the nth sensor, where the total time is the sum of the durations of each phase. Obtain the set data fusion target period duration. The objective function for optimizing computation time is to minimize the deviation between the total acquisition time of each sensor and the target cycle time. Through dynamic adjustment or The values ​​are set so that the total acquisition time of each sensor is as close as possible to the target cycle time, ensuring that the multi-source data is fused in a timely and accurate manner within a fixed period.

[0062] Set the target period for data fusion, which is the total time required for the multi-source fusion sensing system to complete one round of data acquisition from all sensors and achieve precise temporal fusion.

[0063] Set the start and end times for each of the three stages of data acquisition from each sensor.

[0064] The execution time coordination strategy coordinates the start and stop of each switching node based on the delay time of each sensor, calculates the start time of each stage of each sensor, and optimizes the total data acquisition time of the overall sensors.

[0065] Implement energy consumption modeling and optimization strategies, calculate the energy consumption of each sensor at each stage of data acquisition, and optimize the energy consumption objective function to achieve a reasonable allocation of airborne energy.

[0066] Based on time-series precise fusion of multi-source sensor data, the system performs obstacle identification and obstacle avoidance path planning in the low-altitude environment of UAVs, completes precise navigation and obstacle avoidance, and feeds back the obstacle avoidance planning results to the UAV flight control system in real time.

[0067] Example 1:

[0068] This embodiment takes a low-altitude inspection scenario of a small industrial drone as an example. The drone is equipped with four sensing sensors: lidar (n=1), visual camera (n=2), ultrasonic sensor (n=3), and millimeter-wave radar (n=4). The drone operates at a low altitude of 50 meters and in a complex environment around urban buildings, which places high demands on the acquisition delay and fusion accuracy of each sensor.

[0069] Step 1: Start-up and parameter acquisition of the multi-source fusion sensing system

[0070] When the drone enters a 50-meter low-altitude inspection environment, the flight control system triggers navigation and obstacle avoidance commands, activates the multi-source fusion perception system, and the switch nodes of each sensor are initially in the off state; the basic parameters of each sensor are extracted through the sensor attribute modeling module:

[0071] Hardware acquisition delay coefficient k_{dn}}: LiDAR Visual cameras Ultrasonic sensors Millimeter-wave radar ;

[0072] Environmental interference coefficient k_{hn}}: uniformly set in urban building environments ;

[0073] Detection range LiDAR Visual cameras Ultrasonic sensors Millimeter-wave radar Set the sensor's basic acquisition frequency f=10Hz, empirical correction coefficient k=1.0, and minimum basic waiting time for the sensor. Sensor start-up interval base duration Data fusion target cycle duration Sensor basic operating power Energy consumption threshold per unit time for low-altitude drone operations .

[0074] Step 2: Hardware architecture adaptation for multi-source fusion sensing system

[0075] The multi-source fusion sensing system consists of four distributed, symmetrical sensing drive modules, each mounted around the UAV fuselage and connected to a central fusion control module. The central fusion control module includes an onboard energy storage management unit and a central timing linkage unit. The onboard energy storage management unit recovers idle energy consumed during the sensor's waiting period and provides power for the sensing startup and data transmission phases. The central timing linkage unit couples each sensing drive module to control the precise start and stop of each switching node. Each sensing drive module is configured with an independent switching node and a signal filtering unit at the signal transmission end to reduce electromagnetic and occlusion interference from the urban built environment and improve the accuracy of the raw data. The system associates each of the four sensors with the switching nodes of the sensing drive modules, completing hardware adaptation.

[0076] Step 3: Execute the delay time modeling strategy and calculate the duration of each stage for each sensor.

[0077] The delay time modeling module calculates the sensing activation of each sensor according to the simple formula in claim 3. Data transmission Waiting for scheduling Three-stage duration:

[0078] Startup time is perceived:

[0079] , , , ;

[0080] Data transmission duration:

[0081] , , , ;

[0082] Waiting time for scheduling: ,in

[0083] , , , .

[0084] Step 4: Implement time coordination strategies, regulate the start and stop of switch nodes, and optimize the total data collection time.

[0085] Switch node start / stop control: The hardware acquisition delay coefficients k_{dn} are sorted from smallest to largest as follows: ultrasonic sensor (n=3) < lidar (n=1) < millimeter-wave radar (n=4) < visual camera (n=2), prioritizing the activation of sensors with smaller delay coefficients; the start time of each stage for each sensor is calculated according to the formula in claim 4:

[0086] Perception activation: ,Right now , , , ;

[0087] Data transmission: ,Right now , , , ;

[0088] Waiting for scheduling: ,Right now , , , ;

[0089] The central timing linkage unit precisely controls the start and stop of each switch node according to the above time, so as to realize the peak-shifting data collection of the sensor.

[0090] Total data acquisition time optimization: Calculate the total data acquisition time for each sensor. ,get , , , Optimize the objective function based on duration Fine-tuning This allows the total acquisition time of each sensor to be converted to the target period duration. By closely monitoring the data, the system ensures that multi-source data can be fused in a time sequence within a 0.5s cycle, thus meeting the real-time requirements of low-altitude obstacle avoidance for drones.

[0091] Step 5: Implement energy consumption modeling and optimization strategies to calculate energy consumption and optimize energy allocation.

[0092] The energy consumption modeling module is used to calculate the energy consumption, net energy consumption, and total net energy consumption of each sensor at each stage.

[0093] Energy consumption / recovery at each stage:

[0094] Startup power consumption: ;

[0095] Data transmission power consumption: ;

[0096] Amount awaiting scheduling and recovery: ;

[0097] Net power consumption per sensor: ;

[0098] Total net energy consumption of the system: ; Optimize the objective function based on energy consumption Fine-tuning the sensor's basic acquisition frequency This makes the system's total net energy consumption equal to Matching is achieved to avoid excessive consumption of airborne energy; the airborne energy storage management module allocates energy to each sensor and recovers energy consumed during the waiting period based on the calculation results, thereby realizing the rational scheduling of airborne energy.

[0099] Step 6: Multi-source data fusion and low-altitude precision obstacle avoidance

[0100] The AI-integrated obstacle avoidance module receives raw data from various sensors, collected in a precise, time-series manner. First, it removes environmental noise using a feature extraction algorithm, then employs a weighted fusion algorithm to accurately fuse the multi-source data. Based on the fused data, it uses the YOLO target detection algorithm to identify and locate low-altitude buildings, cables, obstacles, and other targets. Combining this with the drone's current flight speed, altitude, and heading, it uses the A* algorithm to plan the optimal obstacle avoidance path. The obstacle avoidance path planning results are fed back to the drone's flight control system in real time. The flight control system adjusts the drone's flight attitude according to instructions to achieve precise low-altitude navigation and obstacle avoidance. Simultaneously, it feeds back changes in environmental interference during obstacle avoidance to the sensor attribute modeling module, dynamically adjusting the environmental interference coefficient k_{hn}} to achieve adaptive optimization of the sensor's collected parameters.

[0101] Step 7: Execute repeatedly

[0102] During low-altitude inspections, the duration of data fusion targets is determined by the number of drones. Using this as a unit, steps 3-6 are executed cyclically to continuously complete multi-source sensor time-series acquisition, data fusion, obstacle identification, and path planning, achieving precise navigation and obstacle avoidance throughout the entire process.

[0103] Based on the disclosure and teachings of the foregoing specification, those skilled in the art can make changes and modifications to the above embodiments. Therefore, the present invention is not limited to the specific embodiments disclosed and described above, and some modifications and changes to the present invention should also fall within the protection scope of the claims of the present invention. Furthermore, although some specific terms are used in this specification, these terms are only for convenience of explanation and do not constitute any limitation on the present invention.

Claims

1. A method for low-altitude precision navigation and obstacle avoidance of unmanned aerial vehicles (UAVs) based on multi-source fusion perception, characterized in that, include: When the UAV performs navigation and obstacle avoidance tasks in a low-altitude environment, it activates a multi-source fusion perception system. The multi-source fusion perception system is equipped with at least four perception sensors with independent data acquisition capabilities. Each perception sensor is equipped with an independent switch node, which is used to precisely control the start and stop of the corresponding sensor's data acquisition. The data acquisition process for each sensor is divided into three stages in sequence: sensing start-up, data transmission, and waiting for scheduling. The hardware attributes and low-altitude environment sensing parameters of each sensor are acquired, and each sensor is associated with a switch node of the multi-source fusion sensing system. The multi-source fusion sensing system is then controlled to perform data acquisition operations, specifically: For any given sensor, obtain its hardware acquisition delay coefficient, environmental interference coefficient, and detection range; set the sensor's basic acquisition frequency; execute a delay time modeling strategy to calculate the duration of each sensor's data acquisition in the three stages of sensing initiation, data transmission, and waiting for scheduling. Set the target period for data fusion, which is the total time required for the multi-source fusion sensing system to complete one round of data acquisition from all sensors and achieve precise temporal fusion. Set the start and end times for each of the three stages of data acquisition from each sensor. The execution time coordination strategy coordinates the start and stop of each switching node based on the delay time of each sensor, calculates the start time of each stage of each sensor, and optimizes the total data acquisition time of the overall sensors. Implement energy consumption modeling and optimization strategies, calculate the energy consumption of each sensor at each stage of data acquisition, and optimize the energy consumption objective function to achieve a reasonable allocation of airborne energy. Based on time-series precise fusion of multi-source sensor data, the system performs obstacle identification and obstacle avoidance path planning in the low-altitude environment of UAVs, completes precise navigation and obstacle avoidance, and feeds back the obstacle avoidance planning results to the UAV flight control system in real time.

2. The method for low-altitude precision navigation and obstacle avoidance of unmanned aerial vehicles based on multi-source fusion perception according to claim 1, characterized in that, The hardware architecture of the multi-source fusion sensing system includes: A number of perception drive modules are set up to match the number of sensors. The perception drive modules are distributed and symmetrically installed on the fuselage of the UAV and are connected to the central fusion control module. Each perception drive module can independently drive the corresponding sensor to perform data acquisition operations. The central fusion control module includes an airborne energy storage management unit and a central timing linkage unit; The airborne energy storage management unit is used to store the energy of the UAV's airborne power supply and to allocate energy to each perception drive module and sensor. Among them, the idle energy generated by the sensor during the waiting scheduling phase is recovered and stored in the airborne energy storage management unit, and the airborne energy storage management unit outputs energy to power the sensor's sensing start-up and data transmission phases; The central timing linkage unit is used to couple each sensing drive module and control the start and stop of each switch node according to the calculation result of the time coordination strategy, so as to realize the parallel and time-sequential data acquisition operation of each sensor. A signal filtering unit is provided at the signal transmission end of each of the sensing drive modules. The signal filtering unit is used to reduce the interference of complex low-altitude environments on sensor data transmission and improve the accuracy of the original acquired data.

3. The method for low-altitude precision navigation and obstacle avoidance of unmanned aerial vehicles based on multi-source fusion perception according to claim 1, characterized in that, The execution delay time modeling strategy calculates the duration of each stage of data acquisition for each sensor, including: Let the hardware acquisition delay coefficient of the nth sensor be denoted as k_{dn}}, the environmental interference coefficient be denoted as k_{hn}}, and the detection distance be denoted as L. n n = 1, 2, 3, 4; Obtain the sensor's basic acquisition frequency f, and set an empirical correction coefficient k, where k is a constant between 0.8 and 1.2; Calculate the duration T1(n) of the sensor's sensing initiation phase: ; Calculate the duration T2(n) of the sensor data transmission phase: ; Calculate the duration T3(n) of the sensor waiting scheduling phase: T3(n) = |T1(n) + T2(n) − T 均 |+t0; where, Add the average data transmission time to the start-up of all sensor sensing. This is the minimum waiting time for the sensor. A constant of 0.01-0.05 s; The duration of the waiting and scheduling phase is dynamically adjusted based on the difference between the total duration of the first two phases of the sensor and the average duration, thereby achieving a balance in the total acquisition time of each sensor.

4. The method for low-altitude precision navigation and obstacle avoidance of UAVs based on multi-source fusion perception according to claim 3, characterized in that, The execution time coordination strategy coordinates the start and stop of each switching node, calculates the start time of each stage for each sensor, and optimizes the overall data acquisition time of the sensors, including: Calculate the start time of each stage for the nth sensor, where n = 1, 2, 3, 4; Let t1(n) be the start time of the perception initiation phase. start The start time of the data transmission phase is denoted as t2(n). start The start time of the waiting scheduling phase is denoted as t3(n). start ; Sort each sensor by its hardware acquisition delay coefficient k_{dn}} from smallest to largest, and prioritize activating the sensor switching nodes with the smallest k_{dn}} to coordinate the sensing startup phase: t1(n) start =n×t s , where t s t is the base duration of the sensor start-up interval. s A constant value of 0.02~0.08s is used to achieve staggered triggering of sensor activation. Coordinated data transmission phase: That is, after the sensor sensing start-up phase ends, the data transmission phase begins immediately; Coordination and waiting for scheduling phase: That is, after the sensor data transmission phase ends, it immediately enters the waiting scheduling phase.

5. The method for low-altitude precision navigation and obstacle avoidance of unmanned aerial vehicles based on multi-source fusion perception according to claim 4, characterized in that, The execution time coordination strategy also includes optimizing the overall data acquisition time of the sensors, specifically: Calculate the total data acquisition time T(n) of the nth sensor. The total time is the sum of the time of each stage: T(n) = T1(n) + T2(n) + T3(n). Obtain the set data fusion target cycle time T0. The objective function for optimizing computation time is to minimize the deviation between the total acquisition time of each sensor and the target cycle time. By dynamically adjusting t0 or t s The values ​​are set so that the total acquisition time of each sensor is as close as possible to the target cycle time, ensuring that the multi-source data is fused in a timely and accurate manner within a fixed period.

6. The method for low-altitude precision navigation and obstacle avoidance of UAVs based on multi-source fusion perception according to claim 2, characterized in that, The energy consumption modeling and optimization strategy calculates the energy consumption of each sensor at each stage of data acquisition and optimizes the energy consumption objective function, including: Calculate the energy consumption of the nth sensor at each stage of the data acquisition process, and set the basic operating power of the sensor as P0; Energy consumption during the computational sensing initiation phase ; Energy consumption during the data transmission phase of the calculation The power factor during the data transmission stage is 1.2, which is suitable for the energy consumption requirements of signal transmission. Calculate the energy recovery during the waiting scheduling phase. During the waiting and scheduling phase, the sensor operates at low power to achieve partial energy recovery. Among them, positive energy consumption indicates the consumption of energy by the airborne energy storage management unit, and positive energy recovery indicates the replenishment of energy to the airborne energy storage management unit.

7. The method for low-altitude precision navigation and obstacle avoidance of unmanned aerial vehicles based on multi-source fusion perception according to claim 1, characterized in that, The energy consumption modeling and optimization strategy also includes optimizing the energy consumption objective function, specifically: Calculate the net energy consumption of the entire data acquisition process for the nth sensor. ; Calculate the total net energy consumption of the multi-source fusion sensing system ; Set the energy consumption threshold E per unit time for low-altitude drone operations. th The energy consumption optimization objective function is: By adjusting the sensor's basic acquisition frequency f, the total net energy consumption of the system can be matched with the energy consumption threshold, thus avoiding excessive consumption or waste of airborne energy.

Images