An unmanned aerial vehicle automatic collision avoidance method and system fusing multi-source information

Abstract

The application provides an unmanned aerial vehicle automatic collision avoidance method and system fusing multi-source information, and belongs to the unmanned aerial vehicle collision avoidance field.The method comprises the following steps: acquiring an omnidirectional sound sensor data set by omnidirectional acoustic real-time monitoring through a sound sensor array to identify and determine the direction of an invading unmanned aerial vehicle; collecting an image of the invading unmanned aerial vehicle and the distance between the two unmanned aerial vehicles; identifying the model of the invading unmanned aerial vehicle and constructing a virtual electromagnetic protection area; judging whether the invading unmanned aerial vehicle is about to enter the virtual electromagnetic protection area according to the distance between the two unmanned aerial vehicles; and triggering autonomous collision avoidance of the target unmanned aerial vehicle when the invading unmanned aerial vehicle is about to enter the virtual electromagnetic protection area, so that the target unmanned aerial vehicle is kept away from the invading unmanned aerial vehicle.The application solves the technical problem that the unmanned aerial vehicle collision avoidance in the prior art adopts a fixed safety distance and cannot perform individualized collision avoidance according to the electromagnetic interference characteristics of unmanned aerial vehicles of different models, and achieves the technical effects of dynamically adjusting the collision avoidance distance according to the model of the invading unmanned aerial vehicle and realizing individualized electromagnetic interference collision avoidance.

Description

Technical Field

[0001] This invention relates to the field of drone collision avoidance, and in particular to an automatic collision avoidance method and system for drones that integrates multi-source information. Background Technology

[0002] With the rapid development and popularization of drone technology, drones are widely used in aerial photography, recreational flying, and agricultural plant protection. In open airspace, it is common to see multiple drones operated by different users flying simultaneously, which places higher demands on drone collision avoidance technology.

[0003] In scenarios where multiple drones fly simultaneously, differences in technical parameters such as communication frequency bands, transmission power, and antenna configurations between different brands and models of drones can cause electromagnetic interference when the distance between them decreases to a certain range. This electromagnetic interference can affect the stability of the communication link, leading to control command transmission failures or abnormal flight control, which in turn can cause the drones to go out of control and ultimately result in collisions.

[0004] Existing drone collision avoidance methods primarily employ a fixed safety distance strategy, applying the same avoidance standard to all types of drones. This fixed-distance collision avoidance approach fails to account for the differences in electromagnetic interference characteristics among different drone models, leading to improper collision avoidance distance settings and an inability to effectively prevent electromagnetic interference. Collision avoidance is only activated when the drone's control capabilities are compromised due to interference, resulting in poor collision avoidance performance. Summary of the Invention

[0005] This invention addresses the technical problem in existing technologies where drone collision avoidance uses a fixed safety distance and cannot provide personalized collision avoidance based on the electromagnetic interference characteristics of different drone models. It provides a method and system for automatic collision avoidance of drones that integrates multi-source information to solve this problem.

[0006] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0007] In a first aspect, the present invention provides an automatic collision avoidance method for unmanned aerial vehicles (UAVs) that integrates multi-source information, comprising: performing omnidirectional acoustic real-time monitoring through an array of sound sensors to acquire an omnidirectional sound sensing dataset, and identifying and determining the location of an intruding UAV based on the omnidirectional sound sensing dataset; activating corresponding visual and ultrasonic sensors according to the location of the intruding UAV, acquiring an image of the intruding UAV through the visual sensor, and acquiring the distance between the intruding UAV and the target UAV through the ultrasonic sensor; identifying the model of the intruding UAV based on the image of the intruding UAV, acquiring an electromagnetic interference distance constraint corresponding to the model of the intruding UAV, and constructing a virtual electromagnetic protection zone centered on the target UAV and in conjunction with the electromagnetic interference distance constraint; determining whether the intruding UAV is about to enter the virtual electromagnetic protection zone based on the distance between the two UAVs, and triggering the target UAV to autonomously avoid collision when the intruding UAV is about to enter the virtual electromagnetic protection zone, thereby causing the target UAV to move away from the intruding UAV.

[0008] Secondly, this invention provides an automatic collision avoidance system for unmanned aerial vehicles (UAVs) that integrates multi-source information, comprising: an omnidirectional acoustic monitoring module, used for real-time omnidirectional acoustic monitoring through an array of sound sensors, acquiring an omnidirectional sound sensing dataset, and identifying and determining the location of an intruding UAV based on the omnidirectional sound sensing dataset; a visual-ultrasonic activation module, used for activating corresponding visual and ultrasonic sensors according to the location of the intruding UAV, acquiring images of the intruding UAV through the visual sensors, and acquiring the distance between the intruding UAV and the target UAV through the ultrasonic sensors; an electromagnetic protection zone construction module, used for identifying the model of the intruding UAV based on the image of the intruding UAV, acquiring electromagnetic interference distance constraints corresponding to the model of the intruding UAV, and constructing a virtual electromagnetic protection zone centered on the target UAV and in conjunction with the electromagnetic interference distance constraints; and a collision avoidance decision execution module, used for determining whether the intruding UAV is about to enter the virtual electromagnetic protection zone based on the distance between the two UAVs, and triggering the target UAV to autonomously avoid collision when the intruding UAV is about to enter the virtual electromagnetic protection zone, thereby moving the target UAV away from the intruding UAV.

[0009] The beneficial effects of this invention are:

[0010] Omnidirectional acoustic real-time monitoring is performed using an array of sound sensors to acquire an omnidirectional sound sensing dataset. Based on this dataset, the location of the intruding drone is identified, enabling comprehensive perception of drones in the surrounding environment and avoiding blind spots. Corresponding visual and ultrasonic sensors are activated based on the intruding drone's location. The visual sensor acquires images of the intruding drone, while the ultrasonic sensor obtains the distance between the intruding and target drones, achieving precise positioning and distance measurement. This provides accurate data for subsequent model identification and collision avoidance decisions. Based on the intruding drone images, the drone model is identified, and electromagnetic interference distance constraints corresponding to that model are obtained. A virtual electromagnetic protection zone is constructed centered on the target drone, combining these distance constraints to establish personalized safety protection ranges based on the electromagnetic characteristics of different drone models, avoiding insufficient or excessive protection due to fixed distances. The distance between the two drones determines whether the intruding drone is about to enter the virtual electromagnetic protection zone. When the intruding drone is about to enter, the target drone is triggered to autonomously avoid collisions, moving away from the intruding drone. This preventative collision avoidance before electromagnetic interference occurs effectively prevents the risk of uncontrolled collisions caused by electromagnetic interference.

[0011] The above technical solution achieves the effect of dynamically adjusting the collision avoidance distance according to the model of the intruding drone and performing personalized electromagnetic interference collision avoidance. It effectively solves the technical problem in the existing technology that uses a fixed safety distance and cannot perform personalized collision avoidance for different models of drones with different electromagnetic interference characteristics. Attached Figure Description

[0012] Figure 1 A flowchart illustrating an automatic collision avoidance method for unmanned aerial vehicles (UAVs) that integrates multi-source information, provided by the present invention;

[0013] Figure 2 This is a schematic diagram of the structure of an automatic collision avoidance system for unmanned aerial vehicles that integrates multi-source information, provided by the present invention.

[0014] In the attached diagram, the components represented by each number are as follows:

[0015] Omnidirectional acoustic monitoring module 11, visual ultrasonic activation module 12, electromagnetic protection zone construction module 13, collision avoidance decision execution module 14. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0018] In the description of this invention, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed herein.

[0019] Example 1, as Figure 1 As shown, this embodiment of the invention provides an automatic collision avoidance method for unmanned aerial vehicles (UAVs) that integrates multi-source information, including:

[0020] S1. Perform omnidirectional acoustic real-time monitoring through an array of sound sensors to obtain an omnidirectional sound sensing dataset, and identify and determine the location of the intruding drone based on the omnidirectional sound sensing dataset.

[0021] Specifically, the sound sensor array includes multiple sound sensors distributed in different spatial locations around the target drone. These sound sensors are configured according to a predetermined spatial layout to form a three-dimensional monitoring network covering all directions of the target drone, including the front, rear, left, right, top, and bottom.

[0022] Once the drone is activated, multiple sound sensors simultaneously acquire acoustic signals, obtaining sound information from all directions in real time to form an omnidirectional sound sensing dataset. This omnidirectional sound sensing dataset contains acoustic environment information of the airspace surrounding the target drone at the current monitoring moment.

[0023] In the process of identifying the location of an intruding drone, spatial positioning is achieved by utilizing the physical characteristics of sound propagation. When an intruding drone approaches a target drone, the sound signal generated by its propellers is captured by the target drone's sound sensor array. At this point, the sound sensors closer to the intruding drone will detect a stronger sound signal intensity, while the sensors farther away will detect a relatively weaker sound signal intensity, thus forming a directional sound intensity distribution pattern in space, which helps identify the location of the intruding drone.

[0024] The target drone is the drone being controlled by the user; the intruding drone refers to other drones that enter the target drone's airspace and may pose a security threat to the target drone, and are controlled by other users. When the two drones are too close, electromagnetic interference may occur, leading to unstable communication links or failure of control command transmission, which may result in safety accidents such as loss of control and crashes.

[0025] Traditional solutions for comprehensive threat detection require the simultaneous activation of multiple high-power devices such as visual sensors, lidar, or millimeter-wave radars, all kept constantly on to ensure continuous monitoring. This operational mode leads to a sharp increase in power consumption, severely impacting the drone's endurance. In contrast, the acoustic sensor array used in this embodiment consumes extremely low power, enabling continuous operation for extended periods without significantly increasing energy consumption. Furthermore, the pre-positioning function achieved through acoustic monitoring selectively activates visual and ultrasonic sensors in the corresponding direction based on the threat's location, avoiding the resource waste of operating all sensors simultaneously in traditional solutions. This on-demand activation not only significantly reduces overall power consumption but also extends the lifespan of critical sensors.

[0026] S2. Activate the corresponding visual sensor and ultrasonic sensor according to the location of the intruding drone. Collect images of the intruding drone through the visual sensor and obtain the distance between the intruding drone and the target drone through the ultrasonic sensor.

[0027] Specifically, the target drone is equipped with multiple visual and ultrasonic sensors distributed in different locations. These sensors are installed in key positions on the front, rear, left, and right sides of the drone according to a predetermined spatial configuration. Once the acoustic sensor array determines the location of the intruding drone, it automatically selects and activates the visual and ultrasonic sensors facing that direction, while sensors in other directions remain in standby mode.

[0028] The activated visual sensor immediately aligns with the direction of the intruding drone to acquire images, obtaining image data containing the drone's external features, providing the necessary data foundation for subsequent model identification. Simultaneously, the ultrasonic sensor in the corresponding location emits ultrasonic signals and receives the reflected echoes. By calculating the round-trip time of the ultrasonic signals, the straight-line distance between the intruding and target drones is accurately measured, obtaining the real-time distance between the two drones.

[0029] Compared to traditional solutions where all sensors operate simultaneously, the directional activation strategy significantly reduces power consumption. Acoustic pre-positioning provides directional information that accurately pinpoints the threat source, avoiding unnecessary omnidirectional scanning and detection, thus optimizing sensor resource allocation. This layered detection mechanism, based on coarse acoustic localization and fine directional detection, not only improves detection efficiency but also extends sensor lifespan. Image information acquired by the visual sensor and distance data measured by the ultrasonic sensor provide reliable data support for subsequent threat assessment and collision avoidance decisions.

[0030] S3. Based on the image of the intrusion drone, identify the model of the intrusion drone, obtain the electromagnetic interference distance constraint corresponding to the model of the intrusion drone, and construct a virtual electromagnetic protection zone with the target drone as the center and in combination with the electromagnetic interference distance constraint.

[0031] Specifically, the acquired images of intruding drones are analyzed to achieve drone model identification and personalized electromagnetic protection. Specifically, the acquired images of the intruding drones are input into a pre-trained drone identifier, which is built based on deep learning algorithms and can identify the specific model of the intruding drone to obtain its model number.

[0032] After obtaining the model of the intruding drone, a pre-built electromagnetic interference (EMI) feature library is retrieved. This library contains EMI distance constraints for different drone models, obtained through actual testing, reflecting the critical values ​​at which each drone model may cause EMI to the target drone at specific distances. Based on the identified intruding drone model, the corresponding EMI distance constraints are matched and retrieved from the EMI feature library.

[0033] Based on the acquired electromagnetic interference distance constraint, a spherical virtual electromagnetic protection zone is constructed in three-dimensional space, with the target UAV's current position as the center and the electromagnetic interference distance constraint as the radius. This virtual protection zone fully considers the UAV's motion characteristics in three-dimensional space and can more comprehensively prevent electromagnetic interference threats from any direction and altitude.

[0034] Model identification enables differentiated electromagnetic interference prevention. Personalized protection distances are set for different intruding drone models, avoiding both overly conservative settings that lead to frequent false triggers and overly lenient settings that may cause electromagnetic interference risks. This model-based dynamic protection mechanism extends collision avoidance from simple physical collision prevention to electromagnetic interference prevention, intervening in security protection at an earlier stage to ensure the normal operation of drone communication and control.

[0035] S4. Determine whether the intruding drone is about to enter the virtual electromagnetic protection zone based on the distance between the two drones. When the intruding drone is about to enter the virtual electromagnetic protection zone, trigger the target drone to autonomously avoid collision, so that the target drone moves away from the intruding drone.

[0036] Specifically, based on the acquired distance between the two drones, the movement trajectory of the intruding drone is dynamically monitored and threat assessed. The current distance between the two drones is continuously compared with the radius of the constructed virtual electromagnetic protection zone. When an intruding drone is detected about to enter the virtual electromagnetic protection zone, an autonomous collision avoidance mechanism is immediately activated.

[0037] To ensure the timeliness of collision avoidance actions, a takeover response time is introduced, reflecting the time required from threat assessment to collision avoidance execution completion. The relative approach rate between the two drones is calculated based on continuously sampled distance data, and the takeover response distance is determined in conjunction with the takeover response time. The takeover response distance is then added to an electromagnetic interference distance constraint to obtain an intrusion distance threshold. When the real-time monitored distance between the two drones is less than or equal to the intrusion distance threshold, it is determined that the intruding drone is about to enter the virtual electromagnetic protection zone, and a collision avoidance response is immediately triggered.

[0038] Upon collision avoidance triggering, the target drone temporarily assumes manual control from the user, ensuring the collision avoidance maneuver is executed quickly and without human intervention. Based on the determined location of the intruding drone, the avoidance direction is determined, and the target drone is controlled to maneuver away from the intruding drone. During the collision avoidance process, the distance between the two drones is continuously monitored. When the distance between the two drones remains consistently greater than the intrusion distance threshold within a preset time period, the threat is confirmed to be neutralized, and flight control is smoothly returned to the user.

[0039] This intelligent takeover mechanism ensures flight safety while maximizing the user's control experience. It allows for a brief takeover of flight control and execution of necessary safety maneuvers without affecting normal user operation, immediately restoring user control after collision avoidance. This ensures a smooth transition and continuity of user experience. Through early intervention and precise control, effective avoidance can be achieved before electromagnetic interference occurs, fundamentally preventing communication interruptions and loss of control risks caused by electromagnetic signal conflicts, thus avoiding collisions and ensuring flight safety in environments with multiple drones coexisting.

[0040] Furthermore, omnidirectional acoustic real-time monitoring is performed using an array of sound sensors to acquire an omnidirectional sound sensing dataset, and the location of the intruding drone is identified and determined based on the omnidirectional sound sensing dataset, including:

[0041] S11. The sound sensor array includes multiple sound sensors distributed in different directions around the target UAV. Multiple sound sensors are activated simultaneously to perform real-time acoustic monitoring and obtain an omnidirectional sound sensing dataset.

[0042] S12. Obtain a benchmark omnidirectional sound sensing dataset, wherein the benchmark omnidirectional sound sensing dataset is sound sensing data collected by multiple sound sensors in an environment without external sound source interference.

[0043] S13. Compare and analyze the omnidirectional sound sensing dataset with the benchmark omnidirectional sound sensing dataset to obtain the distribution characteristics of sound intensity difference of multiple sound sensors.

[0044] S14. Determine the location of the intruding drone based on the distribution characteristics of the sound intensity difference between multiple sound sensors.

[0045] In one feasible implementation, the sound sensor array adopts a distributed layout configuration, including multiple sound sensors installed at different spatial locations such as in front, behind, to the left, to the right, above, and below the target drone. After the target drone starts, the multiple sound sensors synchronously collect acoustic signals, acquiring sound intensity data from all directions in real time, forming an omnidirectional sound sensing dataset containing complete airspace acoustic environment information at the current moment.

[0046] Next, a baseline omnidirectional sound sensor dataset is acquired. This dataset consists of sound sensor data collected by multiple sound sensors in an environment free from external sound source interference. Specifically, after the target UAV takes off and reaches a stable flight state, and in a clean environment confirming the absence of other aircraft in the surrounding airspace, the baseline data acquisition program is automatically initiated. This process lasts 30-60 seconds, acquiring and storing a baseline omnidirectional sound sensor dataset containing only the sound of the target UAV's own propellers and the current ambient background noise. Due to the physical characteristic of the target UAV's propeller noise spreading uniformly in all directions, the sound sensors in all directions acquire sound signals of essentially the same intensity, forming a stable acoustic baseline and providing an accurate reference standard for subsequent abnormal sound source detection.

[0047] Subsequently, the omnidirectional sound sensor dataset obtained from real-time monitoring was compared and analyzed one by one with a pre-established benchmark omnidirectional sound sensor dataset. When an intruding drone enters the monitoring airspace, the acoustic environment changes significantly, and each sensor needs to simultaneously collect signals from both the target drone and the intruding drone. Due to the different spatial distances between the intruding drone and each sound sensor, the signal intensity collected by each sound sensor exhibits significant spatial distribution differences. By calculating the difference between the current detection value of each sound sensor and the benchmark detection value, the sound intensity difference distribution characteristics reflecting the relative position of the intruding drone are obtained.

[0048] Next, directional analysis is performed based on the physical property that sound signal intensity is inversely proportional to distance. When an intruding drone enters the monitoring airspace, sound sensors closer to the drone detect a significantly enhanced sound signal, resulting in a larger sound intensity difference. Sensors further away show relatively smaller signal amplification and correspondingly smaller sound intensity differences, thus forming a sound intensity gradient distribution with distinct directional characteristics. By identifying the location of the sound sensor with the largest sound intensity difference and combining this with the trend of difference changes in adjacent sensors, the directional location of the intruding drone is determined.

[0049] The orientation recognition method based on the distribution characteristics of sound intensity difference makes full use of the physical propagation law of sound signals. It achieves fast and accurate orientation judgment through simple amplitude comparison, without the need for complex signal processing algorithms. It has the advantages of simple calculation, fast response and strong anti-interference ability, and provides reliable directional guidance for subsequent orientation detection and collision avoidance decision-making.

[0050] Furthermore, based on the image of the intruding drone, the model of the intruding drone is identified, and an electromagnetic interference distance constraint corresponding to the model of the intruding drone is obtained. A virtual electromagnetic protection zone is constructed centered on the target drone, in conjunction with the electromagnetic interference distance constraint, including:

[0051] S31. Input the image of the intruding drone into the drone identifier to obtain the model of the intruding drone;

[0052] S32. Retrieve the electromagnetic interference feature library, which includes multiple UAV models and multiple electromagnetic interference distance constraints, with each UAV model corresponding to one of the multiple electromagnetic interference distance constraints.

[0053] S33. Match the intrusion drone model with the multiple drone models to obtain the matching drone model, and retrieve the corresponding electromagnetic interference distance constraint to obtain the electromagnetic interference distance constraint corresponding to the intrusion drone model.

[0054] S34. Using the target UAV as the center and the electromagnetic interference distance constraint as the radius, construct a spherical virtual electromagnetic protection area in three-dimensional space.

[0055] In one feasible implementation, the acquired images of the intruding drone are first input into a pre-trained drone identifier for analysis. This drone identifier, built on a deep learning algorithm, has the ability to identify drone models. By extracting key feature information from the images, the drone identifier accurately determines the specific model of the intruding drone and outputs the intruding drone model.

[0056] Then, a pre-built electromagnetic interference (EMI) feature library is retrieved. This library is a database established through extensive experimental testing. The EMI feature library contains EMI characteristic parameters for multiple UAV models, with each UAV model corresponding to an experimentally verified EMI distance constraint. These parameters reflect the differences in technical specifications of different UAV models in terms of communication frequency bands, transmit power, antenna configuration, control protocols, etc., and the resulting electromagnetic compatibility characteristics.

[0057] Subsequently, the identified intruding drone model is compared one by one with multiple drone models stored in the electromagnetic interference feature database to find a completely matching drone model. Once the matching drone model is determined, the corresponding electromagnetic interference distance constraint is immediately retrieved based on the matching drone model to obtain the electromagnetic interference distance constraint for the current intruding drone.

[0058] Next, based on the acquired electromagnetic interference distance constraints, a virtual electromagnetic protection zone is constructed in three-dimensional space. Specifically, a spherical virtual electromagnetic protection zone is established with the target UAV's current spatial position as the center and the electromagnetic interference distance constraints as the radius. This virtual electromagnetic protection zone fully considers the UAV's maneuverability in three-dimensional space and can defend against electromagnetic interference threats from any direction and altitude.

[0059] Compared to traditional collision avoidance systems that use fixed safety distance settings, this system achieves personalized electromagnetic protection based on model characteristics. Different brands and models of drones, due to significant differences in their technical parameters, may generate varying degrees of electromagnetic interference during close-range flight. Through precise model identification and corresponding parameter retrieval, a reasonable protection distance can be set for each intruding drone. This avoids both overly conservative settings that lead to frequent false triggers and unnecessary avoidance maneuvers, and overly lenient settings that may cause electromagnetic interference risks, achieving precise and intelligent collision avoidance operation.

[0060] Furthermore, the construction steps of the drone identifier include:

[0061] S311. Establish a drone model library, which includes multiple drone models. Collect drone image sets corresponding to each drone model to obtain multiple drone image sets.

[0062] S312. Construct a sample drone image set and a sample drone model set based on the multiple drone image sets and the multiple drone models;

[0063] S313. Using the sample drone image set as input and the sample drone model set as labels, train and obtain the drone identifier.

[0064] In a preferred embodiment, firstly, a drone model database is established, which includes multiple drone models. For each drone model in the database, a multi-angle, multi-environment image acquisition strategy is employed to acquire high-quality image data of that model under different lighting conditions, shooting distances, and attitude angles. Through systematic data acquisition, a dedicated image set containing hundreds to thousands of images is established for each drone model, ensuring coverage of the model's visual performance characteristics in various actual flight scenarios, thereby obtaining multiple drone image sets corresponding to multiple drone models.

[0065] Then, multiple drone image sets from different drone models are integrated to construct a sample drone image set with a unified format. Simultaneously, a corresponding model label is created for each image in the sample drone image set, converting the drone model name into a standardized classification identifier, forming a sample drone model set that corresponds one-to-one with the image data. This image-label paired data structure provides a complete training data foundation for subsequent supervised learning.

[0066] Subsequently, a deep convolutional neural network architecture was used to train and construct the drone identifier. Using a set of sample drone images as network input and a set of sample drone models as the desired output labels, the network was trained to learn the visual feature representations of different drone models through backpropagation and gradient descent optimization methods. This established a mapping relationship from image features to model classification, resulting in the drone identifier, which possesses the ability to accurately determine the model of a new input image. This drone identifier can quickly identify the specific model of an intruding drone in complex flight environments based on only one or a small number of images, providing reliable model information support for subsequent electromagnetic interference parameter matching and protection zone construction.

[0067] Furthermore, the steps for constructing the electromagnetic interference feature library include:

[0068] S321. Determine a first drone model from the plurality of drone models, and determine a first experimental drone based on the first drone model;

[0069] S322. Control the first experimental UAV and the target UAV to conduct multiple electromagnetic interference tests. In each electromagnetic interference test, gradually reduce the distance between the two UAVs until electromagnetic interference occurs. Record the distance data when interference occurs to obtain multiple electromagnetic interference test distances.

[0070] S323. Select the maximum distance value from the plurality of electromagnetic interference test distances as the first electromagnetic interference distance constraint corresponding to the first UAV model;

[0071] S324. Associate the first UAV model with the first electromagnetic interference distance constraint and store it in the electromagnetic interference feature library;

[0072] S325. Following the method of obtaining the first electromagnetic interference distance constraint for the first UAV model, obtain the electromagnetic interference distance constraints for the other UAV models and establish the electromagnetic interference feature library.

[0073] Specifically, firstly, the first UAV model to be tested is determined from the UAV model library in a predetermined order, and a corresponding first experimental UAV is prepared according to the specifications of that model. The experimental UAV needs to ensure that its technical parameters are completely consistent with those of standard products sold on the market, including key electromagnetic characteristic parameters such as communication frequency band configuration, transmit power level, antenna type and gain, and control protocol version, to ensure the representativeness and reliability of the test results.

[0074] Then, standardized electromagnetic interference (EMI) tests were conducted in an open, interference-free experimental environment. The first experimental UAV and the target UAV were controlled to fly simultaneously from a considerable distance, maintaining relative stillness or low-speed flight, gradually reducing the spatial distance between them. During the distance reduction process, key indicators such as the stability of the communication link between the two UAVs, control command response time, and signal reception strength were continuously monitored. When communication interruption, control delay, signal attenuation, or other EMI phenomena were detected, the current precise distance data between the two UAVs was immediately recorded. Through repeated tests, multiple EMI test distances were obtained under different environmental conditions and flight attitudes.

[0075] Subsequently, statistical analysis was performed on the obtained electromagnetic interference test distances. Since electromagnetic interference can be influenced by various factors such as the environmental electromagnetic background, the UAV's attitude angle, and relative position, the interference distance may vary under different test conditions. To ensure the reliability and safety of electromagnetic protection, the maximum electromagnetic interference test distance was selected from all electromagnetic interference test distances as the first electromagnetic interference distance constraint for this first UAV model. This conservative design strategy can effectively prevent electromagnetic interference even under the most stringent conditions.

[0076] Subsequently, a mapping relationship was established between the experimentally verified first UAV model and its corresponding electromagnetic interference distance constraint, and stored in the data structure of the electromagnetic interference feature library. Following the standardized process for processing the first UAV model, the same tests and parameter acquisition operations were performed on the remaining UAV models in the library. Through systematic experimental testing, a complete electromagnetic interference feature library containing multiple UAV models and their corresponding electromagnetic interference distance constraints was gradually established. This electromagnetic interference feature library, built based on a large amount of actual test data, can provide accurate electromagnetic safety distance references for collision avoidance of different UAV models, ensuring the effectiveness and reliability of collision avoidance in various practical application scenarios.

[0077] Furthermore, based on the distance between the two drones, it is determined whether the intruding drone is about to enter the virtual electromagnetic protection zone. When the intruding drone is about to enter the virtual electromagnetic protection zone, the target drone is triggered to autonomously avoid collision, causing the target drone to move away from the intruding drone, including:

[0078] S41. Obtain the takeover response time of the target UAV, and determine the takeover response distance based on the takeover response time;

[0079] S42. Sum the interception response distance with the electromagnetic interference distance constraint corresponding to the intrusion drone model to obtain the intrusion distance threshold;

[0080] S43. When the distance between the two drones is less than or equal to the intrusion distance threshold, it is determined that the intruding drone is about to enter the virtual electromagnetic protection zone;

[0081] S44. Trigger the autonomous control system of the target drone and temporarily take over the user's manual control authority;

[0082] S45. Determine the avoidance direction based on the location of the intruding drone, and control the target drone to move away from the intruding drone;

[0083] S46. Continuously monitor the distance between the two devices. When the distance between the two devices exceeds the intrusion distance threshold within a preset time period, return control to the user.

[0084] In a preferred embodiment, firstly, the takeover response time of the target UAV is acquired, reflecting the time required from threat assessment to the complete execution of collision avoidance actions. Based on real-time monitored distance changes between the two UAVs, the relative approach rate is calculated, and this rate is multiplied by the takeover response time to obtain the takeover response distance. This takeover response distance represents the minimum spatial buffer space required to complete the collision avoidance response at the current relative approach rate between the two UAVs.

[0085] Subsequently, the determined takeover response distance and the obtained electromagnetic interference distance constraints corresponding to the intruding drone models were numerically summed to obtain an intrusion distance threshold that comprehensively considers electromagnetic safety and response delay. This intrusion distance threshold ensures sufficient electromagnetic safety distance while reserving the necessary reaction time and space for collision avoidance, thus achieving an organic combination of electromagnetic protection requirements and dynamic response capabilities.

[0086] Subsequently, the real-time distance between the two drones is continuously compared with the calculated intrusion distance threshold. When the distance between the two drones is detected to be less than or equal to the intrusion distance threshold, it is immediately determined that the intruding drone is about to enter the virtual electromagnetic protection zone, triggering a collision avoidance warning. This dynamic threshold-based judgment method can provide early warnings by considering movement trends, ensuring sufficient time to execute effective avoidance maneuvers.

[0087] Once a threat is confirmed, the target drone's autonomous control is immediately activated, temporarily taking over manual control from the user via software commands. During this takeover, a smooth transition strategy is employed to avoid any impact on flight stability from a sudden switch in control, ensuring a seamless transition from manual to autonomous control. Then, based on the determined location of the intruding drone, the optimal avoidance direction is calculated, prioritizing airspace directions that are far from the intruding drone and meet flight safety requirements. The target drone is then controlled to maneuver according to a preset avoidance strategy, including adjusting altitude, changing heading angle, or increasing lateral distance, ensuring the distance between the two drones is rapidly increased to a safe level.

[0088] Simultaneously, throughout the collision avoidance process, the real-time changes in the distance between the two aircraft are continuously monitored. When the distance between the two aircraft is detected to remain stably greater than the intrusion distance threshold within a preset time period, it is confirmed that the threat has been completely eliminated. At this point, flight control authority is automatically and smoothly returned to the user, restoring normal manual control mode, ensuring that the user can continue to carry out the flight mission according to the original plan.

[0089] This collision avoidance control mechanism ensures electromagnetic safety while maximizing the user's control experience and the continuity of flight missions. Through precise timing and smooth control transfer, it provides reliable safety protection at critical moments and immediately restores user autonomy after collision avoidance, achieving an optimal balance between safety and user experience.

[0090] Furthermore, determining the takeover response distance based on the takeover response duration includes:

[0091] S411. Calculate the relative approach rate of the two machines based on the continuously sampled distance between them;

[0092] S412. Multiply the relative approach rate of the two machines by the control response time to obtain the control response distance.

[0093] In a preferred embodiment, real-time analysis of the relative motion state of the two drones is performed based on continuously sampled distance data provided by ultrasonic sensors. By performing numerical differentiation on the distance data sampled at consecutive time points, the distance change between adjacent sampling moments is calculated and divided by the corresponding time interval to obtain the instantaneous approach rate between the two drones. A sliding window averaging method is used to smooth the instantaneous approach rate of multiple sampling points, eliminating the influence of measurement noise and random fluctuations, and obtaining a stable and reliable relative approach rate that reflects the real-time dynamic characteristics of the intruding drone approaching the target drone.

[0094] Subsequently, the obtained relative approach speed between the two aircraft is multiplied by the preset takeover response time to obtain the takeover response distance, which reflects the spatial distance required to complete the entire process from threat detection to collision avoidance. The dynamic calculation of the takeover response distance ensures that collision avoidance can adaptively adjust the triggering timing according to different relative motion states, avoiding unnecessary avoidance caused by premature triggering and preventing collision avoidance failure that may be caused by late triggering.

[0095] By employing a dynamic threshold setting method based on real-time speed calculation, the diversity and complexity of relative motion states during UAV flight are fully considered. When the relative approach speed between two aircraft is relatively fast, the takeover response distance is automatically increased to initiate collision avoidance procedures in advance; when the relative approach speed is relatively slow, the takeover response distance is correspondingly reduced to avoid frequent false triggers. Through this adaptive distance threshold adjustment mechanism, optimal response performance and collision avoidance effects can be maintained under various flight scenarios.

[0096] Example 2, as Figure 2 As shown, based on the same inventive concept as the UAV automatic collision avoidance method that integrates multi-source information provided in Embodiment 1, this embodiment of the invention also provides an UAV automatic collision avoidance system that integrates multi-source information, including:

[0097] The omnidirectional acoustic monitoring module 11 is used to perform real-time omnidirectional acoustic monitoring through an array of sound sensors, acquire an omnidirectional sound sensing dataset, and identify and determine the location of the intruding drone based on the omnidirectional sound sensing dataset.

[0098] The visual ultrasonic activation module 12 is used to activate the corresponding visual sensor and ultrasonic sensor according to the location of the intruding drone, acquire images of the intruding drone through the visual sensor, and obtain the distance between the intruding drone and the target drone through the ultrasonic sensor.

[0099] The electromagnetic protection zone construction module 13 is used to identify the model of the intruding drone based on the image of the intruding drone, obtain the electromagnetic interference distance constraint corresponding to the model of the intruding drone, and construct a virtual electromagnetic protection zone with the target drone as the center and in combination with the electromagnetic interference distance constraint.

[0100] The collision avoidance decision execution module 14 is used to determine whether the intruding drone is about to enter the virtual electromagnetic protection zone based on the distance between the two drones. When the intruding drone is about to enter the virtual electromagnetic protection zone, the target drone is triggered to autonomously avoid collision, so that the target drone moves away from the intruding drone.

[0101] Furthermore, the execution steps of the omnidirectional acoustic monitoring module 11 include:

[0102] The sound sensor array includes multiple sound sensors distributed in different directions around the target drone. Multiple sound sensors are activated simultaneously to perform real-time acoustic monitoring and acquire an omnidirectional sound sensing dataset.

[0103] Obtain a benchmark omnidirectional sound sensing dataset, which is sound sensing data collected by multiple sound sensors in an environment without external sound source interference;

[0104] The omnidirectional sound sensing dataset is compared and analyzed with the benchmark omnidirectional sound sensing dataset to obtain the distribution characteristics of sound intensity difference from multiple sound sensors.

[0105] The location of the intruding drone was determined based on the distribution characteristics of the sound intensity difference from multiple sound sensors.

[0106] Furthermore, the execution steps of the electromagnetic protection zone construction module 13 include:

[0107] The image of the intruding drone is input into the drone identifier to obtain the model of the intruding drone;

[0108] Retrieve the electromagnetic interference feature library, which includes multiple UAV models and multiple electromagnetic interference distance constraints, with each UAV model corresponding to one of the multiple electromagnetic interference distance constraints.

[0109] The intrusion drone model is matched with the multiple drone models to obtain the matching drone model, and the corresponding electromagnetic interference distance constraint is retrieved to obtain the electromagnetic interference distance constraint corresponding to the intrusion drone model.

[0110] Centered on the target UAV, and with the electromagnetic interference distance constraint as the radius, a spherical virtual electromagnetic protection zone is constructed in three-dimensional space.

[0111] Furthermore, the construction steps of the drone identifier include:

[0112] Establish a drone model library, which includes multiple drone models, and collect drone image sets corresponding to each drone model to obtain multiple drone image sets;

[0113] Based on the multiple drone image sets and the multiple drone models, construct a sample drone image set and a sample drone model set;

[0114] Using the sample drone image set as input and the sample drone model set as labels, the drone identifier is trained and obtained.

[0115] Furthermore, the steps for constructing the electromagnetic interference feature library include:

[0116] A first drone model is determined from the plurality of drone models, and a first experimental drone is determined based on the first drone model;

[0117] The first experimental UAV and the target UAV were controlled to conduct multiple electromagnetic interference tests. In each electromagnetic interference test, the distance between the two UAVs was gradually reduced until electromagnetic interference occurred. The distance data when interference occurred was recorded to obtain multiple electromagnetic interference test distances.

[0118] The maximum distance value is selected from the plurality of electromagnetic interference test distances and used as the first electromagnetic interference distance constraint corresponding to the first UAV model.

[0119] The first UAV model is associated with the first electromagnetic interference distance constraint and stored in the electromagnetic interference feature library;

[0120] Following the method of obtaining the first electromagnetic interference distance constraint for the first UAV model, the electromagnetic interference distance constraints for the other UAV models are obtained, and the electromagnetic interference feature library is established.

[0121] Furthermore, the execution steps of the collision avoidance decision execution module 14 include:

[0122] Obtain the takeover response time of the target UAV, and determine the takeover response distance based on the takeover response time;

[0123] The intrusion distance threshold is obtained by summing the interception response distance with the electromagnetic interference distance constraint corresponding to the intrusion drone model.

[0124] When the distance between the two drones is less than or equal to the intrusion distance threshold, it is determined that the intruding drone is about to enter the virtual electromagnetic protection zone;

[0125] Trigger the autonomous control system of the target drone and temporarily take over the user's manual control authority;

[0126] Determine the avoidance direction based on the location of the intruding drone, and control the target drone to move away from the intruding drone;

[0127] The system continuously monitors the distance between the two devices. When the distance between the two devices exceeds the intrusion distance threshold within a preset time period, control permissions are returned to the user.

[0128] Furthermore, the execution steps of the collision avoidance decision execution module 14 also include:

[0129] The relative approach rate of the two machines is calculated based on the continuously sampled distance between them.

[0130] The relative approach rate of the two machines is multiplied by the control response time to obtain the control response distance.

[0131] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0132] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0133] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0134] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0135] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0136] Although preferred embodiments of the invention have been described, those skilled in the art, once they have learned the basic inventive concept, can make other changes and modifications to these embodiments.

[0137] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of this invention and its equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for automatic collision avoidance of unmanned aerial vehicles (UAVs) that integrates multi-source information, characterized in that, The method includes: Omnidirectional acoustic real-time monitoring is performed using an array of sound sensors to acquire an omnidirectional sound sensing dataset, and the location of the intruding drone is identified and determined based on the omnidirectional sound sensing dataset. Based on the location of the intruding drone, the corresponding visual and ultrasonic sensors are activated. The visual sensors acquire images of the intruding drone, and the ultrasonic sensors obtain the distance between the intruding drone and the target drone. Based on the image of the intrusion drone, the model of the intrusion drone is identified, and the electromagnetic interference distance constraint corresponding to the model of the intrusion drone is obtained. A virtual electromagnetic protection zone is constructed with the target drone as the center and the electromagnetic interference distance constraint. Based on the distance between the two drones, it is determined whether the intruding drone is about to enter the virtual electromagnetic protection zone. When the intruding drone is about to enter the virtual electromagnetic protection zone, the target drone is triggered to autonomously avoid collision, causing the target drone to move away from the intruding drone, including: Obtain the takeover response time of the target UAV, and determine the takeover response distance based on the takeover response time; The intrusion distance threshold is obtained by summing the interception response distance with the electromagnetic interference distance constraint corresponding to the intrusion drone model. When the distance between the two drones is less than or equal to the intrusion distance threshold, it is determined that the intruding drone is about to enter the virtual electromagnetic protection zone; Trigger the autonomous control system of the target drone and temporarily take over the user's manual control authority; Determine the avoidance direction based on the location of the intruding drone, and control the target drone to move away from the intruding drone; The system continuously monitors the distance between the two devices. When the distance between the two devices exceeds the intrusion distance threshold within a preset time period, control permissions are returned to the user.

2. The method according to claim 1, characterized in that, Omnidirectional acoustic real-time monitoring is performed using an array of sound sensors to acquire an omnidirectional sound sensing dataset, and the location of the intruding drone is identified and determined based on the omnidirectional sound sensing dataset, including: The sound sensor array includes multiple sound sensors distributed in different directions around the target drone. Multiple sound sensors are activated simultaneously to perform real-time acoustic monitoring and acquire an omnidirectional sound sensing dataset. Obtain a benchmark omnidirectional sound sensing dataset, which is sound sensing data collected by multiple sound sensors in an environment without external sound source interference; The omnidirectional sound sensing dataset is compared and analyzed with the benchmark omnidirectional sound sensing dataset to obtain the distribution characteristics of sound intensity difference from multiple sound sensors. The location of the intruding drone was determined based on the distribution characteristics of the sound intensity difference from multiple sound sensors.

3. The method according to claim 1, characterized in that, Based on the image of the intrusion drone, the model of the intrusion drone is identified, and an electromagnetic interference distance constraint corresponding to the model of the intrusion drone is obtained. A virtual electromagnetic protection zone is constructed centered on the target drone and in conjunction with the electromagnetic interference distance constraint, including: The image of the intruding drone is input into the drone identifier to obtain the model of the intruding drone; Retrieve the electromagnetic interference feature library, which includes multiple UAV models and multiple electromagnetic interference distance constraints, with each UAV model corresponding to one of the multiple electromagnetic interference distance constraints. The intrusion drone model is matched with the multiple drone models to obtain the matching drone model, and the corresponding electromagnetic interference distance constraint is retrieved to obtain the electromagnetic interference distance constraint corresponding to the intrusion drone model. Centered on the target UAV, and with the electromagnetic interference distance constraint as the radius, a spherical virtual electromagnetic protection zone is constructed in three-dimensional space.

4. The method according to claim 3, characterized in that, The construction steps of the drone identifier include: Establish a drone model library, which includes multiple drone models, and collect drone image sets corresponding to each drone model to obtain multiple drone image sets; Based on the multiple drone image sets and the multiple drone models, construct a sample drone image set and a sample drone model set; Using the sample drone image set as input and the sample drone model set as labels, the drone identifier is trained and obtained.

5. The method according to claim 3, characterized in that, The steps for constructing the electromagnetic interference feature library include: A first drone model is determined from the plurality of drone models, and a first experimental drone is determined based on the first drone model; The first experimental UAV and the target UAV were controlled to conduct multiple electromagnetic interference tests. In each electromagnetic interference test, the distance between the two UAVs was gradually reduced until electromagnetic interference occurred. The distance data when interference occurred was recorded to obtain multiple electromagnetic interference test distances. The maximum distance value is selected from the plurality of electromagnetic interference test distances and used as the first electromagnetic interference distance constraint corresponding to the first UAV model. The first UAV model is associated with the first electromagnetic interference distance constraint and stored in the electromagnetic interference feature library; Following the method of obtaining the first electromagnetic interference distance constraint for the first UAV model, the electromagnetic interference distance constraints for the other UAV models are obtained, and the electromagnetic interference feature library is established.

6. The method according to claim 1, characterized in that, Determining the takeover response distance based on the takeover response duration includes: The relative approach rate of the two machines is calculated based on the continuously sampled distance between them. The relative approach rate of the two machines is multiplied by the control response time to obtain the control response distance.

7. An automatic collision avoidance system for unmanned aerial vehicles (UAVs) that integrates multi-source information, characterized in that, The system for implementing the method as described in any one of claims 1 to 6 includes: The omnidirectional acoustic monitoring module is used to perform real-time omnidirectional acoustic monitoring through an array of sound sensors, acquire an omnidirectional sound sensing dataset, and identify and determine the location of the intruding drone based on the omnidirectional sound sensing dataset. The visual ultrasonic activation module is used to activate the corresponding visual sensor and ultrasonic sensor according to the location of the intruding drone. The visual sensor acquires images of the intruding drone, and the ultrasonic sensor obtains the distance between the intruding drone and the target drone. The electromagnetic protection zone construction module is used to identify the model of the intruding drone based on the image of the intruding drone, obtain the electromagnetic interference distance constraint corresponding to the model of the intruding drone, and construct a virtual electromagnetic protection zone with the target drone as the center and in combination with the electromagnetic interference distance constraint; the collision avoidance decision execution module is used to determine whether the intruding drone is about to enter the virtual electromagnetic protection zone based on the distance between the two drones, and when the intruding drone is about to enter the virtual electromagnetic protection zone, trigger the target drone to autonomously avoid collision, so that the target drone moves away from the intruding drone.

Images