Self-adaptive anti-interference unmanned aerial vehicle distribution system and method for denial environment

Abstract

The invention discloses a denial environment-oriented self-adaptive anti-interference unmanned aerial vehicle distribution system and a denial environment-oriented self-adaptive anti-interference unmanned aerial vehicle distribution method. The system integrates a navigation module, a communication module, a path planning module and a safety redundancy module, high-precision positioning is achieved through multi-source sensor fusion, and the stability of a communication link is ensured by combining frequency hopping spread spectrum communication, laser communication and an ad hoc network technology. The path planning module dynamically generates an obstacle avoidance path by using a three-dimensional threat map and a deep reinforcement learning algorithm, and switches to an offline mode when communication is interrupted. A safety redundancy mechanism guarantees the fault tolerance of the system through dual-system hot backup, an emergency parachute and a data self-destruction function. According to the invention, the anti-interference capability and the distribution efficiency of the unmanned aerial vehicle in a complex denial environment are effectively improved, and a solution is provided for the practical application of the unmanned aerial vehicle distribution system.

Description

Technical Field

[0001] This application relates to the field of aircraft anti-interference technology, specifically to a system and method for adaptive delivery in a denied environment. Background Art

[0002] Drone delivery technology currently demonstrates significant advantages in logistics, emergency rescue, and the transportation of specialized materials. However, existing systems still face multiple technical bottlenecks in complex, denied environments. Specifically, drones in these environments must achieve precise positioning without satellite navigation, maintain reliable communications amidst strong interference, dynamically evade real-time threats, and ensure high system fault tolerance. Traditional technical solutions suffer from significant shortcomings in these dimensions.

[0003] The defects of the existing technology are analyzed as follows:

[0004] 1. Navigation and positioning rely on a single sensor and lack anti-interference capabilities: Existing drone systems often use inertial navigation units for dead reckoning under GNSS denial conditions, but IMUs have cumulative errors and lack multi-source sensor fusion, geomagnetic sensors, and acoustic receivers to collect heterogeneous data, and use extended Kalman filtering and particle filtering algorithms for dynamic weighted calculations to output high-precision posture information. The mechanism causes positioning accuracy to decrease significantly over time. For example, the positioning error of a single IMU within 10 seconds may exceed 10 meters, which cannot meet the needs of high-precision delivery. In addition, the existing navigation system has not effectively integrated multimodal data such as visual SLAM, geomagnetic matching, and acoustic beacons, making it difficult to achieve robust positioning in complex electromagnetic environments;

[0005] 2. Rigid communication links and poor dynamic adaptability: Traditional drone communication modules mostly use fixed-frequency radio transmission, which is prone to link interruption in areas of electromagnetic interference or communication silence. Existing frequency-hopping spread spectrum systems lack real-time spectrum sensing capabilities and are unable to dynamically select low-interference channels. Laser communication and ad hoc network relay technologies do not achieve seamless switching, resulting in insufficient coverage in communication blind spots. For example, when interference intensity exceeds -80dBm, the existing communication system's switching delay can reach hundreds of milliseconds, seriously affecting the real-time performance of control commands.

[0006] 3. Static path planning and a lack of intelligent obstacle avoidance capabilities: Current drone path planning is mostly based on preset routes or simple obstacle avoidance algorithms, and is unable to integrate multi-dimensional data such as electromagnetic interference sources, terrain obstacles, and weather threats in real time to generate dynamic obstacle avoidance strategies. For example, existing systems typically perform hovering or return operations after communication interruptions, lacking the ability to replan local paths based on offline decision-making, significantly increasing the risk of mission interruption.

[0007] 4. Weak redundancy mechanisms and insufficient system fault tolerance: Existing drone systems often utilize a single master control architecture, resulting in failover times exceeding seconds and a lack of emergency landing and data protection mechanisms. For example, if the master control system fails, the backup system takes several seconds to restart, failing to meet millisecond-level high reliability requirements. Furthermore, data storage units lack an encrypted self-destruct function, making it vulnerable to sensitive information leakage in the event of a communications hijack.

[0008] In summary, existing drone delivery systems in denied environments face core challenges such as low positioning accuracy, poor communication robustness, rigid path planning, and insufficient safety redundancy. A comprehensive solution with adaptive anti-interference capabilities is urgently needed. This invention addresses these technical challenges and, through multi-source sensor fusion positioning, dynamic communication switching, intelligent path planning, and a multi-level safety redundancy mechanism, enables reliable drone operation and efficient delivery in complex denied environments. Summary of the Invention

[0009] In response to the above technical bottlenecks, the present invention provides an adaptive anti-interference UAV delivery system and method for denied environments.

[0010] According to a first aspect of the present application, an adaptive anti-interference UAV delivery system for denied environments is proposed, comprising: a navigation module for achieving anti-interference positioning through multi-source sensor fusion under GNSS signal denial conditions and outputting six-degree-of-freedom pose data of the UAV; a communication module for dynamically switching communication modes to maintain the communication link under electromagnetic interference or communication interruption conditions, wherein the communication modes include frequency hopping spread spectrum communication, laser communication, and ad hoc network relay transmission communication based on the OLSR protocol; a path planning module for integrating electromagnetic interference location data, terrain elevation data, and meteorological data to construct a three-dimensional threat map, and generating an obstacle avoidance path through a deep Q-network algorithm, and switching to an offline replanning mode to perform local path replanning when the communication interruption exceeds 10 seconds; a safety redundancy module for achieving fault switching within 10ms through dual-system hot backup, triggering an emergency parachute based on crash risk determination of vibration amplitude and attitude angle deviation, and activating a data self-destruct mechanism in the event of communication hijacking and controlling the UAV to land at a preset safe coordinate. The system achieves anti-interference positioning, dynamic communication switching, intelligent path planning, and adaptive adjustment of multi-level safety response through the collaborative work of various modules.

[0011] Furthermore, the navigation module includes: a multi-source sensor, which integrates an inertial measurement unit (100Hz output angular velocity / acceleration data), a binocular vision camera (performing ORB feature matching), a solid-state lidar (10cm precision three-dimensional point cloud), a three-axis geomagnetic sensor (geomagnetic fingerprint matching error <2m) and an ultrasonic receiver (40kHz acoustic beacon TDOA positioning), which is used to collect environmental data around the drone; a data fusion unit, which uses a preset extended Kalman filter and particle filter hybrid algorithm to dynamically weighted fuse the environmental data generated by the multi-source sensors, and output six-degree-of-freedom posture data with a position error of <3m and an attitude angle error of <1°; a navigation controller, which generates flight attitude correction instructions for the drone based on the six-degree-of-freedom posture data.

[0012] Furthermore, the communication module includes: a frequency hopping spread spectrum transceiver, which is used to select channels with interference strength less than -90dBm through real-time spectrum sensing (FFT1024 point scanning), and perform channel hopping at intervals of 1ms in the 2.4GHz frequency band; a laser communication unit, which is used to establish a 1Gbps directional optical link in a radio silence area using a 1550nm wavelength laser (beam divergence angle less than 2mrad); a self-organizing network protocol stack, which is used to establish a multi-hop Mesh network based on the OLSR protocol, maintain the topology through HELLO messages (200ms period) and TC messages (500ms period), so that the end-to-end delay is less than 500ms, and enable multiple drones to form a dynamic relay network to bypass communication blind spots.

[0013] Furthermore, the path planning module includes: a three-dimensional threat map construction unit, which is used to integrate electromagnetic interference source location data (±10m accuracy), 5m resolution terrain elevation data and real-time wind speed and precipitation meteorological information to generate a 64×64×8 rasterized three-dimensional threat map; a reinforcement learning decision engine unit, which is used to generate an obstacle avoidance strategy based on the three-dimensional threat map through a deep Q network algorithm, and perform offline path tracking by fusing the visual odometry with the A* algorithm when communication is interrupted.

[0014] Furthermore, the safety redundancy module includes: a dual-system hot backup architecture unit, which adopts a master-slave FPGA architecture and realizes fault detection through a heartbeat packet mechanism (10ms period), and the backup system takes over control within 10ms; an emergency parachute trigger unit, which triggers the gunpowder ejection mechanism to deploy the parachute when the vibration amplitude is greater than 15g (duration greater than 50ms) and the pitch angle deviation is greater than 30°; a data self-destruction unit, which performs three-level data erasure (all 0→all 1→pseudo-random number) and controls the drone to land at an encrypted safety coordinate (error ±5m) when a communication hijacking failure is detected (including SSL certificate forgery attacks).

[0015] According to a second aspect, the present application provides an adaptive anti-interference drone delivery method, characterized by comprising:

[0016] The denial environment determination step includes detecting whether the current environment of the UAV belongs to at least one of the denial types of GNSS signal denial, electromagnetic interference or communication interruption;

[0017] Multi-source sensor data collection and fusion positioning steps, including anti-interference positioning using multi-source sensors through the navigation module when GNSS signal denial is detected;

[0018] The steps of dynamically switching the anti-interference communication link and relay networking include dynamically switching at least one of frequency hopping spread spectrum communication, laser communication, and ad hoc network relay transmission communication through the communication module to maintain the communication link when electromagnetic interference or communication interruption is detected;

[0019] Dynamic threat map construction and reinforcement learning path decision-making steps, including generating obstacle avoidance paths based on the three-dimensional threat map and reinforcement learning algorithms through the path planning module, and executing offline decisions when communication interruptions exceed a set threshold;

[0020] Multi-level safety redundancy mechanism triggering and emergency response steps include detecting main control system failure, flight attitude failure or communication hijacking failure through the safety redundancy module, triggering the corresponding fault-tolerant response in the dual-system hot backup, emergency parachute or data self-destruction mechanism.

[0021] Furthermore, in the multi-source sensor data collection and fusion positioning step, it specifically includes: collecting angular velocity and acceleration data at a frequency of 100Hz through an inertial measurement unit, performing ORB feature matching to generate an environmental point cloud through a binocular vision camera, generating a three-dimensional point cloud map with a precision of 10cm through a solid-state lidar, performing geomagnetic fingerprint matching through a three-axis geomagnetic sensor (matching error <2m), and analyzing the arrival time difference of a 40kHz acoustic beacon through an ultrasonic receiver; using a preset extended Kalman filter and particle filter hybrid algorithm to dynamically weighted fuse environmental data (angular velocity and acceleration data, environmental point cloud, three-dimensional point cloud map, geomagnetic fingerprint, time difference) to generate six-degree-of-freedom posture data (position error <3m); and generating a flight attitude correction instruction for the drone based on the six-degree-of-freedom posture data.

[0022] Furthermore, when electromagnetic interference or communication interruption occurs, the communication mode is dynamically switched to maintain the communication link, specifically including: when switching to frequency hopping spread spectrum communication, selecting an undisturbed communication frequency band through real-time spectrum sensing, including performing channel hopping in the 2.4GHz frequency band; when switching to laser communication, establishing a directional optical link in the radio silence area, including activating the 1550nm laser communication link; when switching to self-organizing network relay transmission communication, multiple drones are formed into a dynamic relay network to bypass communication blind spots, including establishing a multi-hop mesh network based on the OLSR protocol to relay transmission control instructions and telemetry data.

[0023] Furthermore, in the steps of dynamic switching and relay networking of the anti-interference communication link, it specifically includes: when switching to frequency hopping spread spectrum communication, scanning the 0.1-6GHz frequency band through real-time spectrum sensing and selecting a 2.4GHz channel with an interference intensity of less than -90dBm, and performing channel hopping at intervals of 1ms according to a pseudo-random sequence; when switching to laser communication, activating a 1550nm wavelength laser communication link and establishing directional optical transmission with a beam divergence angle of less than 2mrad (rate 1Gbps); when switching to self-organizing network relay transmission communication, sending a HELLO message with a period of 200ms based on the OLSR protocol, and forming a multi-hop Mesh network with an end-to-end delay of less than 500ms, so that multiple drones form a dynamic relay network to bypass communication blind spots.

[0024] Furthermore, the dynamic threat map construction and reinforcement learning path decision steps specifically include: integrating electromagnetic interference source location data (±10m accuracy), 5m resolution terrain elevation data, and real-time wind speed and precipitation meteorological data to construct a 64×64×8 rasterized three-dimensional threat map; based on the three-dimensional threat map, generating an obstacle avoidance strategy through a deep Q network algorithm, the obstacle avoidance strategy includes adjusting the heading angle, flight altitude, and speed of the UAV; switching to offline replanning mode when the communication interruption exceeds 10 seconds, and performing local path replanning with a period of 1 second by fusing the visual odometry with the A* algorithm.

[0025] Furthermore, in the multi-level safety redundancy mechanism triggering and emergency response steps, it specifically includes: obtaining the failure factors of the drone, which include main control system failure, flight attitude failure and communication hijacking failure; when the main control system fails, the heartbeat packet detection mechanism (cycle 10ms) of the master-slave FPGA architecture is used to enable the backup system to take over control within 10ms; when a flight attitude failure occurs, when the vibration amplitude is greater than 15g and the duration is greater than 50ms, and the pitch angle deviation is greater than 30°, the gunpowder ejection mechanism is triggered to deploy the emergency parachute; when a communication hijacking failure occurs, the three-level data erasure process of the AES-256 encrypted storage medium is started (all 0 coverage → all 1 coverage → pseudo-random number coverage), and the drone is controlled to land at the preset safety coordinates (positioning error ±5m).

[0026] The adaptive anti-interference UAV delivery system and method for denied environments provided by the present invention has the following significant advantages and beneficial effects compared to the existing technology:

[0027] 1. High-precision, anti-interference positioning capability: By integrating data from a multi-source sensor, including an inertial measurement unit (IMU), a binocular vision camera, a solid-state lidar, a three-axis geomagnetic sensor, and an ultrasonic receiver, and adopting a hybrid algorithm of extended Kalman filtering and particle filtering for dynamic weighted fusion, the system achieves high-precision positioning under GNSS signal denial conditions, with a positioning error of less than 3 meters. This solves the technical problems of large cumulative errors and poor anti-interference capability of traditional single-sensor positioning.

[0028] 2. Dynamic and stable communication link: The communication module uses a frequency-hopping spread spectrum transceiver to perceive the spectrum environment in real time, dynamically selects channels with interference strength below -90dBm, and performs channel hopping at 1ms intervals. In radio-quiet areas, it switches to a 1550nm laser communication link to ensure the reliability of directional transmission. At the same time, self-organizing network technology based on the OLSR protocol realizes multi-hop relay transmission communication, with end-to-end delay less than 500ms, significantly improving the robustness of drone communications in complex electromagnetic environments.

[0029] 3. Intelligent dynamic path planning: The path planning module integrates electromagnetic interference sources, terrain elevation, and meteorological data through a three-dimensional threat map, and generates an optimal obstacle avoidance strategy based on a deep Q-network algorithm, enabling dynamic adjustment of heading angle, flight altitude, and speed. In the event of a communication interruption, the module automatically switches to offline mode and performs local path replanning to ensure mission continuity and avoid hovering or return issues caused by communication interruptions in traditional systems.

[0030] 4. Multi-level safety redundancy: The safety redundancy module adopts a dual-system hot backup architecture, with a master-slave switching time of less than 10ms, ensuring high system reliability. Vibration sensors and attitude prediction models monitor crash risks in real time. When the risk index exceeds the threshold of 0.85, the emergency parachute is automatically ejected. When the data self-destruct unit detects a communication hijacking attack, it performs a three-level overwrite erase and controls the drone to land at the preset safe coordinates, effectively preventing the leakage of sensitive information.

[0031] 5. Efficient delivery mission adaptability: This invention ensures the UAV's mission execution capability in a denied environment through offline data preloading and data retransmission after network recovery. At the same time, the reinforcement learning-based path planning algorithm can dynamically optimize flight strategies based on real-time threat maps, significantly improving delivery efficiency and mission success rate.

[0032] In summary, the present invention effectively solves the technical problems faced by drones in denied environments, such as low positioning accuracy, communication interruption, rigid path planning, and insufficient system fault tolerance, through the synergistic effect of multi-source sensors, dynamic communication switching, intelligent path planning, and multi-level safety redundancy mechanisms. It provides a highly reliable and adaptable solution for drone delivery in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This application presents an architecture diagram of an adaptive anti-interference drone delivery system for denied environments;

[0034] Figure 2 This is a schematic diagram of a navigation module in one embodiment of the present application;

[0035] Figure 3 This application presents a flow chart of an adaptive anti-interference UAV delivery method for a denied environment;

[0036] Figure 4 This is a flowchart of the multi-source sensor data collection and fusion positioning steps in one embodiment of the present application;

[0037] Figure 5 This is a flowchart of the steps of dynamic switching of anti-interference communication links and relay networking in one embodiment of the present application;

[0038] Figure 6 This is a flowchart of the steps of dynamic threat map construction and reinforcement learning path decision-making in one embodiment of the present application;

[0039] Figure 7 This is a flowchart of the multi-level safety redundancy mechanism triggering and emergency response steps in one embodiment of the present application;

[0040] Figure 8 This is a flowchart of the steps of offline preloading of task data and network disconnection recovery processing in one embodiment of the present application;

[0041] Figure 9 This is a general schematic diagram of the operation of an adaptive anti-interference drone delivery system and method for a denied environment according to an embodiment of the present application. DETAILED DESCRIPTION

[0042] The technical solution of the present invention is described in detail below with reference to the accompanying drawings and embodiments, but the protection scope of the present invention is not limited to the embodiments.

[0043] Example 1: System Architecture and Module Implementation

[0044] This embodiment combines the attached Figure 1The system architecture diagram shown in Figure 1 details the specific implementation of the adaptive, anti-interference drone delivery system of the present invention. The system comprises a navigation module 101, a communication module 102, a path planning module 103, and a safety redundancy module 104. These modules achieve high-reliability delivery in denied environments through the deep integration of hardware collaboration and software logic. The implementation details of each module are described below.

[0045] The navigation module 101 is used to achieve anti-interference positioning through multi-source sensor fusion under GNSS signal denial conditions and output the six-degree-of-freedom posture data of the UAV. Figure 2 As shown, this module's hardware implementation consists of a multi-source sensor 21, a data fusion unit 22, and a navigation controller 23. The multi-source sensor 21 integrates an ADIS16470 inertial measurement unit 24, which outputs raw angular velocity and acceleration data at a 100Hz frequency. A binocular vision camera 25 extracts environmental point clouds using the ORB feature matching algorithm. A RoboSense M1 solid-state lidar 26 generates a 3D point cloud map with 10cm accuracy. An HMC5883L three-axis geomagnetic sensor 27 performs least squares matching with a pre-stored geomagnetic fingerprint database to achieve a positioning error of less than 2m. A 40kHz ultrasonic receiver 28 analyzes the time difference of arrival (TDOA) of acoustic beacons. The data fusion unit 22 uses a Xilinx Zynq-7000 FPGA to implement a hybrid extended Kalman filter and particle filter algorithm. The extended Kalman filter layer performs dead reckoning on the IMU data, while the particle filter layer updates particle weights based on the visual SLAM pose and geomagnetic matching results. After dynamic weighted fusion, the resulting data outputs six-degree-of-freedom pose data with a position error of less than 3 meters. The navigation controller 23 runs the PID control algorithm based on the STM32H743 microprocessor, generates the flight attitude correction instructions of the UAV according to the fused posture data, drives the brushless motor through the PWM signal to adjust the rotor speed, and realizes anti-interference positioning and tracking.

[0046] The communication module 102 is used to dynamically switch the communication mode to maintain the communication link under electromagnetic interference or communication interruption conditions, including frequency hopping spread spectrum communication, laser communication and self-organizing network relay transmission communication based on OLSR protocol. Figure 1The communication module 102 includes a frequency hopping spread spectrum transceiver 105 (also referred to as an FHSS transceiver 105), a laser communication unit 106, and an ad hoc network protocol stack 107; wherein, the frequency hopping spread spectrum transceiver 105 selects a channel with an interference strength of less than -90dBm through real-time spectrum sensing, and performs channel hopping at intervals of 1ms in the 2.4GHz frequency band. Specifically, the AD9361 RF front-end chip can be used to scan the 0.1-6GHz frequency band in real time through a 1024-point FFT algorithm, dynamically select a 2.4GHz channel with an interference strength less than -90dBm, perform channel hopping at intervals of 1ms according to a pseudo-random sequence, and enable a 500μs emergency frequency hopping mode when pulse interference is detected. The laser communication unit 106 uses a 1550nm wavelength laser to establish a 1Gbps directional optical link in a radio-quiet area. Specifically, it is equipped with a 1550nm wavelength DFB laser diode. The beam divergence angle is controlled within 2mrad by a collimating lens. The transmission direction is adjusted by a servo motor with a step angle of 0.01° to establish a directional optical link with the ground relay station. The OOK modulation method is used to achieve a 1Gbps data transmission rate. The forward error correction code is RS (255, 239) to ensure a bit error rate of less than 1×10 -6 . The self-organizing network protocol stack 107 establishes a multi-hop Mesh network based on the OLSR protocol, maintains the topology through HELLO messages and TC messages, makes the end-to-end delay less than 500ms, and enables multiple drones to form a dynamic relay network to bypass communication blind spots. Specifically, based on the OLSR routing protocol, a HELLO message with a period of 200ms is defined for neighbor discovery, and a TC message with a period of 500ms is used to broadcast topology information. The shortest path is calculated through the Dijkstra algorithm and the maximum number of hops is limited to 5. When the link is interrupted, the routing table is reconstructed within 50ms, maintaining the end-to-end delay below 500ms.

[0047] The path planning module 103 is used to integrate electromagnetic interference intensity, terrain elevation and meteorological data to build a three-dimensional threat map, and generate obstacle avoidance paths through the deep Q network algorithm. When the communication interruption exceeds 10 seconds, it switches to the offline replanning mode to perform local path replanning. Figure 1The path planning module 103 includes a three-dimensional threat map construction unit 108 and a reinforcement learning decision engine 109. The three-dimensional threat map construction unit 108 is used to integrate electromagnetic interference source location data, 5m resolution terrain elevation data, and real-time wind speed and precipitation meteorological information to analyze the distribution of dynamic obstacles and generate a 64×64×8 rasterized three-dimensional threat map. Specifically, the three-dimensional threat map is mapped to a 64×64×8 three-dimensional grid space by integrating electromagnetic interference source location data (accuracy ±10 meters) collected by an electromagnetic spectrum analyzer, terrain elevation data of a 5m resolution digital elevation model, and wind speed and precipitation meteorological data provided by a meteorological radar. Each grid has a size of 78m×78m×25m and contains interference intensity, obstacle height, and wind speed level parameters. The reinforcement learning decision engine 109 is used to encode the three-dimensional threat map into a tensor input deep Q network, output the obstacle avoidance strategy of heading angle, flight altitude and speed, and perform offline path tracking by integrating the visual odometry with the A* algorithm when the communication is interrupted. Specifically, the deep Q network algorithm is adopted. The input layer receives the rasterized encoding of the threat map and the current state vector of the drone. The output layer generates an obstacle avoidance strategy including heading angle, flight altitude and speed. The network structure includes three fully connected hidden layers (256→128→64 nodes). When the communication link is interrupted for more than 10 seconds, the system automatically switches to the offline decision mode and performs local path replanning with a period of 1 second by integrating the visual odometry with the A* algorithm. The path length deviation is controlled within 5%.

[0048] The safety redundancy module 104 is used to achieve fault switching within 10ms through dual system hot backup, trigger the emergency parachute based on the crash risk determination of vibration amplitude and attitude angle deviation, and activate the data self-destruct mechanism and control the drone to land at the preset safe coordinates in the event of communication hijacking. Figure 1The safety redundancy module 104 includes a dual-system hot backup architecture unit 110, a data self-destruction unit 111, and an emergency parachute trigger unit 112. The dual-system hot backup architecture unit 110 adopts a master-slave FPGA architecture, implements fault detection through a heartbeat packet mechanism, and uses the backup system to take over control within 10ms. Specifically, a master-slave Xilinx Artix-7 FPGA design is adopted. The main control system sends a heartbeat packet containing a CRC32 check code every 10ms. When the backup system detects three consecutive heartbeat losses, it takes over control within 10ms through a CPLD hardware switching circuit. The double buffering mechanism ensures uninterrupted command transmission. The emergency parachute trigger unit 112 is used to trigger the gunpowder ejection mechanism to deploy the parachute when it detects a vibration amplitude greater than 15g and a pitch angle deviation greater than 30°. Specifically, it integrates an ADXL377 three-axis vibration sensor to monitor acceleration amplitude in real time. When any axial vibration exceeds 15g and lasts longer than 50 milliseconds, and the pitch angle deviation predicted by the Kalman filter exceeds 30 degrees, the gunpowder ejection mechanism is triggered to deploy the emergency parachute, with an ejection response time of less than 200 milliseconds. The data self-destruct unit 111 is used to perform a three-level data erase and control the drone to land to encrypted secure coordinates when a communication hijacking failure is detected. Specifically, when an SSL certificate forgery attack is detected, the AES-256 encrypted eMMC5.1 storage medium is overwritten with a three-level erase (all 0s → all 1s → pseudo-random numbers). After the erase is complete, the drone is controlled to disable the communication module and rely on the navigation module to guide it to the preset safe coordinates for landing, with a positioning error controlled within ±5 meters.

[0049] Accordingly, this application also proposes a method for adaptive anti-interference drone delivery in a denied environment.

[0050] like Figure 3 As shown, it includes the following steps S31-S35. It can be understood that the method achieves reliable drone delivery in a denied environment through the coordinated execution of the denial factor determination step, the multi-source sensor anti-interference positioning step, the dynamic communication switching step, the path planning step, and the offline decision-making step. The following steps detail the technical details of each link:

[0051] S31, the denial factor judgment step, includes detecting whether the current environment of the drone belongs to at least one of the denial types of GNSS signal denial, electromagnetic interference or communication interruption. For example, when the drone enters the target area, the multi-source sensor 21 collects environmental data in real time. If the GNSS signal strength is lower than -130dBm for 5 seconds, it is determined to be a GNSS denial environment; if the 2.4GHz band interference power exceeds -80dBm and lasts for 200ms, it is determined to be an electromagnetic interference environment; if the communication link loses three consecutive heartbeat packets, it is determined to be a communication interruption environment. It can be understood that the denial factor judgment function of this step realizes accurate identification of the environment type through a multi-dimensional threshold trigger mechanism.

[0052] It should be noted that the determination threshold is set based on the sensor accuracy: -130 dBm corresponds to the sensitivity limit of the GNSS module, and -80 dBm is the anti-interference critical value of the frequency hopping spread spectrum transceiver 105 .

[0053] S32, multi-source sensor data acquisition and fusion positioning steps, including when GNSS signal rejection is detected, using multi-source sensors for anti-interference positioning through the navigation module. Specifically, angular velocity and acceleration data are collected at a frequency of 100Hz through the inertial measurement unit, ORB feature matching is performed through the binocular vision camera to generate an environmental point cloud, a three-dimensional point cloud map with a precision of 10cm is generated through a solid-state lidar, geomagnetic fingerprint matching is performed through a three-axis geomagnetic sensor, and the arrival time difference of the 40kHz acoustic beacon is analyzed through an ultrasonic receiver; an extended Kalman filter and a particle filter hybrid algorithm are used to perform dynamic weighted fusion of angular velocity and acceleration data, environmental point cloud, three-dimensional point cloud map, geomagnetic fingerprint, and time difference, and output six-degree-of-freedom posture data with a position error of less than 3 meters; based on the six-degree-of-freedom posture data, flight attitude correction instructions for the UAV are generated.

[0054] like Figure 4As shown, the multi-source sensor data acquisition module is first executed. The three-axis angular velocity and acceleration raw data output by the ADIS16470 inertial measurement unit 24 at a frequency of 100Hz are obtained through IMU data acquisition S403; the binocular vision camera 25 is driven by the ORB algorithm to extract feature points and generate an environmental point cloud map through visual data acquisition S404; the RoboSense M1 solid-state laser radar 26 is controlled by LiDAR data acquisition S405 to generate a 10cm precision three-dimensional point cloud; the HMC5883L three-axis geomagnetic sensor 27 is matched with the pre-stored geomagnetic fingerprint library through geomagnetic data acquisition S407 (error <2m); the 40kHz acoustic beacon TDOA data (±0.3m accuracy @50m) received by the ultrasonic receiver 28 is analyzed through acoustic positioning data acquisition S406; the above data are aggregated to the data acquisition node S401 to complete the multi-source information integration. The multi-source data fusion positioning module then executes: Data fusion node S402 concurrently executes an extended Kalman filter S408 (state equation update pose prediction) and a particle filter S409 (1000 particles initialize and update weights). After adaptive weighted output fusion processing S410 (EKF weight α = 0.7, particle filter weight β = 0.3), six-degree-of-freedom pose data (including XYZ position and pitch / roll / yaw angles) with a position error of less than 3 meters is generated. Finally, the navigation controller 23 applies the PID algorithm () to generate flight attitude correction commands. It can be understood that this step fully implements the technical feature of "anti-interference positioning using multi-source sensors under GNSS signal denial conditions." The 1000-particle size of the particle filter S409 ensures a positioning confidence level of >99.7%.

[0055] It should be noted that the position error of <3m of the pose data meets the accuracy requirement of "outputting the six-degree-of-freedom pose data of the drone".

[0056] S33, the steps of dynamic switching of anti-interference communication links and relay networking, including dynamically switching at least one of frequency hopping spread spectrum communication, laser communication and self-organizing network relay transmission communication through the communication module to maintain the communication link when electromagnetic interference or communication interruption is detected. For example, when switching to frequency hopping spread spectrum communication, the 0.1-6GHz frequency band is scanned through real-time spectrum sensing and a 2.4GHz channel with an interference intensity lower than -90dBm is selected, and channel hopping is performed at intervals of 1ms according to a pseudo-random sequence; when switching to laser communication, a 1550nm wavelength laser communication link is activated and a directional optical transmission with a beam divergence angle of less than 2mrad is established, with a modulation rate of 1Gbps; when switching to self-organizing network relay transmission, a HELLO message with a period of 200ms is sent based on the OLSR protocol to form a multi-hop Mesh network with an end-to-end delay of less than 500ms, so that multiple drones form a dynamic relay network to bypass communication blind spots.

[0057] The spectrum sensing unit S51 monitors the environmental interference intensity in real time. When the interference intensity is detected to be greater than -90dBm, the frequency hopping spread spectrum communication S53 mode is activated: the AD9361 RF front end selects an available channel through a 1024-point FFT scan and performs channel hopping at 1ms intervals according to a pseudo-random sequence. If pulse interference is detected, the 500μs emergency frequency hopping mode is enabled. When the interference intensity is ≤-90dBm, the radio silence state is further determined: if silence is confirmed, the laser communication link establishment S52 mode is started. The servo motor adjusts the transmission direction of the laser communication unit 106, and a 1550nm 1Gbps directional optical link is established through a challenge-response protocol. RS255239 encoding ensures a bit error rate of less than 1×10 -6 If the communication is not in silent mode, the communication distance is determined. If the distance is greater than 10 km, the system establishes ad hoc relay transmission mode (S54). Based on the OLSR protocol, HELLO messages are broadcast at a 200ms period. The Dijkstra algorithm calculates the shortest path within 5 hops and reconstructs the routing table within 50ms if the link is interrupted. If the distance is ≤ 10 km, the system returns to frequency-hopping spread spectrum communication mode (S53). This step corresponds to three communication switching modes, with the >-90dBm threshold trigger mechanism in spectrum sensing (S51) implementing real-time spectrum sensing.

[0058] It should be noted that the 10 km distance threshold is set based on the maximum effective transmission distance of the laser communication unit 106 to ensure that the ad hoc network relay transmission S54 is only activated in the beyond-line-of-sight scenario.

[0059] S34, Dynamic Threat Map Construction and Reinforcement Learning Path Decision-Making Steps, includes generating obstacle avoidance paths using a path planning module based on a three-dimensional threat map and reinforcement learning algorithms, and executing offline decision-making when communication interruptions exceed a set threshold. Specifically, a 64×64×8 gridded three-dimensional threat map is constructed by integrating the location of electromagnetic interference sources, 5-meter resolution terrain elevation data, and real-time wind speed and precipitation meteorological data. An obstacle avoidance strategy is generated using a deep Q-network algorithm based on heading angle, flight altitude, and speed. When communication interruptions exceed 10 seconds, the system switches to offline replanning mode, executing a 1-second local path replanning cycle based on visual odometry and the A* algorithm.

[0060] like Figure 6As shown, multi-source data integration and threat modeling are first performed: electromagnetic interference sources (position accuracy ±10m), 5m resolution DEM terrain elevation data, and real-time wind speed information from meteorological radar are input into the three-dimensional threat map construction unit 108 through data integration S63 to generate a 64×64×8 gridded three-dimensional threat map (grid size 78m×78m×25m); then intelligent decision-making and full-modal path planning are performed: the reinforcement learning decision engine 109 performs reinforcement learning path planning S62, processes the threat map based on the DQN decision S64, and outputs a real-time obstacle avoidance strategy S66 with a heading angle, a flight altitude of 50-200m, and a speed of 10-15m / s; at the communication status detection point, it is determined whether the communication interruption exceeds 10 seconds: if so, it switches to offline replanning S65 mode, and performs local path replanning with a cycle of 1 second (path deviation <5%) by fusing the A* algorithm with the visual odometry; if not, it continues to run the DQN decision S64 and real-time obstacle avoidance S66 in a closed loop, forming a path planning mechanism that is adaptive to dynamic threat environments. It can be understood that this step realizes the dual-mode switching function of "generating an obstacle avoidance path based on a three-dimensional threat map" and "executing offline decision-making when communication is interrupted".

[0061] It should be noted that the communication status detection point in the closed-loop decision-making process corresponds to the technical characteristic of "deep Q network algorithm decision delay ≤ 2 seconds", ensuring that the 10-second interruption threshold strictly matches the algorithm performance.

[0062] S35. Multi-level safety redundancy mechanism triggering and emergency response steps include detecting a main control system failure, flight attitude failure, or communication hijacking failure through the safety redundancy module, triggering the corresponding fault-tolerant response in the dual-system hot backup, emergency parachute, or data self-destruction mechanism. Specifically, the failure factors of the drone are obtained, including main control system failure, flight attitude failure, and communication hijacking failure. When a main control system failure occurs, the heartbeat packet detection mechanism of the master-slave FPGA architecture is used to enable the backup system to take over control within 10ms. When the vibration amplitude of the axial acceleration is detected to be greater than 15g and the duration exceeds 50ms, and the Kalman filter predicts a pitch angle deviation greater than 30°, a flight attitude failure is determined, triggering the gunpowder ejection mechanism to deploy the emergency parachute. When a communication hijacking failure occurs, the three-level data erasure process of the AES-256 encrypted storage medium is initiated, and the drone is controlled to land at the preset safe coordinates.

[0063] like Figure 7As shown, when executing fault detection S73, if it is detected that the main FPGA has lost heartbeats for three consecutive times or the CRC32 check fails (any condition is met), the dual-system hot backup architecture unit 110 immediately triggers the switch execution S74, and transfers control to the backup system within 10ms through the CPLD hardware circuit; when executing risk judgment S75, if the ADXL377 vibration sensor detects that the axial acceleration is greater than 15g and lasts for more than 50ms, and the Kalman filter predicts that the pitch angle deviation is greater than 30° (risk index ≥ 0.85), the emergency parachute trigger unit 112 starts the parachute ejection S76, and the ejection mechanism response time is less than 200ms; when an SSL certificate forgery attack is detected, it is determined to be a communication hijacking, and the data self-destruction unit 111 immediately executes encryption erasure S77 to perform three-level overwriting (all 0→all 1→pseudo-random number) on the eMMC5.1 storage medium, and then the navigation module 101 executes emergency landing S78 to the preset safety coordinates (latitude and longitude error ±5m). It can be understood that this step fully realizes the three-level response of "dual system hot backup to achieve fault switching within 10ms", "triggering emergency parachute based on vibration amplitude and attitude angle deviation" and "activating data self-destruction mechanism when communication is hijacked".

[0064] It should be noted that the 0.85 risk threshold in risk assessment S75 is based on the confidence calculation rule of the attitude prediction model. When the predicted crash probability is greater than 98%, the threshold is triggered to ensure accurate response of the emergency mechanism.

[0065] Task data offline preloading and network recovery processing steps: Figure 8 As shown, offline data preloading S81 is first performed: an AES256-encrypted geomagnetic fingerprint library S82 (grid resolution 0.5m×0.5m) and 5m-resolution DEM terrain data S83 are preloaded in parallel to the eMMC5.1 storage medium via a secure USB interface. After communication is restored, network recovery processing S84 is performed: data retransmission S85 is triggered in parallel (automatically uploading IMU logs, threat map update records, and raw sensor data cached during the network outage) and model fine-tuning S86 (updating the DQN policy network weight parameters of the reinforcement learning decision engine 109 with a learning rate of 0.0001 and a batch size of 32). It can be understood that this step provides the data foundation and algorithmic support for "performing offline decisions during communication interruptions," where AES256 encryption implements the "storage medium encryption protection" requirement.

[0066] It should be noted that the 5m resolution DEM data is processed by the spatial interpolation algorithm of the 3D threat map construction unit 108 to meet the ±0.5m elevation accuracy standard of the 3D threat map.

[0067] The above content is a further detailed description of the present application in conjunction with specific implementation methods, and the specific implementation of the present application cannot be considered to be limited to these descriptions. For ordinary technicians in the technical field to which the present application belongs, several simple deductions or substitutions can be made without departing from the inventive concept of the present application.

Claims

1. An adaptive anti-interference UAV delivery system for denied environments, characterized by: include: The navigation module is used to achieve anti-interference positioning through multi-source sensor fusion under GNSS signal denial conditions and output the UAV's six-degree-of-freedom pose data; A communication module, configured to dynamically switch communication modes to maintain a communication link under conditions of electromagnetic interference or communication interruption, wherein the communication modes include frequency hopping spread spectrum communication, laser communication, and ad hoc network relay transmission communication based on the OLSR protocol; The path planning module integrates electromagnetic interference source location data, terrain elevation data, and meteorological data to build a three-dimensional threat map. It also generates obstacle avoidance paths using a deep Q-network algorithm and switches to offline replanning mode to perform local path replanning if communication is interrupted for more than 10 seconds. A safety redundancy module, designed to achieve failover within 10ms through dual-system hot backup, trigger an emergency parachute based on crash risk assessment based on vibration amplitude and attitude angle deviation, and activate a data self-destruct mechanism and control the drone to land at a preset safe coordinate in the event of communication hijacking; The system achieves anti-interference positioning, dynamic communication switching, intelligent path planning and adaptive adjustment of multi-level safety response through the collaborative work between various modules.

2. The adaptive anti-interference UAV delivery system for denied environments according to claim 1 is characterized in that: The navigation module includes: Multi-source sensor, integrating an inertial measurement unit, a binocular vision camera, a solid-state lidar, a three-axis geomagnetic sensor, and an ultrasonic receiver, to collect environmental data around the drone; The data fusion unit uses a preset extended Kalman filter and particle filter hybrid algorithm to dynamically weight the environmental data generated by the multi-source sensors and output six-degree-of-freedom posture data with a position error of less than 3m and an attitude angle error of less than 1°; The navigation controller generates a flight attitude correction instruction for the UAV based on the six-degree-of-freedom posture data through a PID algorithm.

3. The adaptive anti-interference UAV delivery system for denied environments according to claim 1 is characterized in that: The communication module includes: Frequency-hopping spread spectrum transceiver, which uses real-time spectrum sensing to select channels with interference strength less than -90dBm and performs channel hopping at 1ms intervals in the 2.4GHz band; Laser communication unit, used to establish a 1Gbps directional optical link in a radio-quiet area using a 1550nm wavelength laser; The self-organizing network protocol stack is used to build a multi-hop mesh network based on the OLSR protocol. It maintains the topology through HELLO and TC messages, making the end-to-end delay less than 500ms, and enables multiple drones to form a dynamic relay network to bypass communication blind spots.

4. The adaptive anti-interference UAV delivery system for denied environments according to claim 1 is characterized in that: The path planning module includes: a three-dimensional threat map construction unit, configured to integrate electromagnetic interference source location data, 5m resolution terrain elevation data, and real-time wind speed and precipitation meteorological information to analyze the distribution of dynamic obstacles and generate the three-dimensional threat map in a 64×64×8 raster format; A reinforcement learning decision engine unit is used to encode the three-dimensional threat map into a tensor input to the deep Q network, output the obstacle avoidance strategy of heading angle, flight altitude and speed, and perform offline path tracking by integrating visual odometry with the A* algorithm when communication is interrupted.

5. The adaptive anti-interference UAV delivery system for denied environments according to claim 1 is characterized in that: The safety redundancy module includes: Dual-system hot standby architecture unit, using a master-slave FPGA architecture, implements fault detection through a heartbeat packet mechanism, and uses the backup system to take over control within 10ms; The emergency parachute trigger unit is used to trigger the gunpowder ejection mechanism to deploy the parachute when the vibration amplitude is detected to be greater than 15g and the pitch angle deviation is greater than 30°; The data self-destruct unit is used to perform level 3 data erasure and control the drone to land at encrypted safe coordinates when a communication hijacking failure is detected.

6. An adaptive anti-interference UAV delivery method for a denied environment, characterized by: include: The denial environment determination step includes detecting whether the current environment of the UAV belongs to at least one of the denial types of GNSS signal denial, electromagnetic interference or communication interruption; Multi-source sensor data collection and fusion positioning steps, including anti-interference positioning using multi-source sensors through the navigation module when GNSS signal denial is detected; The steps of dynamically switching the anti-interference communication link and relay networking include dynamically switching at least one of frequency hopping spread spectrum communication, laser communication, and ad hoc network relay transmission communication through the communication module to maintain the communication link when electromagnetic interference or communication interruption is detected; Dynamic threat map construction and reinforcement learning path decision-making steps, including generating obstacle avoidance paths based on the three-dimensional threat map and reinforcement learning algorithms through the path planning module, and executing offline decisions when communication interruptions exceed a set threshold; Multi-level safety redundancy mechanism triggering and emergency response steps include detecting main control system failure, flight attitude failure or communication hijacking failure through the safety redundancy module, triggering the corresponding fault-tolerant response in the dual-system hot backup, emergency parachute or data self-destruction mechanism.

7. The adaptive anti-interference UAV delivery method for denied environments according to claim 6 is characterized in that: The multi-source sensor data collection and fusion positioning step specifically includes: The inertial measurement unit collects angular velocity and acceleration data at a frequency of 100Hz, and the binocular vision camera performs ORB feature matching to generate an environmental point cloud. The solid-state lidar generates a 3D point cloud map with a 10cm accuracy. The three-axis geomagnetic sensor performs geomagnetic fingerprint matching, and the ultrasonic receiver analyzes the arrival time difference of the 40kHz acoustic beacon. A hybrid algorithm of extended Kalman filtering and particle filtering is used to dynamically weighted fuse the angular velocity and acceleration data, the environmental point cloud, the three-dimensional point cloud map, the geomagnetic fingerprint, and the time difference, and output six-degree-of-freedom pose data with a position error of less than 3 meters; A flight attitude correction instruction for the UAV is generated based on the six-degree-of-freedom posture data.

8. The adaptive anti-interference UAV delivery method for denied environments according to claim 6 is characterized in that: The steps of dynamically switching the anti-interference communication link and establishing the relay network specifically include: When switching to frequency-hopping spread spectrum communication, the system scans the 0.1-6 GHz frequency band through real-time spectrum sensing and selects a 2.4 GHz channel with an interference strength lower than -90 dBm. It then performs channel hopping at 1 ms intervals according to a pseudo-random sequence. When switching to laser communication, the 1550nm wavelength laser communication link is activated and directional optical transmission with a beam divergence angle of less than 2mrad is established, with a modulation rate of 1Gbps; When switching to ad hoc network relay transmission, HELLO messages with a period of 200ms are sent based on the OLSR protocol to form a multi-hop Mesh network with an end-to-end delay of less than 500ms, allowing multiple drones to form a dynamic relay network to bypass communication blind spots.

9. The adaptive anti-interference UAV delivery method for denied environments according to claim 6, characterized in that: The steps of constructing the dynamic threat map and strengthening the learning path decision-making specifically include: By integrating the location of electromagnetic interference sources, 5-meter resolution terrain elevation data, and real-time wind speed and precipitation meteorological data, a 64×64×8 grid of three-dimensional threat maps was constructed. Generate obstacle avoidance strategies based on heading angle, flight altitude, and speed through a deep Q-network algorithm; When the communication interruption exceeds 10 seconds, the system switches to offline replanning mode and performs local path replanning with a cycle of 1 second based on visual odometry and the A* algorithm.

10. The adaptive anti-interference UAV delivery method for denied environments according to claim 6, characterized in that: The multi-level safety redundancy mechanism triggering and emergency response steps specifically include: Obtaining failure factors of the drone, wherein the failure factors include main control system failure, flight attitude failure, and communication hijacking failure; When the main control system fails, the heartbeat packet detection mechanism of the master-slave FPGA architecture allows the backup system to take over control within 10ms; When the vibration amplitude of the axial acceleration is detected to be greater than 15g and the duration exceeds 50ms, and the Kalman filter predicts that the pitch angle deviation is greater than 30°, a flight attitude failure is determined to have occurred, triggering the gunpowder ejection mechanism to deploy the emergency parachute; When a communication hijack failure occurs, the three-level data erasure process of the AES-256 encrypted storage medium is initiated, and the drone is controlled to land at the preset safe coordinates.

Images