Low-altitude economic flight data management method and system based on edge computing

Abstract

The present invention relates to the field of data management technology, and specifically to a low-altitude economic flight data management method and system based on edge computing, including edge computing nodes deployed on aircraft and ground base stations, sensors for collecting multi-source data, a module for preprocessing data, a ring data queue caching mechanism, a composite biometric code generation module, a distributed storage edge node, an abnormal fluctuation feature extraction module, an equipment health status and trend analysis and calculation module, a virtual power manager, a main power supply and a backup power supply, an abnormal behavior monitoring module, and a defense strategy library. The technical solution collects multi-source data through edge computing nodes and sensors, generates feature codes and encrypts distributed storage after preprocessing and caching, realizes dynamic resource adjustment through state evaluation, digital twins and optimization algorithms, uses neural networks to manage power, and combines abnormal monitoring to ensure data security, forming a full-process management system.

Description

Technical Field

[0001] The present invention relates to the field of data management technology, and in particular to a low-altitude economic flight data management method and system based on edge computing. Background Art

[0002] As an emerging economic form developed based on low-altitude airspace, the low-altitude economy has shown great potential in scenarios such as drone logistics distribution, agricultural plant protection, and emergency rescue. The degree of airspace openness and equipment density continue to increase.

[0003] However, its large-scale development faces the dual challenges of data management efficiency and security. Traditional cloud-based processing models require data to be transmitted over long distances to remote servers, resulting in round-trip latency exceeding 200ms. This makes it difficult for low-altitude flight equipment to respond in real time to sudden environmental changes (such as obstacle avoidance within 50ms). Furthermore, low-altitude flight equipment has limited computing resources, and centralized data storage is an easy target for attack. Globally, malicious intrusions into low-altitude communication links have increased by 67% annually, posing a high risk of data leakage.

[0004] Moreover, different low-altitude missions have very different requirements for computing power and communication resources. Traditional static resource allocation strategies result in only 55% equipment computing power utilization and over 40% waste of communication bandwidth. Existing monitoring relies on threshold alarms and can only identify explicit faults. 73% of accidents caused by hidden faults fail to receive warnings 24 hours in advance. Traditional data management models have become a bottleneck for the large-scale development of the low-altitude economy. There is an urgent need for new data management methods that integrate edge computing, intelligent algorithms, and active safety mechanisms to achieve low-latency processing, dynamic resource optimization, and full-link risk prevention and control.

[0005] Based on the above problems, the present invention provides a low-altitude economic flight data management method and system based on edge computing. Summary of the Invention

[0006] In response to the shortcomings of the existing technology, the present invention provides a low-altitude economic flight data management method and system based on edge computing to solve problems such as significant real-time defects, weak data security system, low resource scheduling efficiency and insufficient fault prediction capabilities, and realize a high real-time, high security and high reliability data management system in low-altitude economic scenarios.

[0007] To achieve the above objectives, one of the present inventions is to provide a low-altitude economic flight data management method based on edge computing, comprising:

[0008] By deploying edge computing nodes on aircraft and ground base stations, multi-source data from low-altitude flight equipment is collected and pre-processed in the edge computing nodes. A ring-shaped data queue cache mechanism is used to manage the data in segments, dynamically allocating cache space for each data segment and performing data storage or cleanup operations based on the occupancy of the cache space.

[0009] Generate a composite biometric code for the flight data in the cache, encrypt the data blocks in pieces using a chaotic encryption algorithm, and store the encrypted data blocks in a distributed manner on multiple edge nodes. Verify on the edge computing node whether the composite biometric code is consistent with the stored code. If not, initiate a secondary verification mechanism to further confirm the user's identity.

[0010] Extract abnormal fluctuation characteristics of flight data, calculate equipment health status coefficient and trend analysis coefficient based on pre-processed data, and simultaneously predict the future status of the aircraft by combining historical flight data and multi-dimensional parameters collected in real time. If the threshold is exceeded, a fault warning is triggered.

[0011] A digital twin model is built based on current flight data and resource allocation plans, and the flight trajectory and energy consumption are predicted through a physical simulation engine. If the simulation results deviate from the preset threshold by more than 5%, an adaptive entropy-weighted hybrid optimization algorithm is triggered, and a priority queue is generated based on the flight mission urgency and resource weights, enabling dynamic adjustment of resource allocation strategies and iterative adjustments to resource allocation parameters.

[0012] Inputting the power consumption characteristic data of the aircraft into a heterogeneous device management neural network to perform power management prediction, generate device power consumption prediction parameters, and perform power response control on the main power supply and backup power supply of the aircraft through a virtual power manager to generate power regulation parameters;

[0013] Monitor user operations on aircraft data. When abnormal behavior is detected, record the behavior and store it in the defense policy library for subsequent security policy optimization. At the same time, use the real-time monitoring capabilities of edge computing nodes to quickly respond to abnormal behavior and ensure the security and reliability of aircraft data.

[0014] The second aspect of the present invention is to provide a low-altitude economic flight data management system based on edge computing, which is used to implement a low-altitude economic flight data management method based on edge computing, including the following modules:

[0015] Edge data acquisition and preprocessing module: Through edge computing nodes deployed on aircraft and ground base stations, it collects multi-source data such as device status, communication signals, and environmental parameters in real time, performs denoising and standardization preprocessing, and adopts a ring data queue caching mechanism to manage data in segments according to time windows, and dynamically cleans up redundant data based on cache occupancy;

[0016] Distributed security storage and verification module: Generates a composite biometric code that includes the device hardware fingerprint and user operation characteristics, encrypts the data blocks in fragments using a dynamic chaotic encryption algorithm, and stores them in multiple edge nodes via a distributed hash table. The code is first verified during access, and abnormalities trigger SMS verification or biometric secondary verification.

[0017] Intelligent status assessment and prediction module: Calculates equipment health status coefficients and trend analysis coefficients based on multi-dimensional data, detects abnormal fluctuations using the Z-score and DBSCAN algorithms, generates sub-fluctuation impact factors, triggers fault warnings when thresholds are exceeded, and outputs fault type and remaining time.

[0018] Digital twin-driven resource optimization module: This module builds a lightweight digital twin model to simulate flight trajectory and energy consumption. If the deviation exceeds 5%, an adaptive entropy weight algorithm is triggered to dynamically allocate computing power and bandwidth resources based on task urgency, generating a priority queue.

[0019] Intelligent power management module: Utilizes LSTM neural networks to predict power consumption parameters, coordinates the main power supply and supercapacitors through a virtual power manager, implements a stepped power supply strategy, and combines Bayesian networks to optimize power distribution and extend battery life.

[0020] Abnormal behavior monitoring module: Establishes a user operation baseline through unsupervised learning, monitors access frequency, path depth and other indicators in real time, blocks the connection when anomalies are detected, records the characteristics to the defense strategy library and updates the security model.

[0021] The third aspect of the present invention is to provide a computer device, which includes: a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to realize a low-altitude economic flight data management method based on edge computing.

[0022] A fourth aspect of the present invention is to provide a computer-readable storage medium, wherein the storage medium stores at least one instruction, at least one program, code set or instruction set, and the at least one instruction, the at least one program, the code set or instruction set is loaded and executed by a processor to implement a low-altitude economic flight data management method based on edge computing.

[0023] Compared with the existing technology, the present invention provides a low-altitude economic flight data management method and system based on edge computing, which has the following beneficial effects:

[0024] This solution deploys edge computing nodes on aircraft and ground base stations to collect multi-source data such as device status, communication signals, and environmental parameters in real time. It implements efficient data management through denoising, standardized pre-processing, and a ring queue caching mechanism, reducing end-to-end latency to less than 50ms and resolving the real-time bottleneck caused by cloud transmission.

[0025] This solution combines a composite biometric code that binds the device's hardware fingerprint with the user's operating habits, with a dynamic chaotic encryption algorithm to fragment and store data blocks on multiple edge nodes. It also introduces a dual verification mechanism (code verification + secondary authentication) to build a data security barrier that prevents leakage and tampering, meeting the requirements of privacy protection regulations.

[0026] This solution uses digital twin technology to build a flight trajectory and energy consumption simulation model. It dynamically adjusts resource allocation, such as computing power and bandwidth, through an adaptive entropy weight algorithm. Combined with task priority queues (e.g., prioritizing emergency tasks), it improves resource allocation efficiency by 30% and computing power utilization to over 90%. It also uses an LSTM neural network to predict power consumption parameters and coordinates the main and backup power supplies to implement a stepped power supply strategy, extending the aircraft's flight range by 15%-20%.

[0027] This solution uses health status coefficient modeling, abnormal fluctuation cluster analysis, and historical baseline comparison to provide early warning of potential equipment failures two hours in advance, reducing maintenance costs by 40%. It also establishes a user operation baseline based on unsupervised learning, blocks abnormal access in real time, and updates the defense policy library to achieve self-evolution of security policies.

[0028] This solution breaks through the performance and security bottlenecks of traditional cloud computing by forming a closed-loop management system of "data collection-secure transmission-intelligent analysis-dynamic optimization", providing core technical support with low latency, high reliability, strong security and low energy consumption for low-altitude scenarios such as drone logistics and emergency rescue, and significantly improving real-time decision-making capabilities and system robustness in low-altitude economic scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0030] Figure 1 This is a block diagram of the system layered architecture of the present invention;

[0031] Figure 2 A waterfall flow chart of the method steps of the present invention;

[0032] Figure 3This is a flow chart of data encryption and storage of the present invention;

[0033] Figure 4 This is a flow chart of intelligent power management of the present invention;

[0034] Figure 5 This is the abnormal behavior monitoring state diagram of the present invention. DETAILED DESCRIPTION

[0035] In order to make the objectives, technical solutions and advantages of this application clearer, the implementation methods of this application will be further described in detail below with reference to the accompanying drawings.

[0036] It aims to address the problems existing in low-altitude economic flight data management, such as high real-time processing delay, prominent data security risks, inefficient resource scheduling, delayed fault prediction and extensive energy consumption management.

[0037] This solution proposes a low-altitude economic flight data management method based on edge computing.

[0038] By deploying edge computing nodes, multi-source data such as device status and environmental parameters can be collected in real time to achieve efficient data management, reduce end-to-end latency to less than 50ms, and resolve the real-time bottleneck caused by cloud transmission.

[0039] Utilizes a composite biometric code that combines the device's hardware fingerprint with the user's operating habits to build a data security barrier that prevents leakage and tampering, meeting privacy protection regulations.

[0040] Based on the flight trajectory and energy consumption simulation model, the adaptive entropy weight algorithm dynamically adjusts computing power, improving resource allocation efficiency by 30% and computing power utilization to over 90%;

[0041] Through health status coefficient modeling, abnormal fluctuation cluster analysis, and historical baseline comparison, potential equipment failures can be warned two hours in advance, maintenance costs can be reduced by 40%, and self-evolution of security strategies can be achieved.

[0042] This solution breaks through the performance and security bottlenecks of traditional cloud computing by forming a closed-loop management system of "data collection-secure transmission-intelligent analysis-dynamic optimization", providing core technical support with low latency, high reliability, strong security and low energy consumption for low-altitude scenarios such as drone logistics and emergency rescue, and significantly improving real-time decision-making capabilities and system robustness in low-altitude economic scenarios.

[0043] Example 1, as Figure 1-Figure 5 As shown, the low-altitude economic flight data management method based on edge computing provided in an embodiment of the present application is exemplified, including the following steps:

[0044] Step S100: Through the edge nodes of the aircraft and ground base stations, the status data of equipment such as IMU and BMS, 2.4GHz / 5.8GHz communication signals, and environmental parameters such as Beidou positioning are collected at a frequency of 100Hz. After Kalman filtering and IQR algorithm to remove outliers, Z-score normalization is applied. A circular data queue is used to cache data in 500ms windows. Redundant data is dynamically cleaned based on the LRU algorithm. When the cache occupancy rate exceeds 80%, non-critical segments from 24 hours ago are preferentially deleted.

[0045] Step S200: The device's MAC address and control pressure curve are integrated to generate a 128-bit composite biometric code. The data block is dynamically encrypted using the Logistic Chaos Algorithm and distributedly stored via IPFS based on network latency (1KB shards + 1 replica for ≤50ms, variable shards + 3 redundant shards for >100ms). The code is first verified upon access. If the Hamming distance is greater than 5 bits, a secondary verification via SMS verification code or iris recognition is triggered.

[0046] Step S300: Calculate the health factor H (H = 0.5S_d / 10 + 0.3S_c + 0.2S_e) based on the device status (engine / battery / attitude scores), communication quality, and environmental risks. Analyze the parameter change rate using a sliding window. Use DBSCAN to cluster abnormal data and use the Weibull model to predict the fault type and remaining time. A yellow alert is triggered when H < 0.6, and a red alert is triggered when H < 0.4.

[0047] Step S400: A lightweight twin model is built in the Unity engine. Trajectory and energy consumption are simulated with a 20ms step size. The adaptive entropy weighting algorithm is triggered when the deviation exceeds 5%. Computing power, bandwidth, and power weights are calculated based on the urgency of the task (with an emergency rescue weight of 1.0). Resources are scheduled across nodes using the Hungarian algorithm. High-priority tasks are dynamically allocated an additional 20% of computing power, with a scheduling delay of <100ms.

[0048] Step S500: Current RMS, voltage ripple, and other parameters are input into a three-layer LSTM network to predict power consumption (error ≤ 5%). The virtual power manager then provides power by load: low loads (≤ 50W) use only supercapacitors, medium loads (50-150W) use a main power supply in parallel with capacitors, and high loads (>150W) use capacitors to provide pulsed current. When the lithium battery temperature is >45°C, the current is reduced to 0.5°C, extending battery life by over 15%.

[0049] Step S600: An operational baseline is established using the Isolation Forest algorithm to monitor for anomalies such as sensitive data access and high-frequency operations during non-working hours. When an anomaly is detected, the first level logs the data, the second level blocks the connection, and the third level initiates isolation and updates the defense policy library. Response time to anomalies is less than 500ms, and the detection rate for new attacks increases by 3-5% per week, ensuring reliable and controllable data operations.

[0050] like Figure 2 As shown, step S100 is used for edge node data collection and cache management, providing raw materials for subsequent analysis, including:

[0051] S110:

[0052] S1101: Equipment status data link construction:

[0053] The engine ECU, BMS battery management system, and IMU inertial measurement unit are connected via the CAN bus (baud rate 1Mbps). A master-slave synchronization mechanism is adopted (Master is the flight control computer, Slave is the sensor node). A synchronization pulse (SyncPulse) is sent every 10ms to calibrate the sampling clock.

[0054] Define the binary data frame format, including:

[0055] Frame header (0xAA55, 2 bytes), timestamp (64-bit UTC nanosecond value, 8 bytes), sensor ID (1 byte, 0x01-0x05 corresponds to IMU / ECU / BMS, etc.), data body (variable length, such as an IMU containing 3-axis acceleration / angular velocity, totaling 18 bytes), CRC checksum (2 bytes). The IMU / barometer frequency is 100 Hz, and the BMS / ECU frequency is 20 Hz, meeting the real-time requirements of flight control.

[0056] S1102: RF front-end design:

[0057] It uses a software defined radio (SDR) architecture based on RTL-SDR (Software Defined Radio), covers the 2.4GHz / 5.8GHz UAV communication frequency band, has a sampling rate of 2MSPS, and uses FPGA (Xilinx Zynq series) for real-time calculation:

[0058] Received signal strength indicator (RSSI, unit dBm): RSSI = 10× (power spectral density × channel bandwidth);

[0059] Round-trip time delay (RTT, in milliseconds): Send ICMPEchoRequest and record the response time, taking the average of 5 samples;

[0060] Signal-to-noise ratio (SNR, in dB): SNR = 10× , the noise power is estimated by sampling the idle channel;

[0061] S1103: It uses a high-precision barometer (accuracy ±0.1hPa, converted altitude error ±1m), a temperature and humidity sensor (accuracy ±0.5℃ / ±2%RH), and an ultrasonic anemometer (range 0-30m / s, accuracy ±0.1m / s), and is connected to the edge node via the I2C bus (rate 400kHz).

[0062] Simultaneously receives GPSL1 (1575.42MHz) and BeiDou B1 (1561.098MHz) signals, fuses multi-source positioning data through the Extended Kalman Filter (EKF), and outputs longitude (accuracy ±1m), latitude (accuracy ±1m), and altitude (accuracy ±2m);

[0063] For example: In drone logistics scenarios, IMU data is used for real-time attitude calculation to ensure smooth transportation of goods; RSSI and SNR data dynamically switch communication channels (such as switching from 2.4GHz to 5.8GHz to avoid interference) to ensure the reliability of command transmission.

[0064] S120:

[0065] S1201: Time synchronization correction:

[0066] The edge node is equipped with a temperature-compensated crystal oscillator (TCXO, accuracy ±1ppm), which is calibrated once an hour using an NTP server. Linear interpolation is performed on the timestamps of the collected data to correct the clock offset between the sensor and the edge node:

[0067] formula: = +Δt× , where Δt is the total drift during the calibration period;

[0068] S1202: Outlier Detection and Repair:

[0069] For each sensor channel (such as the X-axis acceleration of the IMU), calculate the quartiles Q1 and Q3 and determine IQR = Q3-Q1;

[0070] Define the outlier range: [Q1−1.5×IQR,Q3+1.5×IQR], and values ​​outside the range are marked as outliers;

[0071] Outlier repair: Use linear interpolation of previous and next values ​​(if the i-th data point is abnormal, then );

[0072] Convert the air pressure from hPa to Pa (1hPa=100Pa), the temperature from °C to Kelvin (K=°C+273.15), and the acceleration from m / s² to g (1g=9.80665m / s²).

[0073] For each feature dimension (such as battery voltage, RSSI), calculate the mean μ and standard deviation σ. The normalized value is: , so that the data distribution obeys N(0,1);

[0074] Formula derivation: Assuming a battery voltage sequence is [3.7, 3.8, 3.5, 4.0], we calculate μ = 3.75 and σ ≈ 0.176, which after normalization becomes [0.28, 0.85, −1.42, 1.42], which is convenient for neural network input.

[0075] S130:

[0076] S1301: Segment header (20B): timestamp range (start / end nanosecond values, 8B each), data length (4B), reference count (1B), status bit (1B, 0 = free, 1 = occupied, 2 = to be cleaned up);

[0077] Data body (variable length, maximum 4KB): stores preprocessed binary data;

[0078] Initially create 10 empty segments, each with a size of 1KB;

[0079] When a write request arrives, the "free" segment is allocated first. If there is no available segment, a new segment is created dynamically (5 segments are expanded each time, and the size increases by 1.5 times the previous segment, with a maximum of 16KB per segment).

[0080] S1302: Maintain a bidirectional linked list, with the head being the most recently used segment and the tail being the least recently used segment;

[0081] When the cache occupancy is greater than 80%, check the segment status starting from the end:

[0082] If the status is "to be cleaned up" and the reference count = 0, delete it immediately and reclaim the memory;

[0083] If the status is "occupied", mark it as "to be cleaned up", reduce the reference count by 1, and postpone the cleanup;

[0084] Occupancy = ×100%, the total cache capacity is initially 10KB and can be expanded to a maximum of 128KB;

[0085] S1303: Receive data write request → Search or create data segment → Write data body → Update segment header timestamp and length → Move segment to the head of LRU list;

[0086] Find the segment based on the timestamp range → verify the timestamp validity → copy the data body to the read buffer → move the segment to the head of the LRU list → return the data pointer.

[0087] Based on the pre-processed high-reliability data, the system needs to further build a security protection system to ensure the confidentiality and integrity of the data during storage and transmission. Step S200 is used for secure storage and trusted access after data collection and cleaning, ensuring the security of data assets. Specifically:

[0088] S210:

[0089] S2101: Hardware feature collection:

[0090] Motherboard: Read UUID, manufacturer, and model through the SMBIOS protocol;

[0091] Sensor: reads the IMU calibration coefficient and the barometer factory serial number;

[0092] Communication module: obtain the MAC address of the Wi-Fi chip and the BD_ADDR of the Bluetooth module;

[0093] Concatenate the above information into a string such as "UUID_ABC123_MAC_00:11:22:33:44:55" and generate a 64-byte device fingerprint (such as 0x1a2b3c...d4e5f6) using the SHA-512 algorithm.

[0094] S2102: Operational data collection:

[0095] The pressure sensor voltage is sampled through ADC (12-bit accuracy, 0.00024V resolution) and the pressure intensity versus time curve is recorded (sampling rate 20Hz);

[0096] Monitor the time interval between command transmissions from the ground station software (e.g., the time difference between two "route update" commands) and the order in which interface elements are clicked (e.g., selecting "Photo" first and then "Waypoint 1");

[0097] Extract 12 statistical features including mean, standard deviation, peak value, rising edge slope, etc.

[0098] The DTW algorithm is used to calculate the similarity between the current operation sequence and the historical template to generate a 10-dimensional feature vector;

[0099] The device fingerprint is concatenated with the behavioral feature vector and a 128-byte composite biometric code is generated using the HMAC-SHA256 algorithm (the key is the user ID hash value).

[0100] Formula derivation: Let the device fingerprint be F, the behavior feature vector be B, and the user ID be U, then the feature code C = HMAC-SHA256(K,F||B), where K = SHA256(U).

[0101] S220:

[0102] S2201: Chaos Encryption Algorithm Integration:

[0103] Standard Logistic Mapping: +1=μ (1− ), when μ∈(3.5699456,4], it enters a chaotic state;

[0104] Dynamic parameter adjustment: According to the network delay T (unit ms), set μ=3.8+0.1×min( ,2),Ensure encryption strength is improved at high latency;

[0105] S2202: Data Sharding and Distributed Storage:

[0106] Low-latency scenario (T≤50ms): The shard size is fixed at 1KB, and one redundant copy is generated for each data block (total number of shards = number of original blocks × 2).

[0107] High latency scenario (T>100ms): Variable fragment size (512B-4KB), using Reed-Solomon encoding to generate three redundant fragments (total number of fragments = number of original blocks + 3);

[0108] Generate a pseudo-random number r through VDF (Verifiable Random Function) and calculate the storage node index: node_id=(rmodN)+1 (N is the number of available nodes);

[0109] Call VDF every 10 minutes to regenerate r and update the storage nodes of all data slices to ensure dynamic anti-attack;

[0110] When using Reed-Solomon (5,3) encoding, a maximum of three node failures are allowed, and the data recovery probability is 100%. The calculation formula is: =1- ;

[0111] ( is the single node failure rate, assuming =0.01, then ≈99.9999%)

[0112] Where k is the adaptive weight coefficient (ranging from 0.8 to 1.2), which is used to adjust the resource allocation priority.

[0113] S230:

[0114] S2301: Feature code comparison:

[0115] Extract the hardware fingerprint F' and operation behavior characteristics B' of the current device and generate the feature code C';

[0116] Calculate the Hamming distance: d = Hamming(C, C′). If d ≤ 5 (that is, 5 binary bits of difference are allowed), the verification passes.

[0117] Query the permission table based on the user ID (e.g., "normal user" can only read flight data, "administrator" can modify mission parameters), and return the set of allowed operations;

[0118] S2302: Dynamic verification code generation:

[0119] SMS verification code: Generates a 6-digit random number (range 100000-999999) and sends it to the registered mobile phone number via an encrypted channel (such as HTTPS). It is valid for 30 seconds.

[0120] Biometric identification: Minutiae feature matching of fingerprint images (such as extracting the coordinates of ridge endpoints and bifurcation points) is used to calculate the matching score, with the threshold set at 70 / 100.

[0121] Verification logic: When the first level verification fails, the second level verification is automatically triggered. The user must complete the verification within 5 minutes, otherwise the account will be locked for 15 minutes;

[0122] For example: In an emergency rescue scenario, when a device is temporarily authorized for access, the first-level verification fails due to hardware fingerprint differences, triggering the iris recognition secondary verification. After the rescuer passes, the system automatically grants temporary permission to "read real-time video streams", and the permission is immediately revoked after the mission is completed.

[0123] After completing the secure storage of data, the data needs to be deeply analyzed to evaluate the operating status of the equipment, identify potential risks in advance and trigger early warning mechanisms, and convert the securely stored data into quantitative indicators of the equipment health status to support preventive maintenance.

[0124] Specifically, step S300 is used for real-time status assessment and fault prediction, mining data value to identify risks, and includes:

[0125] S310:

[0126] S3101: Equipment status scoring system:

[0127] The value range is 0-10 minutes, according to the speed fluctuation rate (rpm_var= )calculate:

[0128] = ;

[0129] The remaining capacity (SOC) is calculated by the ampere-hour integration method, combined with the internal resistance growth rate (R_delta= ): =8× +2×(1−min(R_delta,0.2));

[0130] Calculate the standard deviation of the IMU attitude angle ( ), scoring formula: =10× Unit is degree);

[0131] S3102: Communication quality score: =0.6× +0.4×(1− );in, =−100dBm, , =200ms;

[0132] =0.5×min( )+0.5×(1− ), when the obstacle distance is >50 meters, the score for this part is 0.5;

[0133] S3103: Health coefficient synthesis: H=0.5 +0.3 +0.2 (in = ), the weights are determined by logistic regression training of historical fault data, and the AUC value reaches 0.92.

[0134] S320:

[0135] S3201: Battery voltage sequence V=[ , ,..., ], calculate the average drop rate when the window size N=50;

[0136] slope= (Δt is the sampling interval, 0.05s), (Δt is the sampling interval, 0.05s);

[0137] Calculate the average voltage drop rate over the past 7 days and standard deviation , if the current slope< − , labeled “battery degradation trend”;

[0138] S3202: Calculate the Z value of the standardized data, and mark the points with |Z|>3 as potential anomalies;

[0139] Define the neighborhood radius ϵ=0.5 (Euclidean distance) and the minimum number of samples MinPts=3;

[0140] Cluster potential anomalies into several categories, such as "communication interruption" (including RSSI < -90dBm, RTT > 500ms) and "power anomaly" (engine speed fluctuation rate > 0.2).

[0141] For each cluster, calculate the proportion of samples within the class , the impact factor is ×The distance from the cluster center to the origin, used to sort the anomaly priority;

[0142] For example, during the 30th minute of flight, the battery voltage drop rate of a drone was -0.02 V / s, which was lower than the baseline average of -0.01 V / s and exceeded 2σ. At the same time, DBSCAN detected five consecutive points belonging to the "power anomaly category," triggering a "battery thermal runaway risk" warning. The actual failure occurred 1 hour and 45 minutes later, and the warning was issued 2 hours and 15 minutes in advance.

[0143] S330:

[0144] S3301: Fault type classifier:

[0145] Feature input: health coefficient H, trend analysis coefficient (such as slope normalization value), and abnormal impact factor set;

[0146] Model architecture: 3-layer fully connected neural network, with a 10-dimensional input layer, 20-dimensional and 10-dimensional hidden layers, respectively. The output layer corresponds to the fault type (such as battery failure, communication failure, and sensor failure). Softmax activation is used, and the loss function is cross entropy.

[0147] Collect over 5,000 fault samples with a positive-to-negative sample ratio of 1:3. Use the Adam optimizer with a learning rate of 0.001 and train to a validation set accuracy of 85%.

[0148] S3302: Use the Weibull proportional hazard model, the formula is:

[0149] h(t)= (t)exp( + +...+ );

[0150] Where h(t) is the failure rate, is a feature (such as H, slope), is the regression coefficient, determined by maximum likelihood estimation;

[0151] When the prediction is battery failure, calculate the survival function S(t)=exp(− , take the t value when S(t)=0.1 as the remaining available time;

[0152] S3303: Alert Generation and Distribution:

[0153] The alarm level is divided into three levels (yellow / orange / red), corresponding to the H value range of [0.4, 0.6), [0.2, 0.4), <0.2;

[0154] Red alerts are simultaneously pushed to the flight control system (triggering automatic return), ground station operators (SMS + voice), and the operation and maintenance center (email + system pop-up window), ensuring that they are delivered within 5 seconds.

[0155] Based on the real-time status assessment results, the system needs to dynamically optimize the resource allocation strategy to adapt to the rapid changes in flight missions and environment, and use status assessment and fault prediction to provide a data basis for resource optimization. Simulation and intelligent algorithms need to be introduced to achieve dynamic scheduling to meet dynamic mission requirements.

[0156] Step S400 is used to optimize resources driven by digital twins and improve system response efficiency, including:

[0157] S410:

[0158] S4101: Multiphysics Modeling:

[0159] Kinetic model: ;

[0160] Where m is the mass of the aircraft, is the thrust, D(v)=0.5 , D(v) is the air resistance, (ρ is the air density, is the drag coefficient, A is the frontal area);

[0161] Energy consumption model: =I×V×Δt+ ×d(current I, voltage V, flight distance d);

[0162] S4102: Proper Orthogonal Decomposition (POD) is used to extract the first 10 modes, reducing the state space from 6 dimensions (position x / y / z, velocity v_x / v_y / v_z) to 3 dimensions, reducing the computational effort by 70%;

[0163] Using GPU instancing in the Unity engine to render multiple drone twins reduced the number of batch draw calls by 90% and maintained a stable frame rate of 60 FPS.

[0164] S4103: The edge node sends a status update (including position, speed, and posture) every 20ms. The digital twin model synchronizes the intermediate frames through linear interpolation to ensure that the visual delay is less than 50ms.

[0165] S420:

[0166] S4201: Normalize the computing power requirement C, bandwidth requirement B, and remaining power E to the range [0, 1].

[0167] = , = , = ;

[0168] For each indicator, calculate the probability of the i-th task = , entropy value: =− ln ;

[0169] Weight = , where m=3 (computing power / bandwidth / power);

[0170] S4202: = Urgency × × +Importance× × +Timeliness× × ;

[0171] Urgency value: emergency rescue (10) > medical transportation (8) > general logistics (5);

[0172] Using preemptive priority scheduling, high-priority tasks can interrupt the resource allocation of low-priority tasks, ensuring that the delay of critical tasks is less than 50ms;

[0173] S4203: After each resource allocation is completed, the deviation between the actual effect and the simulation result is calculated (such as the computing power utilization deviation ΔC = | − ∣ / );

[0174] If ΔC>0.1, then = +0.1×ΔC, converged to a stable value after 5 iterations;

[0175] Specifically, the entropy weighting method determines weights based on indicator variability, avoiding bias in human assignment. When computing power demand fluctuates significantly, its entropy value decreases and its weight automatically increases. For example, if the standard deviation of computing power requirements for an emergency task is three times that of a normal task, its computing power weight will increase from 0.3 to 0.5, which meets actual needs.

[0176] S430:

[0177] S4301: Collects the CPU utilization (%), GPU memory usage (MB), and remaining communication bandwidth (Mbps) of each node in real time and reports them to the central manager via the gRPC protocol.

[0178] Construct the global resource matrix R= , where n is the number of nodes, m is the resource type (CPU / GPU / bandwidth), is the remaining amount of resource j at node i;

[0179] S4302: Convert task allocation into a bipartite graph matching problem. The elements of the cost matrix C are the processing delay of task k at node i (formula: = + is the task data volume, is the node CPU frequency, is the node bandwidth);

[0180] When the CPU utilization of a node is greater than 85%, load migration is triggered, and the unfinished parts of low-priority tasks are migrated to nodes with utilization less than 30%. The migration cost is calculated by estimating the data volume and bandwidth ( = + , S is the state data size);

[0181] S4303: Allocate a dedicated token bucket for high-priority tasks (e.g., emergency task token generation rate 100 Mbps, bucket capacity 500 MB), ensuring priority bandwidth allocation during burst traffic.

[0182] Three levels of service quality are defined: real-time control (latency < 10ms), video streaming (latency < 100ms), and log transmission (best effort). Differentiated services are implemented through DSCP marking.

[0183] One of the goals of resource optimization is to improve energy efficiency. Therefore, it is necessary to further combine power consumption prediction to achieve intelligent management of the power supply system. Resource allocation optimization needs to be coordinated with energy management to balance performance requirements and endurance.

[0184] Step S500 is used for intelligent power management and energy consumption optimization to extend the battery life of the device, and includes:

[0185] S510:

[0186] S5101: Building a high-precision power consumption prediction model:

[0187] Current effective value (RMS), voltage ripple factor (γ= ), task type one-hot encoding (3-dimensional vector, such as logistics = 100, inspection = 010, rescue = 001);

[0188] Output labels: average power (W), peak power (W) for the next 15 minutes;

[0189] Min-Max normalization is used to scale the current / voltage characteristics to [0,1], formula: = ;

[0190] S5102: 3-layer LSTM (64 neurons per layer, dropout rate 0.2) + 2-layer fully connected (32D and 1D), with tanh and linear activation functions respectively;

[0191] Mean absolute error (MAE) + root mean square error (RMSE) weighted sum, formula: Loss = 0.7 × MAE + 0.3 × RMSE;

[0192] Batch size 32, epoch number 50, learning rate 0.0001, early stopping (patience = 10) to prevent overfitting, validation set MAE = 4.2W, RMSE = 5.8W;

[0193] S5103: After each flight, the newly collected 500 samples are added to the training set and the model is retrained to ensure that the model adapts to power consumption drift caused by equipment aging or environmental changes.

[0194] S520:

[0195] S5201: Power status monitoring:

[0196] Monitoring parameters include SOC (accuracy ±2%), internal resistance (accuracy ±5mΩ), and temperature (accuracy ±1°C). When SOC < 20%, it is marked as "low battery state";

[0197] Monitor capacitor voltage (accuracy ±0.1V) and equivalent series resistance (ESR, accuracy ±10mΩ), and start charging when the voltage is less than 80% of the rated value;

[0198] S5202: Stepped power supply strategy:

[0199] Low load mode:

[0200] Only supercapacitors are used for power supply, and the current limit is ≤3A, which prolongs the life of lithium batteries;

[0201] When the supercapacitor voltage is less than 2.5V (rated 3V) and the main power supply SOC is greater than 80%, charge at a current of 1A;

[0202] Medium load mode: The main power supply and supercapacitor are connected in parallel to provide power. The current distribution formula is:

[0203] Ibattery=I× ( is the internal resistance of the main power supply, is the supercapacitor ESR);

[0204] High load mode:

[0205] When the main power supply is fully output, the supercapacitor provides instantaneous pulse current (maximum 10A) with a duration of ≤5 seconds to prevent overcurrent damage to the lithium battery;

[0206] S5203: When the main power supply temperature is greater than 45°C, the charging current is automatically reduced to 0.5C and the supercapacitor is used first in the medium load mode until the temperature drops below 40°C.

[0207] Energy efficiency calculation: Assuming the energy density of lithium batteries is 200Wh / kg and that of supercapacitors is 10Wh / kg, in low-load mode, the hourly energy consumption is 50Wh, of which the supercapacitor contributes 40Wh (80%) and the lithium battery only consumes 10Wh. This saves 80% of energy compared to a full lithium battery solution, effectively extending battery life:

[0208] Δt= × Battery life improvement factor (m is the battery mass, the factor is 1.2);

[0209] S530:

[0210] S5301: Combined with the power consumption prediction model, the A* algorithm is used to calculate the minimum energy consumption path. The constraints are maximum flight speed ≤ 15m / s and obstacle avoidance distance ≥ 2m.

[0211] Place heavy objects as close to the center of gravity of the aircraft as possible to reduce the energy consumption of attitude adjustment. The formula is: ΔE=0.5×J×( − )(Moment of inertia J, angular velocity change ω− );

[0212] S5302: Based on real-time power consumption prediction, an economic throttle is used during level flight (maximum motor efficiency corresponds to 50%-60% throttle), and temporarily increased to 70% during climbing.

[0213] During the landing phase, the motor reverse braking is activated to convert kinetic energy into electrical energy and store it in the supercapacitor, with a recovery efficiency of approximately 25%-30%;

[0214] S5303: Adopts "constant current-constant voltage-trickle current" three-stage charging. When SOC>90%, the current is reduced to 0.1C to avoid overcharging and battery aging.

[0215] If the aircraft is idle for more than 24 hours, it automatically enters sleep mode, with power consumption reduced to <1mW and wake-up time less than 500ms.

[0216] After completing data processing, resource optimization and energy management, it is necessary to build a closed-loop security monitoring system to prevent human errors or malicious attacks. The stability and security of the system depend on real-time monitoring and defense of operational behaviors.

[0217] Step S600 is used to monitor abnormal behavior and strengthen security, thereby enhancing system reliability, and includes:

[0218] S610:

[0219] S6101: User ID, operation time (accurate to milliseconds), data type (such as route data / mission instructions / sensor parameters), operation type (read / modify / delete), file path depth (such as " / data / flight_logs / 2025 / 05" has a depth of 3);

[0220] Use UUID to mark user sessions and record contextual information such as session duration, operation frequency, and data transfer volume;

[0221] S6102: Randomly select features (such as operation frequency and path depth), randomly generate segmentation hyperplanes in the feature space, and calculate the height of the "isolation tree" of each sample;

[0222] Anomaly score formula: s(x)= , where E(h(x)) is the average tree height and c(n) is the expected tree height of n samples;

[0223] Setting the threshold =0.9, scores >0.9 are considered abnormal;

[0224] S6103: Retrain the model based on the previous 24 hours' operation logs every morning, update the normal behavior boundaries, and adapt to the differences in operation patterns during weekdays and weekends (e.g., weekend operation frequency decreases by 30%).

[0225] S620:

[0226] S6201: A user downloads sensitive data (such as files in the / system / waypoints / directory) during non-working hours (10:00 PM to 6:00 AM), and the transfer volume exceeds 10 MB.

[0227] Formula: alert = (time period∈non-working)∧(path depth≥4)∧(transmission volume>10MB);

[0228] Monitor ordinary users attempting to perform administrator operations (such as modifying flight control parameters) and match operation permissions through access control lists (ACLs). The probability of misoperation is less than 0.1%.

[0229] S6202: The ARP spoofing detection algorithm compares the IP-MAC binding table with the real-time ARP packets. If the same IP address corresponds to different MAC addresses and the attack occurs without a device replacement, it is determined to be a spoofing attack.

[0230] Use machine learning models trained on metadata of encrypted traffic (such as packet size distribution and connection duration) to identify hidden C2 (Command and Control) communications with 88% accuracy;

[0231] S6203: Graded Response:

[0232] Level 1 anomaly (such as incorrect operation): record the log and send an early warning to the user;

[0233] Level 2 anomalies (such as unauthorized access): Block the current connection and generate an incident ticket that is pushed to the security team.

[0234] Level 3 anomalies (such as ransomware attacks): trigger a full system quarantine, shut down non-essential services, and initiate data recovery processes;

[0235] S630:

[0236] S6301: Analyzes attack sample features from the defense policy library, such as IP address segments (e.g., 192.168.1.100-192.168.1.200), operation modes (e.g., 10 consecutive failed logins), and protocol fields (e.g., HTTP request headers containing ".. / .. / ").

[0237] Use the TF-IDF algorithm to vectorize features and construct the attack feature vector space;

[0238] S6302: Whenever a new attack type is detected, its feature vector is added to the training set and the anomaly detection boundary is dynamically expanded using One-Class SVM;

[0239] Simulate attack scenarios (such as brute force cracking and SQL injection) to generate adversarial sample injection models, improve robustness, and reduce the success rate of adversarial samples from 30% to 15%;

[0240] S6303: Updated detection rules are distributed to all edge nodes through a security policy management platform (such as Fortinet Security Fabric), using hash verification to ensure rule consistency.

[0241] Conduct simulated attack drills weekly to verify the effectiveness of defense strategies, record drill results, and generate improvement reports;

[0242] At this point, the system has formed a complete closed loop of "data collection-safe storage-status assessment-resource optimization-energy consumption management-safety monitoring", realizing intelligent management of the entire life cycle of low-altitude flight data.

[0243] Experimental example:

[0244] Purpose of the experiment

[0245] Verify the real-time, safety and energy efficiency of the low-altitude flight data management method based on edge computing in drone logistics.

[0246] Hardware: DJI Matrice 300 RTK drone (equipped with NVIDIA Jetson Xavier NX edge computing node), ground base station (Intel i7-12700H, 16GB RAM);

[0247] Sensors: IMU (Bosch BMI085), barometer (MS5611), RF sensor (RTL-SDR);

[0248] Software: ROS2Humble, Unity2021.3, Python3.9 (with integrated PyTorch / LSTM / Scikit-learn).

[0249] Step 1: The drone carries 5kg of cargo and executes a 3km logistics route. The edge node collects IMU data (such as acceleration) at 100Hz. , , )、BMS battery voltage( ), RSSI signal strength;

[0250] Apply Kalman filter to denoise the original data and normalize it to the [-1,1] interval using Z-score;

[0251] Use a circular queue to cache data, set the cache threshold to 80%, and automatically clear historical data from 24 hours ago;

[0252] Experimental data record table of sensor raw data and preprocessing results:

[0253]

[0254] Step 2: Extract the drone's MAC address (00:12:34:AB:CD:EF) and the pressure curve characteristics of the pilot's control handle, and generate a 128-bit signature code through SHA-256;

[0255] The data blocks are encrypted using Logistic Chaos Encryption (μ=3.8) and stored in three edge nodes through IPFS sharding.

[0256] Simulate unauthorized device access, trigger secondary verification (SMS verification code), and record the verification time.

[0257] Step 3: Calculate the health status coefficient H: equipment status score =8.2, Communication Quality Rating =0.75, Environmental Risk Score =0.6, comprehensive H=0.5×8.2 / 10+0.3×0.75+0.2×0.6=0.745;

[0258] If the battery voltage drop rate slope = -0.015V / s (lower than the baseline -0.01V / s) in three consecutive windows is detected, a "battery degradation trend" warning is triggered;

[0259] Health status score and warning parameter experimental data record table:

[0260]

[0261] Step 4: Build a digital twin model in Unity and simulate the predicted flight trajectory with a deviation of 4.2 meters (threshold 5 meters), triggering the adaptive entropy weight algorithm.

[0262] The computing power / bandwidth / power weights are 0.5 / 0.3 / 0.2 respectively, allocating an additional 15% computing power for logistics tasks. After adjustment, the trajectory deviation is reduced to 0.8 meters.

[0263] Step 5: The LSTM model predicts an average power of 118W for the next 15 minutes (the actual value is 122W, with an error of 3.3%).

[0264] The battery is powered by a "main power supply + supercapacitor" in parallel. Under low load, the supercapacitor takes on 70% of the current, extending the battery life from 45 minutes to 55 minutes.

[0265] Resource allocation weight and power consumption comparison experimental data record table:

[0266]

[0267] Step 6: Simulate a user downloading sensitive route data during non-working hours (23:00), with a transmission volume of 12MB, triggering a Level 1 abnormality alarm;

[0268] Blocking the connection and recording the characteristics to the defense policy library, anomaly detection takes 470ms;

[0269] Safety verification and response time experiment data record table:

[0270] Example 2: A low-altitude economic flight data management system based on edge computing, which is used to implement a low-altitude economic flight data management method based on edge computing, includes the following modules;

[0271] Edge data acquisition and preprocessing module: Through edge computing nodes deployed on aircraft and ground base stations, it collects multi-source data such as device status, communication signals, and environmental parameters in real time, performs denoising and standardization preprocessing, and adopts a ring data queue caching mechanism to manage data in segments according to time windows, and dynamically cleans up redundant data based on cache occupancy;

[0272] Distributed security storage and verification module: Generates a composite biometric code that includes the device hardware fingerprint and user operation characteristics, encrypts the data blocks in fragments using a dynamic chaotic encryption algorithm, and stores them in multiple edge nodes via a distributed hash table. The code is first verified during access, and abnormalities trigger SMS verification or biometric secondary verification.

[0273] Intelligent status assessment and prediction module: Calculates equipment health status coefficients and trend analysis coefficients based on multi-dimensional data, detects abnormal fluctuations using the Z-score and DBSCAN algorithms, generates sub-fluctuation impact factors, triggers fault warnings when thresholds are exceeded, and outputs fault type and remaining time.

[0274] Digital twin-driven resource optimization module: This module builds a lightweight digital twin model to simulate flight trajectory and energy consumption. If the deviation exceeds 5%, an adaptive entropy weight algorithm is triggered to dynamically allocate computing power and bandwidth resources based on task urgency, generating a priority queue.

[0275] Intelligent power management module: Utilizes LSTM neural networks to predict power consumption parameters, coordinates the main power supply and supercapacitors through a virtual power manager, implements a stepped power supply strategy, and combines Bayesian networks to optimize power distribution and extend battery life.

[0276] Abnormal behavior monitoring module: Establishes a user operation baseline through unsupervised learning, monitors access frequency, path depth and other indicators in real time, blocks the connection when anomalies are detected, records the characteristics to the defense strategy library and updates the security model.

[0277] Embodiment 3: A computer device, comprising: a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement a low-altitude economic flight data management method based on edge computing.

[0278] Embodiment 4: A computer-readable storage medium, characterized in that the storage medium stores at least one instruction, at least one program, code set or instruction set, and the at least one instruction, the at least one program, the code set or instruction set is loaded and executed by a processor to implement a low-altitude economic flight data management method based on edge computing.

[0279] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A low-altitude economic flight data management method based on edge computing is characterized by: include: S100: Edge computing nodes deployed on aircraft and ground base stations collect multi-source data from low-altitude flight equipment, including power consumption characteristic data, and pre-process the data in the edge computing nodes. A ring-shaped data queue cache mechanism is used to manage the data in segments, dynamically allocating cache space for each data segment, and performing data storage or cleanup operations based on cache space occupancy. S200: Generate a composite biometric code for the flight data in the cache, encrypt the data blocks in fragments using a chaotic encryption algorithm, and store the encrypted data blocks in a distributed manner on multiple edge nodes. Verify on the edge computing node whether the composite biometric code is consistent with the stored code. If not, initiate a secondary verification mechanism to further confirm the user's identity. S300: Extracts abnormal fluctuation characteristics of flight data and calculates equipment health status coefficients and trend analysis coefficients based on preprocessed data. Simultaneously, it combines historical flight data with multi-dimensional parameters collected in real time to predict the aircraft's future state. If a threshold is exceeded, a fault warning is triggered. S400: Builds a digital twin model based on current flight data and resource allocation plans, and uses a physical simulation engine to predict flight trajectory and energy consumption. If the simulation results deviate from a preset threshold by more than 5%, an adaptive entropy-weighted hybrid optimization algorithm is triggered. It generates a priority queue based on mission urgency and resource weights, dynamically adjusting resource allocation strategies and iteratively adjusting resource allocation parameters. S500: Inputting the power consumption characteristic data of the aircraft into a heterogeneous device management neural network to perform power management prediction, generate device power consumption prediction parameters, and perform power response control on the main power supply and backup power supply of the aircraft through a virtual power manager to generate power regulation parameters; S600: Monitor user operations on aircraft data. When abnormal behavior is detected, record the behavior and store it in the defense policy library for subsequent security policy optimization. At the same time, use the real-time monitoring capabilities of edge computing nodes to quickly respond to abnormal behavior and ensure the security and reliability of aircraft data.

2. The low-altitude economic flight data management method based on edge computing according to claim 1 is characterized in that: The multi-source data acquisition includes: collecting the IMU's three-axis acceleration and angular velocity data at a frequency of 100Hz through the CAN bus, monitoring the RSSI signal strength and RTT transmission delay in the 2.4GHz / 5.8GHz frequency bands in real time through the RTL-SDR radio frequency sensor, and obtaining positioning data with an accuracy of ±1m through the Beidou module; the preprocessing operation includes applying Kalman filtering to the IMU data for denoising, eliminating outliers through the IQR algorithm, and using Z-score normalization to map the data to the [-1,1] interval.

3. The low-altitude economic flight data management method based on edge computing according to claim 1 is characterized in that: The composite biometric code is generated using the SHA-256 algorithm and includes the device MAC address, sensor serial number hardware fingerprint, and 12-dimensional statistical features of the control handle pressure curve. Chaotic encryption uses Logistic mapping, and the parameter μ is dynamically adjusted according to network delay. When the delay is greater than 100ms, μ=3.

9. Data sharding uses Reed-Solomon (5,3) encoding to support 3-node fault tolerance.

4. The low-altitude economic flight data management method based on edge computing according to claim 1 is characterized in that: The health status coefficient H is calculated as follows: H = 0.5 × (S_engine + S_battery + S_att) / 3 + 0.3 × S_c + 0.2 × S_e, where S_engine is the engine status score (0-10 points), S_c is the communication quality score (0-1), S_e is the environmental risk score (0-1), S_battery is the battery status score (0-10 points), and S_att is the attitude sensor status score (0-10 points). When H < 0.6, a yellow warning is triggered.

5. The low-altitude economic flight data management method based on edge computing according to claim 1 is characterized in that: The digital twin model is built using the Unity engine, with a simulation step of 20ms, a trajectory prediction error of ≤1 meter, and an energy consumption prediction error of ≤5%; the adaptive entropy weight algorithm calculates the computing power, bandwidth, and power weights using the entropy method, with the formula ω_j=(1-e_j) / Σ(1-e_j), where e_j is the entropy value of the j-th indicator.

6. The low-altitude economic flight data management method based on edge computing according to claim 1 is characterized in that: The neural network consists of three layers of 64 neurons. The input features are the RMS current, voltage ripple coefficient and task type, which are one-hot encoded. The output power prediction error is ≤5%; the power switching delay is <5ms, and the supercapacitor bears ≥70% of the current under low load.

7. The low-altitude economic flight data management method based on edge computing according to claim 1 is characterized in that: The isolation forest algorithm is used to establish a behavioral baseline for monitoring user operations on aircraft data. The neighborhood radius ε = 0.5, the minimum number of samples MinPts = 3, and an alarm is triggered when the anomaly score s(x) > 0.9; The three-level response mechanism includes: first-level exception logging, second-level exception connection blocking, and third-level exception initiation of system isolation.

8. A low-altitude economic flight data management system based on edge computing, used to implement the method according to any one of claims 1 to 7, characterized in that: Includes the following modules: Edge data acquisition and preprocessing module: Through edge computing nodes deployed on aircraft and ground base stations, it collects multi-source data such as device status, communication signals, and environmental parameters in real time, performs denoising and standardization preprocessing, and adopts a ring data queue caching mechanism to manage data in segments according to time windows and dynamically clean up redundant data based on cache occupancy. Distributed security storage and verification module: Generates a composite biometric code that includes the device hardware fingerprint and user operation characteristics, encrypts the data blocks in fragments using a dynamic chaotic encryption algorithm, and stores them in multiple edge nodes via a distributed hash table. The code is first verified during access, and abnormalities trigger SMS verification or biometric secondary verification. Intelligent status assessment and prediction module: Calculates equipment health status coefficients and trend analysis coefficients based on multi-dimensional data, detects abnormal fluctuations using the Z-score and DBSCAN algorithms, generates sub-fluctuation impact factors, triggers fault warnings when thresholds are exceeded, and outputs fault type and remaining time. Digital twin-driven resource optimization module: This module builds a lightweight digital twin model to simulate flight trajectory and energy consumption. If the deviation exceeds 5%, an adaptive entropy weight algorithm is triggered to dynamically allocate computing power and bandwidth resources based on task urgency, generating a priority queue. Intelligent power management module: Utilizes LSTM neural networks to predict power consumption parameters, coordinates the main power supply and supercapacitors through a virtual power manager, implements a stepped power supply strategy, and combines Bayesian networks to optimize power distribution and extend battery life. Abnormal behavior monitoring module: Establishes a user operation baseline through unsupervised learning, monitors access frequency and path depth indicators in real time, blocks connections when anomalies are detected, records features to the defense strategy library, and updates security models.

9. A computer device, characterized in that: The computer device includes: a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the low-altitude economic flight data management method based on edge computing as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the low-altitude economic flight data management method based on edge computing as described in any one of claims 1 to 7.

Images