MPQUIC-based low-altitude internet-of-intelligence multi-path data transmission system

Abstract

The invention provides a low-altitude networking of intelligence multipath data transmission system based on MPQUIC, and belongs to the technical field of novel network protocols, and the system comprises a network state analysis module which is used for obtaining the network state indexes of all communication paths of an unmanned aerial vehicle under different position coordinates; the flow scheduling module is used for predicting the track of the unmanned aerial vehicle, scoring the communication paths by combining the network state indexes of the communication paths, and obtaining the communication paths corresponding to the data flows with different priorities of the unmanned aerial vehicle according to the priorities of the data flows with the scores of the communication paths; and the path selection optimization module is used for dynamically adjusting weight parameters in communication path scores based on deep reinforcement learning by taking network throughput maximization and time delay minimization as targets to obtain optimal communication paths corresponding to data streams with different priorities. According to the invention, the problem that the prior art is difficult to have high reliability and low time delay due to the dynamic change of the network environment caused by the position change of the unmanned aerial vehicle in the low-altitude Internet of Things is solved.

Description

Technical Field

[0001] This invention belongs to the field of novel network protocol technology, and in particular relates to a low-altitude intelligent network multi-path data transmission system based on MPQUIC. Background Technology

[0002] With the rapid development of drone technology, low-altitude communication networks, as a crucial infrastructure supporting drone communication, monitoring, and positioning, have demonstrated enormous potential in various fields such as agriculture, emergency rescue, logistics, and environmental monitoring. However, the communication environment in which low-altitude drones operate during missions is complex, and the long flight distances and high speeds of drones make network signals susceptible to obstruction, multipath fading, and dynamic interference. Existing technologies often employ multipath protocols such as MPQUIC to address these issues, which is effective; however, MPQUIC deployment in low-altitude intelligent networks still presents the following problems: The real-time nature of network status information is difficult to guarantee: Low-altitude UAV networks rely on network status information, but the high speed and frequent changes in the position of UAVs lead to unstable network status and frequent switching of access points. Under these circumstances, the ACK information fed back by data packets is difficult to provide timely feedback on the real-time network status, while frequent sending of probe packets will consume a lot of communication resources and reduce system efficiency.

[0003] Network status awareness lag: Existing path switching schedulers rely on ACK feedback to obtain the transmission status of paths. However, in low-altitude, high-dynamic network environments, due to the delay in network status feedback, when some paths are idle but have not yet received ACKs, the congestion window cannot be updated in time, resulting in wasted available bandwidth resources and unbalanced load across multiple paths.

[0004] Out-of-order data arrival: During multipath transmission and path switching, the same data stream may be transmitted through multiple paths with different bandwidth and latency characteristics, resulting in data packets arriving at the receiving end out of order. Due to the dependencies between data streams, this out-of-order delivery can cause buffer blockage at the receiving end, affecting the timely delivery of data and thus reducing transmission efficiency and communication quality.

[0005] The path selection scoring criteria fail to accurately reflect path quality: In low-altitude UAV networks, network scenarios are diverse, network states fluctuate greatly, and the influence of different indicators changes dynamically. However, the weight coefficients in existing technologies are set statically by humans, which makes it difficult to adapt to the needs of UAV networks. Therefore, the scoring criteria are difficult to evaluate the true network quality. Summary of the Invention

[0006] To address the aforementioned shortcomings in existing technologies, this invention provides a low-altitude intelligent network multi-path data transmission system based on MPQUIC. This system solves the problems of low data transmission efficiency, wasted bandwidth resources, and low communication quality caused by the existing MPQUIC protocol's inability to adapt to low-altitude UAVs due to network status transmission difficulties, delayed network status perception, out-of-order data transmission, and the inability of path scoring to be applied to low-altitude UAVs.

[0007] To achieve the above objectives, the technical solution adopted by this invention is: a low-altitude intelligent network multipath data transmission system based on MPQUIC, comprising: The network status analysis module is used to obtain network status indicators for each communication path of the UAV at different location coordinates. The flow scheduling module is used to predict the trajectory of the UAV and score the communication path by combining the network status indicators of each communication path. Based on the priority of the data stream with the score of each communication path, the communication path corresponding to the UAV's different priority data streams is obtained. The path selection optimization module is used to dynamically adjust the weight parameters in the communication path scoring based on deep reinforcement learning, with the goal of maximizing network throughput and minimizing latency, to obtain the optimal communication path corresponding to data streams of different priorities.

[0008] Furthermore: the network state analysis module includes: The network status detection unit is used to send and receive detection data packets to each communication path of the UAV when the UAV reaches a new location coordinate, and to obtain the initial network status parameters of each communication path of the UAV. The network status measurement unit is used to calculate the network status detection values ​​of each communication path of the UAV under the new position coordinates based on the probe data packets and a sliding window. The network status detection values ​​include round-trip time, bandwidth and packet loss rate. The network state prediction unit is used to process the new location coordinates, the initial parameters of the network state of each communication path and the network state detection value of each communication path through a trained long short-term memory network model to obtain the estimated value of the network state of each communication path under the new location coordinates. By integrating the initial network status parameters of each communication path, the network status detection values ​​of each communication path, and the network status estimation values ​​of each communication path, the network status index of each communication path is obtained.

[0009] The further beneficial effects mentioned above are as follows: the network status analysis module can measure the network status when the UAV's position changes rapidly. Through active detection, real-time measurement and LSTM prediction, the network status of the UAV can be accurately determined and predicted, which improves the accuracy and timeliness of the network status of the communication path and provides a reliable data foundation for the subsequent scheduling of data streams and communication paths.

[0010] Further: The specific steps of the network state detection unit include: When controlling the drone to fly to a new location coordinate, the available access points are determined by scanning with the multiple network cards carried by the drone; By establishing a one-to-one connection and session registration between the drone and an available access point through multiple network cards carried by the drone, several communication paths of the drone can be obtained. By using multiple network cards carried by the drone, probe data packets containing the drone's coordinates are sent to several communication paths to obtain the initial network status parameters of each communication path of the drone.

[0011] The further beneficial effects mentioned above are: by actively scanning available access points and establishing connections, the network environment initialization and baseline state acquisition for multiple communication paths are achieved. This ensures that the system can quickly sense changes in network status and establish initial connections after the UAV's position changes, providing a stable communication path for subsequent real-time measurement and prediction.

[0012] Further: The specific steps of the network state measurement unit include: The round-trip delay of each communication path is calculated based on the timestamp in the ACK of the probe data packet; The bandwidth is calculated based on QUIC's built-in bottleneck bandwidth and round-trip propagation time algorithm; The packet loss rate is calculated based on the historical number of lost packets and the total number of data packets sent. Based on the sliding window, the round-trip delay, bandwidth and packet loss rate of each communication path are updated using a weighted average to obtain the network status detection value of each communication path. The expression for the weighted average is as follows:

[0013]

[0014]

[0015] in, , and Communication paths The calculated round-trip delay, packet loss rate, and bandwidth at the current moment. , and These are the calculated round-trip delay, packet loss rate, and bandwidth values ​​for the previous time step, respectively. For communication path The round-trip delay of each communication path is calculated based on the timestamp in the ACK of the probe data packet. For communication path The bandwidth is calculated based on QUIC's built-in bottleneck bandwidth and round-trip time algorithm. For communication path The packet loss rate is calculated based on the number of packet losses and the total number of data packets sent within a certain time window. For update rate, .

[0016] The further beneficial effects mentioned above are as follows: the network status measurement unit quickly obtains the round-trip delay, bandwidth and packet loss rate of the communication path through the QUIC protocol, and updates it through the sliding window method. Finally, it obtains a stable and reliable network status detection value that can show the real-time trend of the communication path, providing a data foundation for subsequent communication path scoring, selection and scheduling.

[0017] Furthermore: the flow scheduling module includes: The UAV trajectory prediction unit is used to filter the UAV's position coordinates and velocity based on the Kalman filter algorithm to predict the UAV's short-term motion trajectory. The path scoring and data stream priority calculation unit is used to calculate the score of each communication path based on the short-term motion trajectory and the network status indicators of each communication path, and to determine the priority of each data stream based on the data stream's deadline. The adaptive scheduling unit is used to match the corresponding communication path for each data stream in a round-robin manner according to the priority of the data stream and the score of each communication path. The data stream transmission unit is used to match the corresponding communication path for each data stream and transmit the data stream, and to determine whether the data stream arrives out of order. If so, the data stream is not transmitted using the matching communication path and is marked as a data stream that needs to be retransmitted; otherwise, it is transmitted normally. The retransmission control unit is used to divide the data streams that need to be retransmitted into high-priority data streams and return them to the path scoring and data stream priority calculation unit, so that they can be retransmitted preferentially through the matching communication path in the next scheduling cycle.

[0018] The further beneficial effects mentioned above are as follows: the flow scheduling module can alleviate the problem of communication path performance degradation caused by UAV movement by predicting UAV trajectory, scoring communication paths, and prioritizing data streams. Furthermore, by judging out-of-order arrival and retransmitting, it reduces the impact of out-of-order and packet loss when transmitting data through multiple communication paths, thus ensuring the reliability and real-time performance of high-priority data.

[0019] Furthermore, the expressions for the position coordinates and velocity of the UAV are as follows:

[0020]

[0021]

[0022] in, For drones Position coordinates and velocity vector at any given moment Here is the state transition matrix. For drones Position coordinates and velocity vector at any given moment For process noise, The horizontal coordinate is the coordinate of the space. The vertical coordinate is the coordinate of the space. The vertical coordinate in space, The velocity is the horizontal axis velocity in space. The velocity is the vertical axis of space. For the velocity along the vertical and horizontal axes in space, For time intervals.

[0023] The further beneficial effects mentioned above are: by incorporating the position and velocity of the UAV into a linear discrete form in the form of vectors and matrices, the UAV's continuous motion trajectory can be accurately represented, and the short-term position and velocity state of the UAV can be predicted more accurately based on the Kalman filter algorithm.

[0024] Furthermore, the expression for the score of each communication path is as follows:

[0025]

[0026]

[0027]

[0028] in, For the first Scoring of communication paths , and All are weighted parameters. This is the average round-trip time for all communication paths. This is the average packet loss rate across all communication paths. This is the average bandwidth of all communication paths. No. Round-trip delay of the communication path, No. Packet loss rate of each communication path No. Bandwidth of each communication path The number of communication paths, and These are all indexes of the communication path.

[0029] The priority expressions for each data stream are as follows:

[0030] in, For the first Data stream, For the first The deadline for each data stream. For the current moment, The number of data streams excluding the highest priority data stream. For indexing data streams.

[0031] The further beneficial effects mentioned above are as follows: by scoring communication paths through normalized multi-dimensional communication path network status indicators and calculating data flow priorities through deadlines, the quantification of communication paths and data flows is realized, which facilitates the matching of data flows with the optimal communication paths and improves the overall efficiency of network resource utilization and the timeliness of data transmission.

[0032] Furthermore, the specific steps of the adaptive scheduling unit include: Sort the data streams according to their priority to obtain a data stream queue; The communication paths are sorted according to their scores to obtain a communication path queue. Using a polling approach, the process starts with the highest priority data stream in the data stream queue. It queries all available communication paths for the currently processed data stream, selects the communication path that meets the data stream transmission requirements and has the highest score from the available communication paths, and binds it to the currently processed data stream to obtain the communication path corresponding to the current data stream. Poll all data streams in the data stream queue and match the corresponding communication path for each data stream.

[0033] The further beneficial effects mentioned above are as follows: the adaptive scheduling unit achieves dynamic adaptation of UAV network resources and data transmission requirements by sorting and polling data streams and communication paths, ensuring that high-priority data streams get the optimal transmission communication path first, while taking into account the relative balance of multi-path load, thus improving the overall throughput efficiency of the UAV network and the real-time data transmission guarantee capability.

[0034] Furthermore: the determination of whether the data stream transmission arrives out of order specifically includes:

[0035]

[0036]

[0037] in, For the current communication path number During this data transmission The estimated arrival time of the data stream transmitted at any given time. For the current communication path number During this data transmission The estimated arrival time of the data stream transmitted at any given time. For the round-trip time of the current communication, For the current communication path number The length of the transmission queue during this data transmission. For the current communication path number Bandwidth during this data transmission.

[0038] The further beneficial effects mentioned above are: by judging out-of-order arrival, potential data transmission errors can be identified in the early stages of data stream transmission, and out-of-order delivery can be proactively avoided to prevent unnecessary data retransmission, thereby improving the reliability and efficiency of multi-communication path transmission in UAV networks.

[0039] Furthermore, the expression for the path selection optimization module is as follows:

[0040]

[0041]

[0042] in, For state, For action, As a reward, , and These are the round-trip time, bandwidth, and packet loss rate for the first communication path, respectively. , and These are the round-trip latency, bandwidth, and packet loss rate for the last communication path, respectively. The number of communication paths, , and The weight parameters in the scoring expression for each communication path. For total throughput, The rate of change of total throughput This represents the average transmission delay of all data packets within the window period, based on the ACK statistics. The rate of change of the average time delay As a weight of the total throughput, The weight is the delay.

[0043] The further beneficial effects mentioned above are as follows: Based on state, action and reward, a deep reinforcement learning task can be established, which can learn and dynamically optimize the weight parameters used for frequency division of communication paths according to the real-time network status and historical data of the UAV. Compared with the traditional static fixed weight parameters, it can more accurately allocate network resources for the UAV’s ever-changing network and perform optimal dynamic matching of global network performance, thereby maximizing throughput and minimizing latency.

[0044] The beneficial effects of this invention are: By integrating real-time network state analysis, UAV trajectory Kalman filter prediction, and dynamic weight adjustment through deep reinforcement learning, this invention can pre-match and optimize the data flow and communication path of UAVs based on the real-time changing network state and UAV movement trends, thereby improving transmission efficiency and stability in complex dynamic UAV environments. This invention actively determines out-of-order arrival and combines it with data stream priority obtained based on data stream deadlines to schedule data stream priorities and perform data retransmission. This overcomes the problems of out-of-order arrival and packet loss in multi-communication path transmission, ensuring that high-priority data streams can be transmitted in real time, reliably, and with low latency. The network status analysis module, flow scheduling module, and path selection optimization module of this invention work together to maximize the network throughput of UAVs, minimize data transmission latency, and optimize resource utilization efficiency, thereby enhancing the communication robustness and continuity of low-altitude UAV intelligent networks in scenarios such as emergency rescue and logistics delivery. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of a low-altitude intelligent network multipath data transmission system based on MPQUIC. Detailed Implementation

[0046] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0047] Example 1 like Figure 1The diagram shown is a schematic of a multi-path data transmission system for low-altitude intelligent networks based on MPQUIC. The present invention provides a multi-path data transmission system for low-altitude intelligent networks based on MPQUIC, comprising: The network status analysis module is used to obtain network status indicators for each communication path of the UAV at different location coordinates. The flow scheduling module is used to predict the trajectory of the UAV and score the communication path by combining the network status indicators of each communication path. Based on the priority of the data stream with the score of each communication path, the communication path corresponding to the UAV's different priority data streams is obtained. The path selection optimization module is used to dynamically adjust the weight parameters in the communication path scoring based on deep reinforcement learning, with the goal of maximizing network throughput and minimizing latency, to obtain the optimal communication path corresponding to data streams of different priorities.

[0048] To address the challenges of high-speed UAV movement in low-altitude intelligent networks (LAHs) leading to dynamic changes in the network environment, significant performance differences across multiple communication paths, and the difficulty of traditional transmission protocols simultaneously achieving high reliability and low latency, this invention proposes a multi-path data transmission system for LHAs based on MPQUIC. By using a network state analysis module to monitor and predict the performance indicators of each communication path of the UAV in real time, and combining UAV trajectory prediction and adaptive data flow and multi-communication path scheduling in the flow scheduling module, intelligent dynamic matching of data flow and communication paths is achieved. This effectively reduces data transmission interruptions and out-of-order data arrival caused by UAV communication node movement and network jitter. Furthermore, through a path selection optimization module, deep reinforcement learning is used to adjust the weight parameters in the communication path scoring, enabling the LHA to continuously optimize with the goal of maximizing throughput and minimizing latency. This invention improves the reliability, real-time performance, and network resource utilization efficiency of data transmission across multiple communication paths for UAVs, enhancing the robustness and continuity of communication in scenarios such as emergency rescue and logistics delivery.

[0049] In one embodiment of the present invention, addressing the problem of insufficient accuracy of single communication measurement in low-altitude dynamic networks of unmanned aerial vehicles (UAVs) due to rapid changes in communication path status caused by UAV position changes, the present invention actively detects and acquires initial parameters through a network state analysis module. Combined with real-time sliding window measurement and LSTM network prediction, network state measurement and prediction are performed, achieving accurate perception of UAV communication network performance. This provides a reliable and comprehensive basis for scheduling multiple communication paths and improves the system's adaptability to the dynamic environment of UAVs. The network state analysis module includes: a network state detection unit, a network state measurement unit, and a network state prediction unit. In a specific embodiment of the present invention, the network status detection unit is used to send and receive detection data packets to each communication path of the UAV when the UAV reaches a new location coordinate, and to obtain the initial network status parameters of each communication path of the UAV; the specific steps of the network status detection unit include: When controlling the drone to fly to a new location coordinate, the drone uses multiple network cards to scan and determine available access points, including communication base stations, other drones, satellites, etc. By establishing a one-to-one connection and session registration between the drone and an available access point through multiple network cards carried by the drone, several communication paths of the drone can be obtained. At the same time, the IP address and signal strength information of the available communication paths can be recorded. By using multiple network cards carried by the drone, probe data packets containing the drone's coordinates are sent to several communication paths to obtain the initial network status parameters of each communication path of the drone.

[0050] In a specific embodiment of the present invention, the network state measurement unit is used to calculate the network state detection values ​​of each communication path of the UAV under the new position coordinates based on the probe data packets and a sliding window. The network state detection values ​​include round-trip time, bandwidth, and packet loss rate. The specific steps of the network state measurement unit include: The round-trip delay of each communication path is calculated based on the timestamp in the ACK in the probe data packet; ACK stands for Acknowledge character, which is a transmission control character sent by the receiving station to the sending station in data communication to indicate that the data has been confirmed to have been received without error.

[0051] The bandwidth is calculated based on the bottleneck bandwidth and round-trip time algorithm built into QUIC. Specifically, the bandwidth can be calculated using the number of bytes sent and the time difference between the received ACK within the historical time window. The packet loss rate is calculated based on the historical number of lost packets and the total number of data packets sent. Based on the sliding window, the round-trip delay, bandwidth and packet loss rate of each communication path are updated using a weighted average to obtain the network status detection value of each communication path. The expression for the weighted average is as follows:

[0052]

[0053]

[0054] in, , and Communication paths The calculated round-trip delay, packet loss rate, and bandwidth at the current moment. , and These are the calculated round-trip delay, packet loss rate, and bandwidth values ​​for the previous time step, respectively. For communication path The round-trip delay of each communication path is calculated based on the timestamp in the ACK of the probe data packet. For communication path The bandwidth is calculated based on QUIC's built-in bottleneck bandwidth and round-trip time algorithm. For communication path The packet loss rate is calculated based on the number of packet losses and the total number of data packets sent within a certain time window. For update rate, .

[0055] In a specific embodiment of the present invention, the network state prediction unit is used to process the new location coordinates, the initial parameters of the network state of each communication path and the network state detection value of each communication path through a trained long short-term memory network model to obtain the estimated value of the network state of each communication path under the new location coordinates. During the training phase, the trained Long Short-Term Memory (LSTM) network model takes the UAV's own coordinate information vector as input and uses the historically stored real network state as the supervision target to train the parameters of the LSM network model. This gradually establishes a nonlinear mapping relationship between the UAV's position and network performance, ultimately resulting in a trained LSM network model.

[0056] By integrating the initial network status parameters of each communication path, the network status detection values ​​of each communication path, and the network status estimation values ​​of each communication path, the network status index of each communication path is obtained.

[0057] In one embodiment of the present invention, addressing the problems of high mobility of UAVs in low-altitude intelligent networks, easy data out-of-order transmission during multi-path transmission, and difficulty in matching data stream priorities, resulting in low data transmission efficiency, the flow scheduling module of the present invention integrates UAV future trajectory prediction, communication path scoring, and data stream priority scheduling, and combines out-of-order arrival judgment and data stream retransmission mechanism to achieve dynamic matching and recovery of data streams and communication paths. This improves the transmission reliability of high-priority data streams in low-altitude intelligent networks, maximizes the throughput and minimizes the latency of UAV communication networks, and ensures the reliability and real-time performance of UAV communication tasks. The flow scheduling module includes: a UAV trajectory prediction unit, a path scoring and data stream priority calculation unit, an adaptive scheduling unit, a data stream transmission unit, and a retransmission control unit.

[0058] In a specific embodiment of the present invention, the UAV trajectory prediction unit is used to filter the UAV's position coordinates and velocity based on the Kalman filter algorithm to predict the UAV's short-term motion trajectory; wherein, the expressions for the UAV's position coordinates and velocity are as follows:

[0059]

[0060]

[0061] in, For drones Position coordinates and velocity vector at any given moment Here is the state transition matrix. For drones Position coordinates and velocity vector at any given moment For process noise, The horizontal coordinate is the coordinate of the space. The vertical coordinate is the coordinate of the space. The vertical coordinate in space, The velocity is the horizontal axis velocity in space. The velocity is the vertical axis of space. For the velocity along the vertical and horizontal axes in space, The time interval is typically 0.1 to 5 seconds. By unifying the position and velocity of the UAV in the form of vectors and matrices and performing linear discretization, the continuous motion trajectory of the UAV can be accurately represented, and more accurate prediction results of the short-term position and velocity status of the UAV can be obtained.

[0062] In a specific embodiment of the present invention, the path scoring and data flow priority calculation unit is used to calculate the score of each communication path based on the short-term motion trajectory and network status indicators of each communication path, and to determine the priority of each data flow based on the data flow deadline. Through path scoring and dynamic priority judgment of data flows based on deadlines, the round-trip latency, bandwidth, and packet loss rate of each communication path are comprehensively evaluated, and the urgency of the data flow is quantified. This achieves accurate matching between communication path performance and data flow requirements, improves the scientific nature of communication path selection, and optimizes network throughput and real-time performance. The expression for the score of each communication path is as follows:

[0063]

[0064]

[0065]

[0066] in, For the first Scoring of communication paths , and All are weighted parameters. This is the average round-trip time for all communication paths. This is the average packet loss rate across all communication paths. This is the average bandwidth of all communication paths. No. Round-trip delay of the communication path, No. Packet loss rate of each communication path No. Bandwidth of each communication path The number of communication paths, and All are indexes of the communication path; The priority expressions for each data stream are as follows:

[0067] in, For the first Data stream, For the first The deadline for each data stream. For the current moment, The number of data streams excluding the highest priority data stream. For indexing data streams.

[0068] In a specific embodiment of the present invention, the adaptive scheduling unit is used to match a corresponding communication path for each data stream in a round-robin manner according to the priority order of the data streams and the score of each communication path; the specific steps of the adaptive scheduling unit include: Sort the data streams according to their priority to obtain a data stream queue; The communication paths are sorted according to their scores to obtain a communication path queue. Using a polling approach, the process starts with the highest priority data stream in the data stream queue. It queries all available communication paths for the currently processed data stream, selects the communication path that meets the data stream transmission requirements and has the highest score from the available communication paths, and binds it to the currently processed data stream to obtain the communication path corresponding to the current data stream. Poll all data streams in the data stream queue and match the corresponding communication path for each data stream; after completing one poll, if no data stream is matched, recalculate the priority and start the next poll.

[0069] The adaptive scheduling unit of this invention prioritizes the selection of the highest-scoring available communication path for high-priority data streams through data stream sorting, communication path sorting, and round-robin matching, thereby achieving optimal matching between network resources and transmission requirements. This ensures the transmission quality of critical data streams while also balancing the communication task load among multiple communication paths, improving the overall resource utilization efficiency and real-time data transmission performance of the system.

[0070] In a specific embodiment of the present invention, the data stream transmission unit is configured to transmit the data stream according to the corresponding communication path matched for each data stream, and to determine whether the data stream transmission is out of order. If so, the data stream is not transmitted using the corresponding communication path, and the data stream is marked as a data stream that needs to be retransmitted; otherwise, normal transmission is performed. Specifically, determining whether the data stream transmission is out of order includes:

[0071]

[0072]

[0073] in, For the current communication path number During this data transmission The estimated arrival time of the data stream transmitted at any given time. For the current communication path number During this data transmission The estimated arrival time of the data stream transmitted at any given time. For the round-trip time of the current communication, For the current communication path number The length of the transmission queue during this data transmission. For the current communication path number Bandwidth during this data transmission.

[0074] In a specific embodiment of the present invention, the retransmission control unit is used to divide the data streams that need to be retransmitted in the data stream into high-priority data streams and return them to the path scoring and data stream priority calculation unit, so that they can be retransmitted preferentially through the matching communication path in the next scheduling cycle.

[0075] In one embodiment of the present invention, addressing the issue that in low-altitude intelligent networks, the fixed weights in traditional communication path scoring are ill-suited to the dynamic network environment due to real-time changes in UAV location, resulting in communication path scores that do not conform to the real-time network status of the UAV, the present invention utilizes a deep reinforcement learning module to model the weight parameters as a problem of continuous parameter optimization through deep learning. Based on real-time network status and communication path performance feedback, the optimal weight parameters are autonomously learned. With the goal of maximizing UAV network throughput and minimizing data transmission latency, the global optimization of communication path selection for different priority data streams is achieved, improving the system's adaptability and overall transmission performance in dynamic environments. The expression for the path selection optimization module is as follows:

[0076]

[0077]

[0078] in, For state, For action, As a reward, , and These are the round-trip time, bandwidth, and packet loss rate for the first communication path, respectively. , and These are the round-trip latency, bandwidth, and packet loss rate for the last communication path, respectively. The number of communication paths, , and The weight parameters in the scoring expression for each communication path. For total throughput, The rate of change of total throughput This represents the average transmission delay of all data packets within the window period, based on the ACK statistics. The rate of change of the average time delay As a weight of the total throughput, Weighting for latency; Rewards The specific expression can be determined by the optimization goals set by the user, and the corresponding throughput and latency weights can be selected according to different business requirements. and .

[0079] The beneficial effects of this invention are: By integrating real-time network state analysis, UAV trajectory Kalman filter prediction, and dynamic weight adjustment through deep reinforcement learning, this invention can pre-match and optimize the data flow and communication path of UAVs based on the real-time changing network state and UAV movement trends, thereby improving transmission efficiency and stability in complex dynamic UAV environments. This invention actively determines out-of-order arrival and combines it with data stream priority obtained based on data stream deadlines to schedule data stream priorities and perform data retransmission. This overcomes the problems of out-of-order arrival and packet loss in multi-communication path transmission, ensuring that high-priority data streams can be transmitted in real time, reliably, and with low latency. The network status analysis module, flow scheduling module, and path selection optimization module of this invention work together to maximize the network throughput of UAVs, minimize data transmission latency, and optimize resource utilization efficiency, thereby enhancing the communication robustness and continuity of low-altitude UAV intelligent networks in scenarios such as emergency rescue and logistics delivery.

Claims

1. A low-altitude intelligent networking multi-path data transmission system based on MPQUIC, characterized in that, The application relates to a network state analysis module for acquiring network state indexes of each communication path of a UAV at different position coordinates. A flow scheduling module is used for predicting a UAV trajectory, combining network state indexes of each communication path to score the communication paths, and obtaining communication paths corresponding to different priority data streams of the UAV according to the priority of the data streams. A path selection optimization module is used for dynamically adjusting weight parameters in the communication path score based on deep reinforcement learning, with the goal of maximizing network throughput and minimizing time delay, to obtain optimal communication paths corresponding to different priority data streams. The network state analysis module comprises a network state detection unit for sending and receiving detection data packets to each communication path of the UAV when the UAV reaches a new position coordinate, and acquiring initial network state parameters of each communication path of the UAV.

2. The MPQUIC-based low-altitude cognitive networking multipath data transmission system of claim 1, wherein, A network state measurement unit is used for calculating network state detection values of each communication path of the UAV at the new position coordinate based on a sliding window according to the detection data packets, wherein the network state detection values include round-trip time delay, bandwidth and packet loss rate. A network state prediction unit is used for processing the new position coordinate, the initial network state parameters of each communication path and the network state detection values of each communication path through a trained long short-term memory network model to obtain network state estimation values of each communication path at the new position coordinate. The initial network state parameters of each communication path, the network state detection values of each communication path and the network state estimation values of each communication path are integrated to obtain network state indexes of each communication path. The specific steps of the network state detection unit comprise controlling the UAV to fly to a new position coordinate, determining available access points through a multi-network card carried by the UAV, establishing one-to-one connection and session registration between the UAV and the available access points through the multi-network card carried by the UAV to obtain a plurality of communication paths of the UAV, 3. The MPQUIC-based low-altitude cognitive networking multipath data transmission system of claim 2, wherein, sending detection data packets containing the coordinates of the UAV to the plurality of communication paths through the multi-network card carried by the UAV to obtain the initial network state parameters of each communication path of the UAV. The specific steps of the network state measurement unit comprise calculating the round-trip time delay of each communication path according to the time stamp in the ACK in the detection data packet, calculating the bandwidth according to the built-in bottleneck bandwidth and round-trip propagation time algorithm of QUIC, calculating the packet loss rate according to the historical number of lost packets and the total number of sent data packets, 4. The MPQUIC-based low-altitude cognitive networking multipath data transmission system of claim 2, wherein, updating the round-trip time delay, bandwidth and packet loss rate of each communication path by weighted average according to the sliding window to obtain the network state detection values of each communication path, and the expression of the weighted average is as follows: The flow scheduling module comprises a UAV trajectory prediction unit for filtering the position coordinates and speed of the UAV based on a Kalman filtering algorithm to predict a short-term motion trajectory of the UAV, a path score and data stream priority calculation unit for calculating the score of each communication path according to the short-term motion trajectory and the network state indexes of each communication path, and determining the priority of each data stream according to the deadline of the data stream. ​ ​ in, , and Communication paths The calculated round-trip delay, packet loss rate, and bandwidth at the current moment. , and These are the calculated round-trip delay, packet loss rate, and bandwidth values ​​for the previous time step, respectively. For communication path The round-trip delay of each communication path is calculated based on the timestamp in the ACK of the probe data packet. For communication path The bandwidth is calculated based on QUIC's built-in bottleneck bandwidth and round-trip time algorithm. For communication path The packet loss rate is calculated based on the number of packet losses and the total number of data packets sent within a certain time window. For update rate, .

5. The MPQUIC-based low-altitude cognitive networking multipath data transmission system of claim 1, wherein, ​ ​ ​ An adaptive scheduling unit is configured to match each data stream with a corresponding communication path in a polling manner according to the data stream priority order and the score of each communication path; A data stream transmission unit is configured to perform data stream transmission according to the matched communication path of each data stream, and determine whether the data stream transmission is out of order, if yes, the data stream is not transmitted through the matched communication path, and the data stream is marked as a data stream needing retransmission, otherwise, the data stream is normally transmitted; A retransmission control unit is configured to divide the data stream needing retransmission into high-priority data streams, and return to the path score and data stream priority calculation unit to be retransmitted through the matched communication path in the next scheduling cycle.

6. The MPQUIC-based low-altitude cognitive networking multipath data transmission system of claim 5, wherein, Expressions of the position coordinates and the speed of the unmanned aerial vehicle are as follows: wherein, is the position coordinate and velocity vector of the UAV at the is the position coordinate and velocity vector of the UAV at the is the state transition matrix, is the position coordinate and velocity vector of the UAV at the is the position coordinate and velocity vector of the UAV at the is the process noise, is the spatial horizontal coordinate, is the spatial vertical coordinate, is the spatial vertical coordinate, is the spatial horizontal axis velocity, is the spatial vertical axis velocity, is the spatial vertical axis velocity, is the time interval.

7. The MPQUIC-based low-altitude cognitive networking multipath data transmission system of claim 5, wherein, Expressions of the scores of the communication paths are as follows: in, For the first Scoring of communication paths , and All are weighted parameters. This is the average round-trip time for all communication paths. This is the average packet loss rate across all communication paths. This is the average bandwidth of all communication paths. No. Round-trip delay of the communication path, No. Packet loss rate of each communication path No. Bandwidth of each communication path The number of communication paths, and These are all indexes of the communication path. Expressions of the priorities of the data streams are as follows: wherein, is the first data flow, is the deadline of the first data flow, is the current time, is the number of data flows other than the highest priority data flow, is the index of the data flow.

8. The MPQUIC-based low-altitude cognitive networking multipath data transmission system of claim 5, wherein, Specific steps of the adaptive scheduling unit include: Data streams are sorted according to the data stream priority order to obtain a data stream queue; The communication paths are sorted according to the score of each communication path to obtain a communication path queue; In a polling manner, the highest-priority data stream in the data stream queue is processed, all available communication paths of the current processed data stream are queried, the communication path meeting the data stream transmission requirement and having the highest score is selected from the available communication paths, and the current data stream is bound to the selected communication path to obtain the corresponding communication path of the current data stream; All data streams in the data stream queue are polled to match each data stream with a corresponding communication path.

9. The MPQUIC-based low-altitude intelligent networking multipath data transmission system according to claim 5, wherein, The determination of whether the data stream transmission is out of order includes: in, For the current communication path number During this data transmission The estimated arrival time of the data stream transmitted at any given time. For the current communication path number During this data transmission The estimated arrival time of the data stream transmitted at any given time. For the round-trip time of the current communication, For the current communication path number The length of the transmission queue during this data transmission. For the current communication path number Bandwidth during this data transmission.

10. The MPQUIC-based low-altitude cognitive networking multipath data transmission system of claim 1, wherein, An expression of the path selection optimization module is as follows: wherein, is the state, is the action, is the reward, , and are the round-trip delay, bandwidth and packet loss rate of the first communication path, respectively, , and are the round-trip delay, bandwidth and packet loss rate of the last communication path, respectively, is the number of communication paths, , and are the weight parameters in the expression of the score of each communication path, is the total throughput, is the rate of change of the total throughput, is the average sending delay of all data packets according to ACK statistics within the window time, is the rate of change of the average delay, is the weight of the total throughput, is the weight of the delay.

Images