Self-adaptive time synchronization method for unmanned aerial vehicle network

Abstract

The invention discloses a self-adaptive time synchronization method for an unmanned aerial vehicle network, and relates to the technical field of time synchronization of an unmanned aerial vehicle self-organizing network. S2, a node network access and neighbor discovery step; s3, a main node selection step: calculating a main node score weight by each node according to the neighbor information table, and selecting a main node; s4, a centralized time synchronization step, wherein the child nodes carry out time synchronization according to the broadcast of the main node; s5, a TDMA time slot distribution and frame control step; s6, a distributed synchronization step; the child nodes perform time synchronization according to the broadcast of the child nodes; s7, a synchronization quality evaluation and backspacing switching step: when a synchronization backspacing condition is satisfied, performing S4; and when the main node reselection condition is met, S3 is carried out, and the current main node does not participate in main node selection. According to the method, a TDMA mechanism and self-adaptive switching of centralized synchronization and distributed synchronization are adopted, the node collision probability and energy consumption are reduced, and error accumulation of a single synchronization mode is avoided.

Description

Technical Field

[0001] This invention relates to time synchronization technology for unmanned aerial vehicle (UAV) self-organizing networks, and more particularly to an adaptive time synchronization method for UAV networks. Background Technology

[0002] Flying Ad-hoc Networks (FANETs), as a highly mobile and dynamic form of wireless ad-hoc network, have network topologies that frequently change with flight and mission formations, and are frequently used for tasks such as reconnaissance and surveillance, coordinated strikes, emergency rescue, and disaster monitoring. In these complex tasks, time synchronization is crucial for the synergistic effects of UAV swarms. Time synchronization in Flying Ad-hoc Networks is typically constrained by the following: First, certain tasks with specific requirements are subject to engineering resource constraints such as payload, power consumption, and cost, requiring network-related synchronization algorithms to be easy to implement, low-overhead, and low-energy-consumption. Second, UAV deployments constantly change with mission requirements, and the hardware circuits between nodes differ, inevitably leading to time deviations due to environmental interference and manufacturing processes; these errors accumulate with multi-hop propagation. Third, UAVs are usually in motion, and high-speed movement causes changes in propagation conditions and link delay fluctuations, such as the Doppler effect, which amplifies synchronization errors. Time synchronization in UAV networks is one of the key factors in Flying Ad-hoc Networks and is fundamental to ensuring the normal operation of UAV networks in various applications. Therefore, high precision, high dynamic adaptability, and low energy consumption are required for time synchronization between drone nodes in drone networks.

[0003] Currently, time synchronization in UAV networks is mainly divided into two categories: one is centralized synchronization, which relies on satellite time synchronization or ground equipment or UAV reference nodes with high time synchronization accuracy to reduce the impact of uncertainties at the transmitting end; the other is distributed synchronization, which does not rely on external clock sources and uses the method of sharing and broadcasting clocks within the cluster to achieve time synchronization within the region, which is beneficial to improving the adaptability of UAVs to dynamic networks.

[0004] Traditional centralized time synchronization methods, such as Reference Broadcast Synchronization (RBS), primarily rely on network reference nodes broadcasting data packets without local timestamps. The receiving end compares the arrival times multiple times to achieve local time alignment, enabling multi-hop synchronization across regions, but accuracy gradually diminishes. While this receiver-to-receiver synchronization method eliminates sender-end uncertainty, it is still susceptible to uncertainty caused by message delays, significantly impacting synchronization accuracy. The greater the uncertainty in message delays, the larger the resulting time error. The Timing-Sync Protocol for Sensor Networks (TPSN) establishes a hierarchical topology with bidirectional handshakes between parent and child nodes, progressively aligning the entire network clock to the reference root node. It specifically proposes handshakes at MAC layer interruption or completion events to minimize transmission latency. However, due to its strict reliance on the hierarchical structure, multi-hop asymmetric delays and progressive propagation lead to error accumulation, and topology maintenance incurs additional overhead. The Flooding Time Synchronization Protocol (FTSP) selects the node with the smallest ID as the root node and periodically broadcasts synchronization data packets. Receiving nodes use linear regression to estimate clock drift for continuous correction, update their local time, and broadcast again to achieve flood synchronization. Its use of bidirectional switching and on-network node monitoring results in high synchronization accuracy and robustness. However, in highly dynamic networks, frequent node entry and exit triggers node monitoring, significantly increasing communication overhead and energy consumption. The Delay Measurement Time Synchronization (DMTS) algorithm, similar to RBS, is a master-slave unidirectional time synchronization algorithm. Receiving nodes reduce transmission errors by estimating all delays along the synchronization message transmission path. Unidirectional broadcasting only requires a single broadcast to synchronize multiple receivers simultaneously, avoiding the additional overhead of round-trip delay estimation in NTP and TPSN bidirectional synchronization. Traditional distributed time synchronization methods, such as the Time-Diffusion Synchronization Protocol (TDP), treat network-wide time consistency as a diffusion process, gradually spreading the reference time within the network through periodic broadcasts. Gradient Time Synchronization Protocol (GTSP) is a typical neighbor-swapping distributed synchronization protocol. All nodes only exchange information with their local neighbors to adjust their local logical clocks, resulting in smaller nearest neighbor errors and larger far neighbor error tolerances, making it suitable for multi-hop networks.Average Time Synchronization (ATS) is a representative algorithm for average time synchronization. It iteratively averages the node clock compensation parameters through neighbor information exchange, gradually bringing the network node parameters closer together. However, because it relies on iterative updates, its convergence speed is slow, leading to prolonged synchronization time for large networks. Consistent Clock Synchronization (CCS) adds a broadcast weight parameter to ATS. Each successful transmission by a node increments the weight to indicate the number of time broadcasts; more broadcasts mean wider coverage, eventually converging to a time approaching that of a single node. Since CCS also relies on iterative updates, it suffers from the same problems as ATS. All these single-function algorithms, under conditions of limited performance, energy constraints, and high dynamics for UAVs, suffer from difficulties in maintaining synchronization accuracy over long periods, high synchronization overhead, and high energy consumption.

[0005] In recent years, in order to address the problem of poor adaptability of single algorithms, the comprehensive consideration of centralized and distributed synchronization methods has received widespread attention. J. Wu et al. [J. Wu, L. Zhang, Y. Bai and Y. Sun, "Cluster-Based Consensus Time Synchronization for Wireless Sensor Networks," in IEEE SensorsJournal, vol. 15, no. 3, pp. 1404-1413, March 2015, doi: 10.1109 / JSEN.2014.2363471.] divided the network into multiple clusters and introduced the concept of local center using the cluster structure. The cluster head, i.e., the reference node, is responsible for the unified information aggregation within the cluster and shares time within the cluster head. At the same time, nodes gradually approach a consistent clock through neighbor interactions, forming a synchronization method in which the cluster head is centralized and the cluster is distributed. P. Jia et al. [P. Jia, X. Wang and K. Zheng, "Distributed Clock Synchronization Based on Intelligent Clustering in LocalArea Industrial IoT Systems," in IEEE Transactions on Industrial Informatics, vol. 16, no. 6, pp. 3697-3707, June 2020, doi: 10.1109 / TII.2019.2937331.] built upon the cluster-head-centric distributed synchronization method within clusters by using clustering to structure topology and link uncertainties. They then performed distributed synchronization within clusters on the structured neighbors. Furthermore, clustering localized link fluctuations, avoiding the frequent retransmissions and increased energy consumption associated with network-wide synchronization methods.M. Aslam et al. [M. Aslam, W. Liu, X. Jiao, J. Haxhibeqiri, G. Miranda, J. Hoebeke, J. Marquez-Barja and I. Moerman, "Hardware Efficient Clock Synchronization Across Wi-Fi and Ethernet-Based Network Using PTP," in IEEE Transactions on Industrial Informatics, vol. 18, no. 6, pp. 3808-3819, Jun. 2022, doi: 10.1109 / TII.2021.3120005.] proposed a hybrid network hierarchical synchronization method, where the upper layer performs centralized synchronization in a master-slave manner, and the lower layer performs distributed synchronization by improving local compensation strategies to address transmission uncertainties.

[0006] Existing centralized synchronization relies on a single reference node, resulting in uneven energy consumption and poor robustness. Distributed synchronization suffers from problems such as multi-hop error accumulation, susceptibility to crystal drift and external interference, and lacks adaptive control mechanisms for synchronization quality degradation and dynamic topology issues.

[0007] In summary, existing methods and technologies struggle to achieve high accuracy, low overhead, and low complexity in the performance- and energy-constrained, dynamically changing unmanned aerial vehicle (UAV) networks. Therefore, there is an urgent need for a time synchronization method capable of adaptively selecting synchronization strategies, synchronization accuracy, and communication overhead under different network conditions. Summary of the Invention

[0008] To address the above issues, this invention proposes an adaptive time synchronization method for UAV networks. The method comprehensively considers parameters such as energy, location, and speed through a master node selection mechanism, ensuring the reliability, stability, and persistence of the selected master node. The joint design of TDMA and time synchronization significantly reduces node collision probability, energy consumption, and additional communication overhead. Adaptive switching between centralized and distributed synchronization avoids error accumulation caused by external interference and drift mutations in a single synchronization method. The master node reselection mechanism prevents single-point failures and improves the long-term operational stability of the network.

[0009] An adaptive time synchronization method for unmanned aerial vehicle (UAV) networks, comprising the following steps:

[0010] S1, Unmanned Aerial Vehicle (UAV) network node parameter initialization steps;

[0011] S2, Node Networking and Neighbor Discovery Steps: Each drone node sends a HELLO data packet, and each node updates the neighbor information in the neighbor information table with the received HELLO data packet.

[0012] S3, Master Node Selection Steps: Each node calculates the master node score weight based on the neighbor information table and broadcasts a score data packet containing the master node score weight; Each node updates the master node score weight in the neighbor information table with the received score data packet and selects the node with the highest master node score weight in the neighbor information table as the master node candidate; Each node broadcasts a consistency confirmation data packet containing the master node candidate; Each node counts the number of votes for the master node candidate in the neighbor information table with the received consistency confirmation data packet, sets the master node candidate with the most votes as the master node, and the remaining nodes as child nodes;

[0013] S4, centralized time synchronization step: the master node broadcasts a reference data packet including a timestamp, and the child nodes receive the reference data packet and synchronize their time with the master node;

[0014] S5, TDMA time slot allocation and frame control steps: The master node broadcasts frame data packets in the time slot at the beginning of the TDMA frame. Each child node receives the frame data packets, updates its local frame parameters, and performs frame calibration. The child nodes are only awakened in the time slot at the beginning of the frame and in the allocated time slot, and are in a listening state during the rest of the time period.

[0015] S6, Distributed synchronization step: Child nodes broadcast distributed synchronization data packets including a consistent logical clock in their allocated time slots; each child node receives the distributed synchronization data packets and updates its consistent logical clock.

[0016] S7, Synchronization Quality Assessment and Rollback Switching Steps: The master node receives the distributed synchronization data packets from the child node and calculates the synchronization residual of the clock deviation between the child node and the master node; when the synchronization rollback condition based on the synchronization residual is met, proceed to S4; when the master node reselection condition is met, proceed to S3 and the current master node does not participate in the master node selection.

[0017] Preferably, the master node scoring weight is represented as follows:

[0018] ;

[0019] ;

[0020] in, Represents a node The main node's scoring weight; Represents a node Connectivity is used to measure the percentage of a node's neighbors; Represents a node The energy factor is based on the proportion of remaining energy. Represents a node The mobility stability; Represents a node The proximity relationship measures the average distance between a node and its neighbors; Represents a node Link stability is based on packet reception rate; , , , and Indicates the weighting coefficient; Indicates upper limit truncation; This represents the link expiration time, and its value is the link boundary condition. The smallest positive root or , Represents a node With nodes The relative position, Represents a node With nodes relative velocity, Indicates the communication radius of the drone node. Indicates time, Indicates modulo; Represents a node Neighbor information table; This indicates taking the minimum value.

[0021] Preferably, step S3 further includes: eliminating ambiguity using deterministic decision rules; specifically:

[0022] When the primary node score weights of two nodes are less than the preset score difference threshold, they are compared in order of priority: connectivity, energy factor, and link stability, and the node with the higher score is selected as the primary node candidate.

[0023] Preferably, the child node receives the reference data packet and synchronizes its time with the master node, as follows:

[0024] Calculate child nodes The time phase deviation from the master node is expressed as:

[0025] ;

[0026] in, Indicates time phase deviation; This represents the one-way transmission delay compensation term; Represents child nodes The local timestamp at the moment the master node's TIMESYN_HEAD data packet is received. ; This indicates the first broadcast by the master node. Round the local timestamp;

[0027] child nodes For continuous The final offset is obtained by averaging the time phase deviation of the secondary reference broadcast.

[0028] Updating child nodes using the final offset is represented as:

[0029] ;

[0030] ;

[0031] in, Indicates the updated node Time offset compensation parameters; Represents a node Time offset compensation parameters; Indicates the final offset; Indicates the offset correction step size; Indicates the updated child nodes Output time; Represents child nodes The output time.

[0032] Preferably, the frame parameters of the Time Division Multiple Access (TDMA) are expressed as follows: ;in, Indicates the transmission frame period; Indicates the number of time slots per frame; Indicates the time slot length; Indicates the protection interval; This indicates the frame tail maintenance window; child nodes are scheduled to arrive before the allocated time slot. Time-based wake-up, delayed after the time slot ends. The node turns off transmission during the designated time slot and enters a low-power sleep state or performs other information transmission tasks during the remaining time slot, while the other nodes remain in the receiving state during this time slot.

[0033] Preferably, the frame tail maintenance window is used to receive broadcast signals for child node joining and leaving. Before joining the network, a newly joining child node listens to the master node's frame control data packets to synchronize the current frame status and time. Then, within the frame tail maintenance window of the current frame, it sends a network joining request REQ_JOIN using the CSMA / CA mechanism. After receiving the request, the master node updates the frame scheduling version and frame time slot mapping. A child node preparing to leave sends a network leaving request REQ_LEAVE within the frame tail maintenance window. If the master node does not hear a broadcast from a child node in the corresponding time slot within a consecutive frame period that reaches the off-network threshold frame number, it marks the child node as off-network. If no recovery broadcast is heard within the off-network recovery threshold frame period, the master node updates the frame time slot table and releases the time slot.

[0034] Preferably, the distributed time synchronization step is as follows:

[0035] When node Upon receiving the node When broadcasting distributed synchronization data packets, the local hardware clock at the time of reception is recorded and the local logical clock at the time of reception is calculated. Based on this, the node... Update its receive time logic clock value, expressed as:

[0036] ;

[0037] in, Indicates the updated node Receive time local logic clock; Represents a node Receive time local logic clock; and Representing nodes respectively and nodes The confidence level is initially set to 1, and is incremented by 1 after each successful parameter update for the corresponding node. Represents a node Broadcast transmission to nodes One-way transmission delay compensation term; Represents a node The transmission time logic clock;

[0038] node Updating the consistent logical clock is represented as:

[0039] ;

[0040] ;

[0041] ;

[0042] in, Represents a node Consistent logical clock; Indicates the updated node Clock frequency compensation parameters; Represents a node Clock frequency compensation parameters; Indicates the local hardware clock; Indicates the updated node Clock skew compensation parameters; Represents a node Clock skew compensation parameters; Indicates the first End of round node Updated logical clock value.

[0043] Preferably, the synchronization residual is expressed as:

[0044] ;

[0045] in, Represents a node The synchronization residual with the master node; Represents a node One-way transmission delay compensation item for broadcast transmission to the master node; Indicates the first wheel; Represents a node The logical clock at the time of transmission; The logical clock that indicates the moment the master node receives the data.

[0046] Preferably, the synchronization backoff condition based on synchronization residuals specifically meets the following conditions: the distributed time synchronization operation requires no less than the minimum number of resident frames and satisfies any one of the following conditions:

[0047] ;

[0048] ;

[0049] ;

[0050] ;

[0051] in, This represents the residual variance of all child nodes; This indicates the preset residual variance threshold; This represents the maximum value among the synchronization residuals of all child nodes; This indicates the preset maximum residual threshold; Indicates the collision rate; This indicates the preset collision rate threshold; Indicates the number of conflict events, i.e., the most recent The number of broadcast collision events detected from multiple nodes within a time slot; This indicates the number of time slots per frame.

[0052] Preferably, the master node reselection condition is specifically: satisfying the requirement that distributed time synchronization operation has a minimum number of resident frames and meets any of the following conditions:

[0053] ;

[0054] ;

[0055] ;

[0056] ;

[0057] in, This indicates the cumulative runtime of the master node; This indicates the preset threshold for the cumulative runtime of the master node; This represents the energy of the master node; Indicates the preset energy threshold; Indicates the degree of dynamic change in network topology; This indicates a preset threshold for the degree of dynamic change in network topology; Indicates recent Total number of intra-frame join and leave network events; Indicates the transmission frame period.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] (1) The master node selection mechanism of the present invention comprehensively considers parameters such as energy, position, and velocity to ensure the reliability, stability and durability of the selected master node;

[0060] (2) The present invention significantly reduces node collision probability, energy consumption and additional communication overhead through the joint design of TDMA and time synchronization;

[0061] (3) This invention avoids the accumulation of errors caused by external interference and drift mutations in a single synchronization method by adaptive switching between centralized synchronization and distributed synchronization; and avoids single point failure by master node reselection mechanism, thereby improving the long-term stability of network operation.

[0062] (4) The synchronization process of the present invention does not require bidirectional ranging or complex regression calculation, the algorithm has low complexity and is easy to implement in engineering;

[0063] (5) This invention achieves the following: during the node entry phase, nodes broadcast data packets and establish a neighbor information table; during the master node selection phase, node scores are calculated based on neighbor information, and the master node is determined through score broadcasting and consistency confirmation; during the synchronization phase, a combination of centralized and distributed synchronization is used to complete synchronization; during the synchronization quality assessment and rollback switching phase, the residual variance and collision rate of the network time deviation during the distributed synchronization phase are statistically analyzed, and when the synchronization quality degrades, it switches back to centralized synchronization for rapid recovery. In addition, when the master node duration exceeds the limit, the energy is below the threshold, or the topology changes significantly, the master node reselection process is triggered and the master node replacement is completed; this invention solves the problems of difficulty in maintaining synchronization accuracy, high synchronization overhead, and high energy consumption caused by limited processor performance, limited node energy, and high topology dynamics in UAV networks; while ensuring synchronization accuracy, it reduces synchronization communication overhead and energy consumption, and improves the robustness and long-term stable operation capability of the network in dynamic environments, making it suitable for UAV network scenarios with limited processor performance, limited energy, and high topology dynamics. Attached Figure Description

[0064] The present invention will now be described in further detail with reference to the accompanying drawings;

[0065] Figure 1 This is a flowchart of an adaptive time synchronization method for unmanned aerial vehicle (UAV) networks according to an embodiment of the present invention;

[0066] Figure 2 This is the overall process architecture of the adaptive time synchronization method for unmanned aerial vehicle networks according to an embodiment of the present invention;

[0067] Figure 3 This is a schematic diagram illustrating the operation flow of the adaptive time synchronization method for unmanned aerial vehicle (UAV) networks according to an embodiment of the present invention.

[0068] Figure 4 This is a schematic diagram illustrating the scoring weight calculation of the adaptive time synchronization method for unmanned aerial vehicle (UAV) networks according to an embodiment of the present invention.

[0069] Figure 5 This is a distributed time synchronization runtime frame structure for an adaptive time synchronization method for unmanned aerial vehicle (UAV) networks according to an embodiment of the present invention.

[0070] Figure 6 This is a schematic diagram of the time synchronization compensation model of the adaptive time synchronization method for UAV networks according to an embodiment of the present invention; wherein, (a) represents the absolute relationship of logical clocks; and (b) represents the relative relationship of logical clocks. Detailed Implementation

[0071] The present invention will be further described below through specific embodiments.

[0072] like Figure 1 and Figure 2 As shown, the adaptive time synchronization method for UAV networks has the following specific steps:

[0073] S1, Initialization steps for drone network node parameters.

[0074] See Figure 3 As shown, in this embodiment, the drone network consists of a set of nodes. The network is structured around self-organizing communication between nodes via wireless links. Parameters include: network size. One drone node (dynamically changing with the network), deployed in the following area. (such as a three-dimensional matrix region) ), communication radius is The physical and link layers employ low-power wireless communication protocols (to reduce time errors caused by data processing delays, such as IEEE 802.15.4), with a data rate of [data rate missing]. The transmission power is .

[0075] The drone node state definitions include: 1. Pending state: The node enters this state upon power-on or first network access, used for initializing communication parameters, clock, and neighbor table; 2. Neighbor discovery state: The node broadcasts and listens for HELLO data packets, building a neighbor information table; 3. Master node selection state: Each node selects a master node based on the neighbor table information and a scoring algorithm, and then broadcasts the selection result to synchronize a consistent master node across all nodes; 4. Centralized synchronization state: The master node broadcasts a data packet containing a reference time, and child nodes receive the packet for rapid time offset correction; 5. Distributed synchronization state: Child nodes broadcast their local time sequentially according to time slot allocation, and the remaining child nodes receive the broadcast and update their local time offset compensation parameters according to a distributed time synchronization algorithm. The master node is responsible for listening to the broadcast to monitor synchronization.

[0076] The HELLO packet is used during the neighbor discovery and master node selection phases. The HELLO packet content includes: node identifier. Node location Node speed Node energy Broadcast order Local timestamp Scoring weight To reduce uncertainty at the sending end, the timestamp of the transmission time is obtained at the MAC layer, i.e., a point is marked at the time of transmission interruption or transmission completion, so as to minimize transmission delay.

[0077] S2, the steps for node joining the network and discovering neighbors.

[0078] Cold start network access: If If no master node synchronization reference packet TIMESYN_HEAD, master node frame control packet FRAME_CTRL, or node distributed synchronization packet TIMESYN_CCS is detected, the node enters the neighbor discovery state.

[0079] Network access in runtime: If If the above data packets are received, the network is determined to be operational, and the process remains in a pending state. Upon receiving the master node frame control data packet FRAME_CTRL, the master node timestamp is extracted to complete fast clock offset synchronization, and the process is performed in the current frame. The node sends a REQ_JOIN request packet to join the network and waits to receive the next master node frame control data packet FRAME_CTRL to parse the time slot allocation information. After the time slot allocation is completed, the node enters the distributed synchronization state and directly enters the S6 synchronization process.

[0080] Nodes are unaware of each other's existence before and after joining the network; they must establish a connection through neighbor discovery. Parameters for node joining and neighbor discovery include: the HELLO packet broadcast period is... The neighbor noticed the window for a period of time. Each node's local clock is driven by a crystal oscillator, and the crystal oscillator frequency deviation is... Calculation, assuming the maximum frequency offset is (Used for subsequent calculation of synchronization error and resynchronization period constraints).

[0081] node After power-on, it enters a waiting state, initializes system parameters and the local clock counter, and after a period of time, enters the neighbor discovery state. At this stage, the node... The content of the broadcast HELLO data packet is as follows:

[0082] ;

[0083] The neighbor discovered the window Within, nodes broadcast cycles based on HELLO data packets. Based on the CSMA / CA mechanism, a HELLO data packet containing its local time and various information parameters is periodically broadcast to nodes within the communication range. Meanwhile, nodes not in a sending state receive the data packet, unpack it, extract neighbor information, and update their local neighbor information table. Upon receiving neighbor nodes Includes the local timestamp of its sending time After receiving the HELLO data packet, record the local timestamp of the received time. and in the neighbor information table Enter and update table entries:

[0084] ;

[0085] in, Represents a node Received node Number of data packets.

[0086] If some nodes fail to receive broadcasts from certain neighboring nodes correctly and effectively, resulting in missing node information in the neighbor information table, this situation will not affect the subsequent selection of the master node, as the final node score broadcast will be synchronized with the master node of all nodes. In other words, errors in some node information do not affect the overall correct selection of the master node by the system. (This is mentioned in the neighbor discovery window.) After the event ends and no node sends or receives a HELLO data packet for a continuous period of time, the node releases the neighbor discovery state and enters the next stage.

[0087] S3, Steps for selecting the master node.

[0088] After completing S2 neighbor discovery, each node has created a neighbor information table. (See also...) Figure 4 As shown, node According to the neighbor information table Calculate the scoring weight of its own master node :

[0089] ;

[0090] in, Represents a node connectivity, Indicates energy factor, Indicates motion stability, Indicates proximity. Indicates link stability. , , , , These are the corresponding weight coefficients, and they satisfy... .

[0091] node Connectivity :

[0092] ;

[0093] in, For nodes The number of neighbors in the neighbor information table. This represents the number of nodes joining the network.

[0094] node Energy Factor :

[0095] ;

[0096] in, For nodes Remaining energy, including energy consumed by communication, flight, sensors, etc.; This is the initial energy.

[0097] The known communication radius of the UAV node is ,node With nodes relative position Relative velocity :

[0098] ;

[0099] ;

[0100] The link boundary condition is that the distance reaches Then we have:

[0101] ;

[0102] Expanding the formula with vectors:

[0103] ;

[0104] Take the smallest positive root as the link expiration time. If there is no positive root and the connection is already established, it can be considered as... .

[0105] Then node Motion stability :

[0106] ;

[0107] in, Truncate to the upper limit to avoid This distorts the results.

[0108] node Neighborhood :

[0109] ;

[0110] in, This represents the maximum communication distance between drone nodes.

[0111] node Link stability :

[0112] ;

[0113] After the score calculation is completed, the node enters the score broadcast window. .node The scoring data packet is broadcast by competing for the channel using the CSMA / CA mechanism. The content of the HELLO broadcast data packet includes:

[0114] ;

[0115] node Upon receiving neighbor nodes After including the HELLO data packet with rating weights, update the local neighbor information table. :

[0116] ;

[0117] Simultaneously establish a candidate set of master nodes When receiving duplicates, the duplicate items are overwritten and updated based on the sequence number field.

[0118] node According to the candidate set Sort the nodes and select the node with the highest score as the candidate for master node. :

[0119] ;

[0120] When scores are the same or similar, such as To eliminate ambiguity, a deterministic decision rule is used, and comparisons are made in the following priority order:

[0121] 1) Connectivity To ensure communication coverage

[0122] 2) Energy Factor Ensure battery life

[0123] 3) Link stability To ensure stable communication

[0124] To avoid a "multiple master node" situation caused by broadcast packet loss, a master node consistency confirmation window is set here. This is to ensure the uniqueness of the primary node. During this window period, the node... Broadcast TIMESYN_CON packets to announce their selection results. The data packet content includes:

[0125] ;

[0126] in, Represents a node Select the node ID that ranks first in the score. This indicates the weight of the score.

[0127] node In the window After receiving TIMESYN_CON messages from other nodes, the system broadcasts the received scores to the corresponding nodes in the neighbor information table according to the above rules for voting. The primary node receiving the most votes is determined as the consensus primary node. In addition to the primary node, the second-ranked node is designated as a backup primary node for rapid takeover in case of sudden primary node failures such as short-term outages or network disconnection.

[0128] After the consensus confirmation is completed, the unified master node is activated, the node is deselected as the master node, and the network proceeds to the next step.

[0129] S4, centralized time synchronization step.

[0130] The master node, acting as the time reference node, periodically sends synchronization reference broadcast data packets (TIMESYN_HEAD) using a one-way broadcast method. Child nodes mark the received timestamp and estimate the clock phase offset (Time Offset) with the master node based on the one-way broadcast synchronization mechanism. They then update their local clock offset compensation parameters to achieve rapid time synchronization across the entire network. The network configuration centralized synchronization window length is [length to be specified]. The master node synchronous broadcast period is The synchronous broadcast rounds are Offset correction step size Offset convergence threshold .

[0131] In this embodiment, child nodes Before this step begins, the master node ID is confirmed, and the neighbor information table is updated. Then, the centralized synchronization state is entered. The master node performs this process in each cycle. Upon arrival, a synchronization reference data packet TIMESYN_HEAD is broadcast. The packet content includes:

[0132] ;

[0133] in, This indicates the current master node ID. Indicates the number of primary nodes Radio broadcast, This indicates the first broadcast by the master node. Round the local timestamp.

[0134] child nodes Record the local timestamp of the received TIMESYN_HEAD data packet from the master node. child nodes With the master node Wheel time phase deviation The following estimation method is used:

[0135] ;

[0136] in, This represents a one-way transmission delay compensation term, which is an equivalent compensation for the delay in propagation and processing.

[0137] Simultaneously, to further resist reception jitter and external interference, the child node... For continuous The final offset is obtained by statistically smoothing the offset estimate from the secondary reference broadcast. :

[0138] ;

[0139] in, Should meet Subnodes that fail to meet the requirements after the window ends will not proceed to the next TDMA time slot allocation step.

[0140] child nodes Time offset compensation parameters renew:

[0141] ;

[0142] child nodes Output time update:

[0143] ;

[0144] S5, TDMA time slot allocation and frame control steps.

[0145] The master node is responsible for configuring and maintaining the TDMA frame structure parameters: defining the transmission frame period as... Each frame contains There are 1 time slot, and the length of a single time slot is 1. The protection interval (Guard Time) is Offline threshold frame count Offline recovery threshold frame count Frame tail maintenance window See Figure 5 As shown. Wherein:

[0146] ;

[0147] To ensure TDMA operates correctly, To absorb residual synchronization errors between nodes and short-term crystal oscillator drift, the master node sets a synchronization threshold for entering TDMA. (Referred to as the offset convergence threshold in the previous step), so that it satisfies the following constraint relationship with the protection interval:

[0148] ;

[0149] In this embodiment, the master node's transmission time slot is fixed as the first time slot of each frame. This time slot is used to broadcast the FRAME_CTRL frame control data packet to achieve frame boundary alignment and update runtime parameters. Child node Upon receiving FRAME_CTRL, the local frame parameters are updated and the frame start point is calibrated based on this data packet, entering the distributed synchronization state. The content of the broadcast frame control data packet FRAME_CTRL includes:

[0150] ;

[0151] in, Indicates the master node ID; Indicates the master node identifier; This indicates the network system status; the current step is the TDMA mechanism time slot allocation status. Indicates the timestamp sent by the master node; Indicates the current frame number; Indicates the current frame scheduling version; This represents the current frame slot mapping, which the master node uses based on its own neighbor information table. according to Time slots are allocated sequentially in ascending order.

[0152] child nodes Only before its own time slot arrives Wake-up is delayed after the time slot ends. Transmission is disabled, and the node enters a low-power sleep state or performs other information transmission tasks during the remaining time slot. Other nodes remain in receiving mode during this time slot. (The last part, "receiving child node," appears to be a separate, unrelated sentence fragment and is left untranslated.) The broadcast content is processed and updated using a distributed synchronization algorithm, while the master node remains in a listening state, monitoring the transmission status of the corresponding node in each time slot. If continuous... Frame did not detect child nodes Broadcasts within the corresponding time slot are marked as offline; if they continue... If no recovery broadcast is detected for the frame, the frame slot table is updated to release the slot.

[0153] To accommodate dynamic changes in topology, a maintenance window is reserved at the end of each frame. This is used to receive broadcast signals related to node joining and leaving the network. Before joining the network, a newly joining node listens to the master node's frame control data packet FRAME_CTRL to synchronize the current frame status and time. Then, within the reserved maintenance window at the end of the current frame, it sends a network join request REQ_JOIN using the CSMA / CA mechanism. After receiving the request, the master node updates the frame scheduling version and frame time slot mapping. A node preparing to leave the network sends a network leave request REQ_LEAVE within the window.

[0154] S6, Distributed synchronization steps.

[0155] In S4 centralized time synchronization, each node has already adjusted the clock skew compensation parameters. To achieve rapid calibration, this step employs a consistent clock synchronization algorithm to compensate the local clock frequency parameters. and clock offset compensation parameters Simultaneously, iterative updates are performed to suppress synchronization degradation caused by crystal oscillator error accumulation and external interference, and the node enters a distributed synchronization state. If no mode switching condition is triggered, this step continues to be repeated. During this phase, child nodes... Only at the beginning of the frame and its own allocated time slot It wakes up during specific time periods and enters a low-power listening state during other times to reduce energy consumption. See the time synchronization compensation model for details. Figure 6 As shown.

[0156] child nodes With hardware local clock And construct a consistent logical clock. :

[0157] ;

[0158] in, With previous nodes Broadcast timestamp Consistent, this step further breaks it down into detailed components; These are frequency compensation parameters used to compensate for clock drift caused by crystal oscillator frequency offset; These are offset compensation parameters used to compensate for node phase offset. The purpose of these two parameters is to map hardware clock readings to a consistent logic clock; they do not represent the actual physical parameters of the node hardware crystal.

[0159] In the In a frame, child nodes In its allocated time slot The internal broadcast includes a logical clock for consistency of its transmission time. The distributed synchronization data packet TIMESYN_CCS contains the following:

[0160] ;

[0161] Current frame child node Consistent logical clock :

[0162] ;

[0163] When node Upon receiving the node When broadcasting TIMESYN_CCS, record the local hardware clock at the time of reception. And calculate the local logic clock at the time of reception. According to this node Update its receive timing logic clock value:

[0164] ;

[0165] in, and They are nodes and nodes The confidence level is initially set to 1, and is incremented by 1 after each successful parameter update for the corresponding node. For nodes Broadcast transmission to nodes The one-way transmission delay compensation term is an equivalent compensation for the propagation and processing delay. In this scenario, it can be approximately obtained by estimating the fixed processing delay. It can be approximately ignored when the communication is close-range and the processing delay is stable.

[0166] node Update clock offset compensation parameters :

[0167] ;

[0168] No. The end of a frame, i.e., the end of the previous synchronization round, is when the node... Record the updated logical clock value:

[0169] ;

[0170] node Update clock frequency compensation parameters :

[0171] ;

[0172] Therefore, the node Update Consistent Logical Clock :

[0173] ;

[0174] S7, Synchronous quality assessment and rollback switching steps.

[0175] This step is based on S6. During the continuous operation of distributed time synchronization, the master node continuously evaluates the network synchronization quality and switches back to centralized time synchronization based on the evaluation results. At the same time, the master node continuously monitors its own status. When the master node's operating status is no longer suitable to continue serving as the master node, it will switch back to master node re-selection.

[0176] In the Frame Node broadcast slots In the middle, the master node receives the node Broadcasts include a logical clock that ensures consistency with their transmission time. The distributed synchronization data packet TIMESYN_CCS records the local logical clock at the time of reception. (The master node does not need to synchronize, and all child node clocks are synchronized based on the master node's clock, so no compensation parameters are needed.) Therefore, the master node's synchronization residual for this link is defined as:

[0177] ;

[0178] in, For nodes The one-way transmission delay compensation term for broadcast transmission to the master node is an equivalent compensation for the propagation and processing delay. In this scenario, it can be approximately obtained by estimating the fixed processing delay. It can be approximately ignored when the communication is close and the processing delay is stable.

[0179] The master node is at the Frame summaries of all node residual sets Calculate the residual variance:

[0180] ;

[0181] To suppress occasional anomalies, the master node uses a sliding window (nearest) window. (Frame) Statistics:

[0182] ;

[0183] The maximum residual is expressed as:

[0184] ;

[0185] Master node statistics recent If a frame detects broadcast collisions from multiple nodes within a time slot, the number of packet collision events is counted. Define the collision rate:

[0186] ;

[0187] Master node statistics recent Total number of intra-frame join and leave network events Define event frequency Estimate the degree of dynamic change in network topology:

[0188] ;

[0189] When any of the following conditions are met, and the distributed time synchronization operation has a minimum number of resident frames, The synchronization switch is triggered and reverts to S4:

[0190] ;

[0191] ;

[0192] ;

[0193] When the synchronization switchback rollback condition is triggered, the master node is in the [missing information]. Frame (next round of synchronization frame) frame start The synchronization switch rollback instruction is broadcast via the FRAME_CTRL control data packet. After receiving the instruction, the child node stops executing the distributed time synchronization phase and rolls back to execute S4: centralized time synchronization across the entire network.

[0194] When any of the following conditions are met, and the distributed time synchronization operation has a minimum number of resident frames, The master node reselection is triggered and the rollback to S3 is reverted:

[0195] ;

[0196] ;

[0197] ;

[0198] When the master node reselection fallback condition is triggered, the master node is in the... Frame (next round of synchronization frame) frame start The master node reselection rollback instruction is broadcast via the FRAME_CTRL control data packet. Upon receiving this instruction, the child node stops executing the distributed time synchronization phase and simultaneously updates the neighbor information table. Mark the current master node as not participating in the next round of master node selection, and roll back the entire network to execute S3: master node selection.

[0199] In summary, the overall process of this embodiment is as follows: Network initialization is performed, and nodes listen to the current network status when joining the network. If the network is running, the node listens to the master node frame control data packet to obtain the current synchronization frame status and sends a network entry request at the end of the frame. If the network is cold-starting and running for the first time, the node enters the neighbor discovery phase. After the neighbor discovery phase, the node enters the master node selection phase. The node broadcasts its pre-selected master node and collects broadcasts from other nodes, selecting the node with the most votes as the master node. After the master node is selected, the node enters the centralized synchronization phase. The master node broadcasts its local timestamp, and child nodes receive it and update their local clock compensation parameters to achieve fast synchronization. After multiple synchronizations, nodes that meet the synchronization accuracy requirements enter the TDMA time slot allocation phase. During the time slot allocation phase, the master node occupies the first time slot of each frame and simultaneously allocates time slots to child nodes. After allocation, the node enters the distributed synchronization phase. The master node broadcasts frame control data packets at the beginning of each frame to adjust frame parameters and broadcasts the time slot status. Child nodes broadcast their local timestamps within the allocated time slots, and other child nodes receive them and iteratively update their local time compensation parameters according to the distributed synchronization algorithm. During this phase, the master node simultaneously listens to the broadcasts of other child nodes to assess the synchronization quality. When certain switching conditions are met, the system's operating status is adjusted, and the switching process is rolled back.

[0200] The above are merely specific embodiments of the present invention, but the design concept of the present invention is not limited thereto. Any non-substantial modifications made to the present invention using this concept shall be considered as infringing upon the protection scope of the present invention.

Claims

1. An adaptive time synchronization method for unmanned aerial vehicle (UAV) networks, characterized in that, Includes the following steps: S1, Unmanned Aerial Vehicle (UAV) network node parameter initialization steps; S2, Node Networking and Neighbor Discovery Steps: Each drone node sends a HELLO data packet, and each node updates the neighbor information in the neighbor information table with the received HELLO data packet. S3, Master Node Selection Step: Each node calculates the master node rating weight based on the neighbor information table and broadcasts a rating data packet containing the master node rating weight; Each node updates the master node rating weight in the neighbor information table with the received rating data packet and selects the node with the highest master node rating weight in the neighbor information table as the master node candidate. Each node broadcasts a consistency confirmation data packet including the candidate for master node; each node counts the number of votes for the candidate for master node in the neighbor information table based on the received consistency confirmation data packet, and sets the candidate for master node with the most votes as the master node, and the remaining nodes as child nodes; S4, centralized time synchronization step: the master node broadcasts a reference data packet including a timestamp, and the child nodes receive the reference data packet and synchronize their time with the master node; S5, TDMA time slot allocation and frame control steps: The master node broadcasts frame data packets in the time slot at the beginning of the TDMA frame. Each child node receives the frame data packets, updates its local frame parameters, and performs frame calibration. The child nodes are only awakened in the time slot at the beginning of the frame and in the allocated time slot, and are in a listening state during the rest of the time period. S6, Distributed synchronization step: Child nodes broadcast distributed synchronization data packets including a consistent logical clock in their allocated time slots; each child node receives the distributed synchronization data packets and updates its consistent logical clock. S7, Synchronization Quality Assessment and Rollback Switching Steps: The master node receives the distributed synchronization data packets from the child node and calculates the synchronization residual of the clock deviation between the child node and the master node; when the synchronization rollback condition based on the synchronization residual is met, proceed to S4; when the master node reselection condition is met, proceed to S3 and the current master node does not participate in the master node selection.

2. The adaptive time synchronization method for unmanned aerial vehicle (UAV) networks according to claim 1, characterized in that, The master node scoring weight is represented as follows: ； ； in, Represents a node The main node's scoring weight; Represents a node Connectivity is used to measure the percentage of a node's neighbors; Represents a node The energy factor is based on the proportion of remaining energy. Represents a node The mobility stability; Represents a node The proximity relationship measures the average distance between a node and its neighbors; Represents a node Link stability is based on packet reception rate; , , , and Indicates the weighting coefficient; Indicates upper limit truncation; This represents the link expiration time, and its value is the link boundary condition. The smallest positive root or , Represents a node With nodes The relative position, Represents a node With nodes relative velocity, Indicates the communication radius of the drone node. Indicates time, Indicates modulo; Represents a node Neighbor information table; This indicates taking the minimum value.

3. The adaptive time synchronization method for unmanned aerial vehicle (UAV) networks according to claim 2, characterized in that, S3 further includes: eliminating ambiguity using deterministic decision rules; specifically: When the primary node score weights of two nodes are less than the preset score difference threshold, they are compared in order of priority: connectivity, energy factor, and link stability, and the node with the higher score is selected as the primary node candidate.

4. The adaptive time synchronization method for unmanned aerial vehicle (UAV) networks according to claim 1, characterized in that, The child node receives the reference data packet and synchronizes its time with the master node, as follows: Calculate child nodes The time phase deviation from the master node is expressed as: ； in, Indicates time phase deviation; This represents the one-way transmission delay compensation term; Represents child nodes The local timestamp at the moment the master node's TIMESYN_HEAD data packet is received. ; This indicates the first broadcast by the master node. Round the local timestamp; child nodes For continuous The final offset is obtained by averaging the time phase deviation of the secondary reference broadcast. Updating child nodes using the final offset is represented as: ； ； in, Indicates the updated node Time offset compensation parameters; Represents a node Time offset compensation parameters; Indicates the final offset; Indicates the offset correction step size; Indicates the updated child nodes Output time; Represents child nodes The output time.

5. The adaptive time synchronization method for unmanned aerial vehicle (UAV) networks according to claim 1, characterized in that, The frame parameters of the Time Division Multiple Access (TDMA) are expressed as follows: ;in, Indicates the transmission frame period; Indicates the number of time slots per frame; Indicates the time slot length; Indicates the protection interval; This indicates the frame tail maintenance window; child nodes are scheduled to arrive before the allocated time slot. Time-based wake-up, delayed after the time slot ends. The node turns off transmission during the designated time slot and enters a low-power sleep state or performs other information transmission tasks during the remaining time slot, while the other nodes remain in the receiving state during this time slot.

6. The adaptive time synchronization method for unmanned aerial vehicle (UAV) networks according to claim 5, characterized in that, The frame tail maintenance window is used to receive broadcast signals for child node joining and leaving. Before joining the network, a newly joining child node listens to the master node's frame control data packets to synchronize the current frame state and time. Then, within the frame tail maintenance window of the current frame, it sends a network joining request REQ_JOIN using the CSMA / CA mechanism. After receiving the request, the master node updates the frame scheduling version and frame time slot mapping. A child node preparing to leave sends a network leaving request REQ_LEAVE within the frame tail maintenance window. If the master node does not hear a broadcast from a child node in the corresponding time slot within a consecutive number of frames that reach the leaving threshold frame count, it marks the child node as leaving the network. If it still does not hear a recovery broadcast within a number of frames that reach the leaving recovery threshold frame count, it updates the frame time slot table and releases the time slot.

7. The adaptive time synchronization method for unmanned aerial vehicle (UAV) networks according to claim 1, characterized in that, The distributed time synchronization steps are as follows: When node Upon receiving the node When broadcasting distributed synchronization data packets, the local hardware clock at the time of reception is recorded and the local logical clock at the time of reception is calculated. Based on this, the node... Update its receive time logic clock value, expressed as: ； in, Indicates the updated node Receive time local logic clock; Represents a node Receive time local logic clock; and Representing nodes respectively and nodes The confidence level is initially set to 1, and is incremented by 1 after each successful parameter update for the corresponding node. Represents a node Broadcast transmission to nodes One-way transmission delay compensation term; Represents a node The transmission time logic clock; node Updating the consistent logical clock is represented as: ； ； ； in, Represents a node Consistent logical clock; Indicates the updated node Clock frequency compensation parameters; Represents a node Clock frequency compensation parameters; Indicates the local hardware clock; Indicates the updated node Clock skew compensation parameters; Represents a node Clock skew compensation parameters; Indicates the first End of round node Updated logical clock value.

8. The adaptive time synchronization method for unmanned aerial vehicle (UAV) networks according to claim 1, characterized in that, The synchronization residual is expressed as: ； in, Represents a node The synchronization residual with the master node; Represents a node One-way transmission delay compensation item for broadcast transmission to the master node; Indicates the first wheel; Represents a node The logical clock at the time of transmission; The logical clock that indicates the moment the master node receives the data.

9. The adaptive time synchronization method for unmanned aerial vehicle (UAV) networks according to claim 1, characterized in that, The specific conditions for synchronization backoff based on synchronization residuals are: the distributed time synchronization operation must meet at least the minimum number of resident frames and satisfy any of the following conditions: ； ； ； ； in, This represents the residual variance of all child nodes; This indicates the preset residual variance threshold; This represents the maximum value among the synchronization residuals of all child nodes; This indicates the preset maximum residual threshold; Indicates the collision rate; This indicates the preset collision rate threshold; Indicates the number of conflict events, i.e., the most recent The number of broadcast collision events detected from multiple nodes within a time slot; This indicates the number of time slots per frame.

10. The adaptive time synchronization method for unmanned aerial vehicle (UAV) networks according to claim 1, characterized in that, The master node reselection conditions are as follows: the distributed time synchronization operation must meet at least the minimum number of resident frames and satisfy any of the following conditions: ； ； ； ； in, This indicates the cumulative runtime of the master node; This indicates the preset threshold for the cumulative runtime of the master node; This represents the energy of the master node; Indicates the preset energy threshold; Indicates the degree of dynamic change in network topology; This indicates a preset threshold for the degree of dynamic change in network topology; Indicates recent Total number of intra-frame join and leave network events; Indicates the transmission frame period.

Images