High-dynamic mobile ad hoc network communication and sensing integrated routing method, device, equipment and storage medium

Abstract

The invention discloses a high-dynamic mobile ad hoc network communication and sensing integrated routing method, device and equipment and a storage medium, an OFDM (Orthogonal Frequency Division Multiplexing) sensing system is adopted to actively acquire feature information of neighbor nodes, and meanwhile, neighbor features passively acquired based on communication feedback are matched and fused, so that the sensing precision and efficiency in an ad hoc network are remarkably improved. In order to reduce noise interference of a wireless communication environment and realize prediction of a future state of a network, a Kalman filter is adopted to carry out dynamic filtering and state estimation on neighbor feature information. And on the basis of the processed sensing data, the node further calculates stability and effectiveness indexes of the link, inputs the stability and effectiveness indexes into a link quality comprehensive evaluation model constructed by fuzzy logic, and finally generates predictive link quality parameters to provide priori knowledge for following routing selection. And the routing efficiency and the resource utilization are optimized through an on-demand routing normal form and a multipath routing mechanism. And meanwhile, through a node early warning mechanism and a fault pre-detection mechanism, the control overhead is remarkably reduced, and the network stability is improved.

Description

Technical Field

[0001] The present invention belongs to the field of wireless communication and network technology, and in particular relates to a highly dynamic mobile ad hoc network interaceptive integrated routing method, device, equipment and storage medium. Background Art

[0002] Highly Dynamic Mobile Ad-hoc Networks (HD-MANET) are distributed wireless communication systems comprised of a variety of high-speed mobile nodes, encompassing a wide range of autonomously mobile devices, including unmanned aerial vehicles (UFPs), ground vehicles, and naval vessels. Flying Ad-hoc Networks (FANET), a typical HD-MANET scenario, have become a key entry point for HD-MANET research due to their unique three-dimensional spatial deployment, high-speed mobility, and diverse mission requirements. FANETs, ​​comprised of drones and unmanned airships, are capable of autonomous networking and dynamically adjusting communication paths, without relying on fixed infrastructure. As an extension of mobile ad-hoc networks in three-dimensional airspace, FANETs have significant applications in military collaborative operations, emergency communications support, and smart city management. They have attracted significant attention due to their flexible deployment, cost-effectiveness, and robust survivability. The core characteristic of this network is its extreme dynamics: the high-speed movement of nodes (up to Mach speeds) causes the network topology to continuously and rapidly change, which poses numerous technical challenges: First, the severe Doppler effect caused by high-speed movement significantly degrades communication quality; second, frequent topology reconfiguration makes it difficult for traditional routing protocols to maintain stable data transmission paths; third, the complex aerial environment places higher demands on the network's real-time and reliability; and fourth, diverse mission scenarios (such as tactical reconnaissance and logistics distribution) require the network to intelligently adapt to differentiated Quality of Service (QoS) requirements. These characteristics make FANET a highly challenging frontier in current wireless network research.

[0003] Routing protocols play a key role in FANETs, ​​responsible for establishing and maintaining efficient data transmission paths in a dynamically changing network environment. Existing FANET routing protocols fall into two main technical approaches: one is to improve traditional MANET routing protocols to adapt them to the highly dynamic nature of FANETs; the other is to employ intelligent optimization algorithms to model the FANET routing problem and discover the optimal routing strategy through iterative exploration. Driven by these two technical approaches, FANET routing protocols can be divided into four categories: topology-based routing, location-based routing, swarm intelligence routing, and reinforcement learning routing. Routing based on network topology requires selecting a transmission path before data is transmitted to ensure that packets reach their destination. These protocols can generally be categorized as proactive or reactive. Proactive routing protocols, also known as table-driven routing, rely on each node in the network to maintain a complete routing table and regularly update this information to ensure it remains up-to-date. These protocols do not rely on packet transmission to discover or maintain routing information. The primary advantage of proactive routing protocols is their ability to rapidly respond to network changes because routing information is precomputed. However, in highly dynamic scenarios, where inter-node links frequently fluctuate and routing stability is low, proactive routing requires constant updates of routing information, resulting in significant overhead. Reactive routing initiates route discovery on demand, sacrificing initial route establishment latency in exchange for higher packet delivery rates, resulting in increased transmission latency. These two routing approaches offer a compromise between real-time performance and overhead, but cannot guarantee real-time performance while minimizing routing overhead. Traditional routing primarily draws on MANET routing protocols, where nodes calculate the shortest path based on passively acquired information such as network topology and link status. However, this approach exhibits significant limitations in three-dimensional, high-speed scenarios: network status information quickly becomes outdated, there is a lack of efficient perception and prediction of future states, and link quality lacks a standardized and comprehensive evaluation. This weakens routing performance, leading to increased latency and severe packet loss. Thanks to the development of the Global Navigation Satellite System (GNSS), geographic information-based routing protocols have become widely popular. Nodes do not need to explore the entire network state, but instead greedily forward based on local information, unifying the data transmission and route discovery processes. However, this protocol requires UFP nodes to be able to accurately locate themselves and their destination nodes. In highly dynamic FANET scenarios, these nodes not only need to frequently obtain their own positioning information but also frequently perform neighbor discovery, consuming significant amounts of energy and spectrum resources. Furthermore, relying on distance for path selection is prone to local optimality and routing holes. In recent years, inspired by emerging intelligent optimization algorithms, swarm intelligence routing and reinforcement learning routing have come into the spotlight. These technologies, due to their distributed, adaptive, and global optimal solution-finding capabilities, are well-suited to solving FANET routing decision-making problems.In particular, the application of deep reinforcement learning, by modeling routing problems as Markov decision processes, enables unmanned aerial vehicles (UAVs) to autonomously learn optimal routing strategies. However, in highly dynamic scenarios, these protocols often suffer from blind search, local optimality, and slow convergence. Breaking through technical bottlenecks and designing specialized routing protocols for the ever-emerging UFP application scenarios remains challenging.

[0004] Existing FANET routing schemes are primarily based on optimization and improvements of traditional terrestrial mobile ad hoc network routing schemes. While these improvements provided an important foundation for initial FANET research, due to the unique characteristics of FANET nodes, such as high-speed mobility and drastic topology changes, direct application of these solutions can lead to frequent routing interruptions, high packet loss rates, and excessive routing maintenance overhead. To address the challenges posed by this highly dynamic environment, researchers have developed a variety of enhanced routing schemes specifically for FANET. These schemes primarily aim to address link instability caused by high-speed node mobility while maintaining node connectivity as much as possible, improving routing stability, reducing latency, and lowering routing overhead, while ensuring reliable quality of service. These enhanced schemes typically employ more complex routing strategies, such as predictive routing, context-aware mechanisms, and multipath routing, to adapt to the high node dynamics and rapid changes in network topology within FANET. Furthermore, these schemes take into account the unique communication environment of FANET, such as three-dimensional spatial deployment constraints, the unique mobility patterns of aerial nodes (such as formation flying and mission-oriented mobility), and the requirements for air-ground collaborative communication, thereby providing more reliable and efficient routing solutions for FANET. However, these improvement solutions still have shortcomings in dealing with the high-speed mobility challenges of FANET.

[0005] Topology awareness is a key technology in FANET, primarily used to monitor and predict network topology changes in real time to optimize communication performance. Its core objectives include dynamically tracking changes in node location, link status, and neighbor relationships; providing topology information to dynamic routing protocols (such as OLSR and AODV) to reduce routing overhead; adjusting bandwidth, power, and other resources based on topology changes to improve network efficiency; and rapidly detecting node failures or link disruptions to trigger topology repair. Key approaches include neighbor discovery and link monitoring, such as periodic beacon broadcasts to determine link stability through a beacon mechanism; dynamically adjusting beacon frequency with an adaptive beacon interval; and link quality assessment based on signal strength (RSSI), expected transmission time (ETX), or latency measurements. Location awareness includes GPS / GNSS positioning, relative positioning, and topology mapping. Link prediction includes predicting future node locations using methods such as Kalman filtering, Markov models, LSTM, and reinforcement learning, and classifying link stability through supervised learning. Finally, cross-layer topology awareness includes collaborative optimization of routing decisions between the physical and network layers and energy-aware topology control. Typical application scenarios include emergency communications, collaborative search, and military reconnaissance. However, existing technologies face challenges in high dynamics, energy limitations, and 3D topological complexity. Future development directions include combining technologies such as federated learning and digital twins to achieve more accurate topology predictions, as well as Integrated Sensing and Communication (ISAC) technology to combine communication and perception capabilities to obtain accurate neighbor information with minimal overhead, predict link status in real time, and optimize routing decisions.

[0006] In the existing technical solutions, such as the invention patent with application number CN202311835326.1, "A perception-assisted collaborative neighbor discovery method for the initial network construction stage", a perception-assisted collaborative neighbor discovery method for the initial network construction stage is proposed, which relates to the field of aerial network technology. The perception list and neighbor list of each node in the unmanned aerial platform self-organizing network are initialized; at the beginning of the current time slot, each node selects its own transceiver mode and target beam direction; the node in the transmitting mode sends an ISAC signal to the target beam direction and receives an echo, updates the perception list according to the echo, and receives an ACK feedback signal to update the perception list and neighbor list; the node in the receiving mode listens to the ISAC signal sent in the target beam direction and sends an ACK feedback signal, and updates the perception list and neighbor list according to the ISAC signal; when each node in the unmanned aerial platform self-organizing network has discovered all neighbors, or when the current time slot exceeds the preset number of time slots, the neighbor discovery process ends.

[0007] This method significantly improves the efficiency and accuracy of neighbor discovery by deeply integrating communication and perception functions. Its core lies in the design of a dynamic collaboration mechanism: at the communication level, nodes achieve indirect perception by exchanging ISAC signals containing perception lists and neighbor lists, accelerating node discovery across the entire network; at the perception level, beam scanning priorities are dynamically updated based on echo signals and ACK feedback, giving priority to detecting the direction of potential targets, while combining parameters such as distance and SNR to filter high-confidence perception information to optimize accuracy. This solution breaks through the limitations of the separation of perception and communication, reduces the number of blind scans through two-way collaboration, improves robustness, and is particularly suitable for highly dynamic and dense deployment scenarios, providing an efficient solution for the integrated synaesthesia networking protocol.

[0008] The invention patent application number CN202210933403.6, "A Method for Updating Routing in an Unmanned Aerial Platform Self-Organizing Network Based on Kalman DNN Model Prediction," improves the accuracy of neighbor information and meets the requirements of high data packet delivery rate and low end-to-end latency for unmanned aerial platform self-organizing network systems. First, a node position prediction model based on Kalman DNN is proposed, which is used to calculate the position of the node at the next moment. Then, a neighbor update event is generated based on the prediction result. After a timeout event occurs, the neighbor table is updated, a neighbor signaling packet is broadcast, and the route is updated based on the neighbor table. Finally, the route is calculated based on the comprehensive link quality value. This invention introduces neighbor node position prediction into traditional routing protocols. The use of position prediction can reduce the impact of rapid topology changes on link quality, overcoming the problem that traditional mobile self-organizing network routing protocols are not suitable for environments with high-speed node movement. While ensuring the stability of the system link, it also improves the packet delivery rate and reduces end-to-end latency.

[0009] Both this method and this proposal optimize the routing update problem in self-organizing networks of unmanned aerial platforms, but they differ significantly in terms of network model, routing update mechanism, and prediction method. First, this method primarily updates routing based on a prediction model combining a Kalman filter and a DNN, focusing on improving the accuracy of neighbor information by modifying the Kalman filter's predictions, thereby optimizing route selection. In contrast, this proposal employs a synaesthesia-integrated architecture, combining motion feature prediction using a Kalman filter with link quality prediction based on fuzzy logic. This approach not only focuses on node location prediction but also comprehensively considers link quality and network topology dynamics. Second, this proposal's routing updates rely on periodic Hello packet exchanges between nodes to discover neighbors and update the neighbor table. This proposal, however, fuses information from both the communication and perception domains, utilizing the OFDM synaesthesia-integrated waveform to actively acquire neighbor feature information. Combined with passive sensing using beacon packets, this approach enables more timely and accurate acquisition of neighbor node status information, providing a more reliable basis for routing selection. Finally, this solution mainly calculates routes by modifying the neighbor table during route updates. In the route discovery phase, this solution introduces an on-demand routing paradigm based on a node early warning mechanism, and adds fault pre-detection and fast path recovery mechanisms in the route maintenance phase. This can more effectively cope with network topology changes in highly dynamic environments and improve network robustness and communication continuity.

[0010] Existing FANET routing schemes are primarily based on optimization and improvements of traditional terrestrial mobile ad hoc network routing schemes. While these improvements provided an important foundation for initial FANET research, due to the unique characteristics of FANET nodes, such as high-speed mobility and drastic topology changes, direct application of these solutions can lead to frequent routing interruptions, high packet loss rates, and excessive routing maintenance overhead. To address the challenges posed by this highly dynamic environment, researchers have developed a variety of enhanced routing schemes specifically for FANET. These schemes primarily aim to address link instability caused by high-speed node mobility while maintaining node connectivity as much as possible, improving routing stability, reducing latency, and lowering routing overhead, while ensuring reliable quality of service. These enhanced schemes typically employ more complex routing strategies, such as predictive routing, context-aware mechanisms, and multipath routing, to adapt to the high node dynamics and rapid changes in network topology within FANET. Furthermore, these schemes take into account the unique communication environment of FANET, such as three-dimensional spatial deployment constraints, the unique mobility patterns of aerial nodes (such as formation flying and mission-oriented mobility), and the requirements for air-ground collaborative communication, thereby providing more reliable and efficient routing solutions for FANET. However, these improvement solutions still have shortcomings in dealing with the high-speed mobility challenges of FANET.

[0011] In summary, the shortcomings of the existing technical solutions are mainly manifested in the following aspects:

[0012] 1. High Dependence on Positioning System Accuracy: Some geolocation-based routing methods use real-time location information for routing, avoiding the overhead of frequent topology discovery. However, these solutions rely on precise positioning systems, such as the Global Navigation Satellite System (GNSS), which not only increases system complexity but also significantly degrades routing performance when positioning information is inaccurate or unavailable.

[0013] 2. Inadequate topology perception and prediction capabilities: Traditional routing primarily draws on MANET routing protocols, where nodes calculate the shortest path based on passively acquired information, such as network topology and link status. However, this approach presents serious limitations in three-dimensional high-speed scenarios: network status information quickly becomes outdated, and there is a lack of efficient perception and prediction of future status. Nodes rely solely on periodically broadcast beacons to obtain digital identifiers and physical characteristics from their neighbors to assess link connectivity. Nodes can only passively obtain feedback from the previous period, which reduces the credibility of neighbor information and increases the probability of incorrect link status estimation.

[0014] 3. Lack of a dynamic link evaluation system: Existing routing solutions lack quantifiable link quality metrics in dynamic networks, primarily because they typically fail to incorporate factors such as link stability, transmission reliability, and latency into their quantitative evaluation. These solutions may overly rely on simplified metrics such as the number of routing hops, ignoring the dynamic nature of link quality over time. In highly dynamic network environments such as FANET, the rapid movement of nodes leads to frequent changes in link status. Traditional quantitative metrics cannot accurately reflect real-time link quality, thus impacting routing effectiveness and network performance.

[0015] 4. Low routing stability and topology adaptability: Existing routing solutions suffer from low stability, primarily due to their inability to effectively cope with the high-speed node mobility and rapid changes in network topology within FANETs. Active routing protocols require frequent updates of routing information to maintain accuracy, but this is difficult to achieve in a high-speed, dynamic environment. Reactive routing protocols also experience significant latency in route discovery and are slow to respond to sudden topology changes. These factors combine to make it difficult for existing solutions to maintain stable routing paths and adapt to frequent changes in network topology.

[0016] 5. High routing maintenance overhead: Existing routing solutions require continuous monitoring and updating of link status and routing tables in the network to adapt to the rapid movement of nodes and the ever-changing network topology. This continuous update process consumes a large amount of communication and computing resources, especially in highly dynamic environments such as FANET, where the frequency of updates is higher, resulting in significantly increased overhead. Summary of the Invention

[0017] In response to the shortcomings of the existing technology, the present invention proposes a highly dynamic mobile ad hoc network synaesthesia integrated routing method, device, equipment and storage medium. Aiming at the highly dynamic flight ad hoc network scenario, a node mutual perception method based on OFDM synaesthesia integrated waveform is proposed, and a low-overhead link quality prediction routing scheme based on synaesthesia integration is designed.

[0018] The highly dynamic mobile ad hoc network synaesthesia integrated routing method of the present invention comprises the following specific steps:

[0019] Step 1: After all nodes initialize their state information, they obtain the location and velocity information of neighboring nodes. Furthermore, they use OFDM synaesthesia integrated waveform to realize neighboring node echo perception.

[0020] Step 2: Calculate vector similarity by improving the probability measurement function of cosine similarity, and use the bipartite graph model and convex optimization method to perform feature matching and fusion between the communication domain feature matrix BF(t) and the perception domain feature matrix SF(t) to achieve complete reconstruction and accurate identification of neighbor node information.

[0021] Step 3: Each node uses the Kalman filter to reduce noise and predict the motion state of neighboring nodes to improve the accuracy of state estimation.

[0022] Step 4: Calculate link quality prediction indicators based on fuzzy logic. Define the normalized predicted link lifetime as the core indicator for evaluating link stability, and define the normalized predicted expected number of transmissions to evaluate the communication effectiveness of the link. Furthermore, input the normalized predicted link lifetime and the normalized predicted expected number of transmissions into the fuzzy logic system, and generate predictive link quality parameters through fuzzification, rule mapping and combination, and defuzzification.

[0023] Step 5: In the routing discovery phase based on the node early warning mechanism, an on-demand routing paradigm is adopted. The network topology is dynamically discovered by flooding routing control packets on demand. A link quality prediction routing protocol integrating communication perception is constructed. At the same time, a lightweight node early warning mechanism is designed to monitor its own remaining energy E. rem and cache queue length Q, quickly identify high-risk nodes, and selectively forward RREQ messages.

[0024] Step 6: During the routing maintenance phase, fault pre-detection and fast path recovery mechanisms are introduced to improve network robustness.

[0025] The advantages of the present invention are:

[0026] 1. This invention significantly improves the perception accuracy and efficiency in highly dynamic mobile ad hoc networks through a node mutual perception method based on an OFDM synaesthesia integrated waveform. This method breaks the isolation between communication and perception, matches and fuses the features of the communication domain and the perception domain, and obtains the highly timely motion status of neighboring nodes with low overhead. This solves the problem of topology perception in highly dynamic environments and provides real-time, accurate network topology information for routing mechanisms. In addition, a Kalman filter is introduced to reduce noise and predict the motion status of neighboring nodes, allowing for early perception of potential link disruptions and providing more accurate prior knowledge for routing decisions.

[0027] 2. The present invention significantly reduces the frequency of route reconstruction and network overhead through active perception mechanism and Kalman filter prediction, ensuring the efficiency and stability of data transmission. The routing adaptability of dynamic networks is significantly improved through link quality prediction and dynamic adaptive routing method based on node early warning mechanism. This solution implements link quality prediction based on Kalman filtering and fuzzy logic. Nodes dynamically adjust routing paths according to the prediction results and can quickly respond to changes in network topology. At the same time, by defining the normalized predicted link lifetime and the normalized predicted expected number of transmissions as prior knowledge, the robustness and agility of route discovery are enhanced. This dynamic adaptive routing method based on link quality prediction enables routing protocols to better adapt to highly dynamic flying self-organizing network environments, especially when the network topology changes rapidly due to frequent node movement.

[0028] 3. The present invention optimizes routing efficiency and resource utilization through an on-demand routing paradigm and a multipath routing mechanism. The on-demand routing paradigm only starts the routing discovery process when needed, avoiding unnecessary routing update overhead. At the same time, the multipath routing mechanism improves the network's fault tolerance and data transmission reliability by selecting the optimal path as the primary path and retaining the suboptimal path as the backup path. In addition, this solution uses link quality prediction to identify potential link failures in advance, optimize path selection, and further improve resource utilization efficiency. Advantage 4: Reduced control overhead and improved network stability

[0029] 4. This invention significantly reduces control overhead and improves network stability through node early warning and fault pre-detection mechanisms. The node early warning mechanism proactively identifies high-risk nodes by monitoring their own status (such as remaining energy and cache queue length) and inhibits their participation in ineffective route discovery processes, reducing network congestion and control overhead. The fault pre-detection mechanism proactively identifies and corrects routing paths approaching fault thresholds by monitoring node status in real time, avoiding data transmission interruptions. The introduction of these mechanisms enables the network to maintain high stability and communication continuity in highly dynamic environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1This is an overall flow chart of the synaesthesia integrated routing method for highly dynamic mobile ad hoc networks of the present invention;

[0031] Figure 2 Schematic diagram of the dual-domain feature matrix matching model based on bipartite graph;

[0032] Figure 3a is the normalized predicted link survival time membership function;

[0033] Figure 3b The membership function for normalized predicted expected transmission times is

[0034] Figure 4 Output the membership function for predicting link quality for defuzzification;

[0035] Figure 5 This is a structural block diagram of the synaesthesia integrated routing device for highly dynamic mobile ad hoc networks of the present invention. DETAILED DESCRIPTION

[0036] The present invention will be described in further detail below with reference to the accompanying drawings.

[0037] The present invention is directed to a synaesthesia-integrated routing method for a highly dynamic mobile ad hoc network. Its physical entities include multiple unmanned aerial platforms (which can be implemented in the form of multiple computing devices) with the same data forwarding capability or other nodes with routing and forwarding capabilities, eliminating the need for ground infrastructure and a centralized control center. The device module of each node includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The computer device can be a server, comprising a processor, a memory, a network interface, and a database connected via a system bus. The processor of the device is used to provide computing and control capabilities. The memory of the device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the device is used to store data related to the highly dynamic mobile ad hoc network, such as node location information, link quality data, etc. The network interface of the device is used to communicate with other external nodes via a network connection. When the computer program is executed by the processor, the synaesthesia-integrated routing method of the present invention is implemented.

[0038] In this embodiment, the highly dynamic FANET is composed of a group of homogeneous unmanned aerial platform nodes that move autonomously in three-dimensional space. These nodes are interconnected through wireless communication links to form a dynamically changing network topology. The network can be represented as a time-varying directed graph in represents a set of N UFPs with unique identifiers, Each UFP node is equipped with an omnidirectional antenna and can obtain its own motion status information regularly through the global navigation satellite system. For example, the node u i The three-dimensional position at time t and velocity vector

[0039] In terms of communication link modeling, the present invention adopts a dual criterion based on distance and signal quality to determine the reachability of the link. Specifically, the reachability of a link is determined if and only if the Euclidean distance d between nodes does not exceed the maximum communication radius R and the received signal power P r Above the receiver sensitivity threshold P th The communication link is considered connected only when . Considering that the UAV network mainly operates in a LoS propagation environment, this paper adopts the free space propagation model + Nakagami-m as the wireless propagation model. The free space propagation model is used to calculate the received signal power, and the Nakagami-m distribution is introduced to characterize the small-scale fading characteristics, thereby more accurately reflecting the channel characteristics in the actual communication environment.

[0040] Node u j and u i Communication distance d ij Calculated by the Euclidean distance formula:

[0041]

[0042] Where p j For node u i The three-dimensional position at time t,

[0043] Received signal power P r It is determined by the free space propagation formula:

[0044]

[0045] Where, P r for u j The received power, P t for u i The transmission power, G t and G r are the transmitting and receiving antenna gains, λ is the radio wavelength, and d ij For node u i and u j The propagation distance between them.

[0046] The statistical characteristics of the received signal amplitude r are described by the probability density function of the Nakagami-m distribution, which can flexibly characterize channels with different fading degrees through the parameter m.

[0047]

[0048] Where Γ(·) represents the gamma function and m is the fading parameter.

[0049] In order to take into account the regularity and randomness of the movement of the unmanned aerial platform, the present invention adopts an improved three-dimensional GM mobile model to describe the node movement pattern in a scenario with a limited boundary range. The model maintains the continuity of movement by introducing a memory factor, and uses Gaussian random variables to simulate the uncertainty in the movement process. Specifically, the velocity v, azimuth angle θ and pitch angle φ of each node are updated according to a specific random process, so that the maneuverability characteristics of the unmanned aerial platform in actual flight can be better reflected under the premise of ensuring the smoothness of movement. The node velocity update formula is as follows:

[0050]

[0051] Where α∈[0,1] is the memory factor, and are the mean values ​​of the velocity magnitude and direction set by the model, τ(t), ρ(t) and They represent the Gaussian random variables added when the velocity is updated.

[0052] When a node reaches the boundary of a region, a dynamic mean update strategy replaces the hard rebound mechanism. The following formula is used to adjust the mean angular value, allowing the UAV's direction of motion to continuously change in a differential form, progressively approaching the new mean value. This avoids sudden changes in the velocity vector and ensures that the node ultimately returns smoothly to the confined airspace.

[0053]

[0054] Based on the above physical entities and high dynamic FANET, the present invention proposes a distributed routing method for high dynamic mobile ad hoc networks, such as Figure 1 As shown, the following steps are included:

[0055] Step 1: Network initialization and node status awareness.

[0056] All nodes are powered on and their status information is initialized. Each node then obtains the location and speed information of its neighboring nodes every 1 second. This mechanism provides the necessary information support for the perception and maintenance of network topology.

[0057] To further enhance node perception and obtain more timely neighbor feature information, this paper adopts a communication system based on OFDM synaesthesia integrated waveform, which realizes neighbor node echo perception by using OFDM synaesthesia integrated waveform. Combined with frequency division duplex mode, each UFP node is assigned independent transmit and receive frequencies. The node is also equipped with three antenna systems, including:

[0058] a transmitting antenna for signal transmission;

[0059] a communication receiving antenna for receiving beacon information;

[0060] A sensing receiving antenna for processing echo signals.

[0061] This architectural design enables nodes to passively acquire neighbor characteristics through periodic beacon interactions, and to actively perceive through echo signal processing.

[0062] Step 2: Communication-perception dual-domain feature matching and fusion.

[0063] At time t, the node builds a communication domain neighbor list by parsing the received beacon packet and obtains N b The feature matrix of neighbor nodes is:

[0064] BF(t)={bf i (t),i=1,...,N b}

[0065] Where bf i (t) represents the feature vector corresponding to the i-th neighbor node in the communication domain, including the node’s Location information and Speed ​​information. i,k (t) represents the kth physical feature of the i-th neighbor node in the communication list. These features are strictly bound to the node's unique digital identifier (such as MAC address or IP address).

[0066] At the same time, the node constructs the perception domain neighbor list by processing the echo signal and obtains N s The feature matrix of neighbor nodes is:

[0067] SF(t)={sf i (t),i=1,...,N s}

[0068] Where sf i (t) represents the feature vector corresponding to the i-th neighbor node in the perception domain; sf j,m (t) represents the m-th dimension physical feature of the j-th neighbor node in the perception list.

[0069] Since the echo signal itself does not contain any digital identification information, the node cannot directly determine the specific source of the acquired perception features. Considering that the beacon packet usually contains identification fields such as the IP address of the sending node, and these identifications have a one-to-one correspondence with their physical features, the present invention calculates vector similarity by improving the probability measurement function of cosine similarity, and uses a bipartite graph model and convex optimization method to perform feature matching and fusion between the communication domain feature matrix BF(t) and the perception domain feature matrix SF(t), thereby achieving complete reconstruction and accurate identification of neighbor node information. The specific method is as follows:

[0070] At time t, the node extracts the communication domain feature vector of the neighbor node in the communication list and perception list respectively and the receptive domain feature vector Among them, K and M represent the feature dimensions of the two domains respectively.

[0071] In order to effectively measure the matching degree of dual-domain feature vectors, we first define a probability measurement function based on improved cosine similarity to calculate vector similarity:

[0072]

[0073] This function quantifies the directional similarity of the communication domain feature vector and the perception domain feature vector by calculating the ratio of the dot product and the Euclidean norm of the two vectors. The output value range is [-1,1]. The larger the value, the higher the probability that the two feature vectors come from the same unmanned aerial platform.

[0074] Based on the above feature vector similarity measurement, the present invention constructs the communication-perception dual-domain feature matrix matching optimization problem:

[0075] First, define the matching cost function as c t (i,j)=s t (i,j) -1 , used to measure the difference between two feature vectors.

[0076] Then, a dual-domain feature matrix matching model based on bipartite graph is established, such as Figure 2 As shown in Figure 2, the communication domain and perception domain feature sets constitute two vertex sets of a bipartite graph, and the edge weights are the corresponding matching costs. The goal of optimal matching is to find the feature vector pair with the minimum total weight.

[0077] In order to achieve the best match, we should strive to minimize the total cost of all successful matching pairs. At the same time, in order to ensure the rationality and feasibility of the matching results and avoid the imbalance of the overall matching effect due to excessively high local matching costs, it is also necessary to introduce a cost balance constraint to make the matching cost distribution of each successful matching vector pair balanced. In addition, in order to avoid cross-domain multi-matching conflicts (the vector of one domain becomes the best matching object for multiple vectors of another domain), a one-to-one matching constraint is also required, that is, each communication domain feature vector matches at most one perception domain feature vector. Therefore, the above optimization problem can be expressed as

[0078]

[0079] Where,

[0080]

[0081] f1 represents sub-problem 1, which is obtained by summing the cost c of all matching pairs. t (i, j) and take the average to calculate the overall matching cost. f2 represents sub-problem 2. This formula quantifies the distribution balance of matching costs by calculating the root mean square of the sum of squared deviations of the matching cost of each matching pair from the overall matching cost average f1. K=min(N b ,N s ), is the matching matrix, a t (i,j) indicates whether the i-th communication domain feature vector matches the j-th perception domain feature vector. When bf i (t) and sf j (t) matches, a t (i,j)=1; otherwise, a t (i,j)=0.

[0082] The above matching problem can be transformed into a convex optimization problem and solved by convex optimization method. t (i,j)∈{0,1}, introducing a slack variable set and impose constraints

[0083]

[0084] Adding the above constraint as a penalty factor to the original objective function, we can get the modified objective function

[0085]

[0086] Where,

[0087]

[0088] η is a penalty factor used to balance the weight relationship between the original objective function (f1+f2) and the newly added constraint (C).

[0089] Under the fixed {A} condition, the optimal B can be obtained by making full use of the first-order optimality condition:

[0090]

[0091] Finally, the CVX optimization toolkit is used to solve the modified objective function to obtain the optimal matching result.

[0092] This solution overcomes the limitations of traditional reliance solely on beacon communications, improving the real-time and reliability of network topology perception through the synergistic and complementary nature of communication and perception domain information. Perception domain information effectively compensates for potential packet loss or latency issues associated with beacon communications, while communication domain information provides the necessary identification and association for perception data. The two complement each other.

[0093] Step 3: Motion feature prediction based on Kalman filtering.

[0094] The node uses the Kalman filter to reduce noise and predict the motion state of neighboring nodes to improve the accuracy of state estimation. Specifically:

[0095] After completing the communication-perception dual-domain feature matching and fusion, the IP address of the neighbor node and its motion state (position, speed, etc.) observation value can be obtained. However, due to factors such as noise interference, data packet loss and high-speed movement of nodes in the wireless communication environment, these observation information often have significant measurement errors. Kalman filter (KF) is an unbiased, high-precision estimation and prediction tool widely used in fields such as image processing and target tracking. It can estimate the system state from a series of noisy, indirect measurements or inaccurate observations. In order to improve the accuracy of state estimation, the present invention uses KF to reduce noise and predict mobility of the observed neighbor motion information.

[0096] It is known that the UFP node adopts the enhanced three-dimensional GM mobility model, a(t)=[a x (t),a y (t),a z (t)] represents the acceleration of the node, so the state transition model and observation model are established as:

[0097]

[0098] Where x(t) = [p(t), v(t), a(t)] T The state vector of the node movement at time t, x(t+1) is the predicted state of x(t) at the next moment, y(t) is the observation vector of the node, and p(t) = [px (t),p y (t),p z (t)] represents the three-dimensional position; v(t) = [v x (t),v y (t),v z (t)] represents the velocity vector; and are the state transfer matrix and the observation matrix, respectively, and μ is the increment matrix. w(t) and z(t) are both zero-mean Gaussian noises, with covariance matrices Q and R.

[0099] According to the definition of the 3D GM motion model, the acceleration of the node is updated according to the following formula.

[0100]

[0101] Where α is the memory factor and λ(t-1) is the Gaussian random variable added when the acceleration is updated.

[0102] Therefore, the observation matrix is ​​H = [I 3×3 ,0 3×6 ], the state transfer matrix is:

[0103]

[0104] And the increment matrix is

[0105] Where Δt is the time step of velocity update.

[0106] According to the standard procedure of KF, the process of predicting and tracking the node motion state is as follows:

[0107] (1) Prediction status

[0108]

[0109] (2) Prediction mean square error

[0110] M(t|t-1)=FM(t-1)F T +Q.

[0111] (3) Calculate KF gain

[0112] K(t)=M(t|t-1)H T (R+HM(t|t-1)H T ) -1 .

[0113] (4) Tracking status

[0114]

[0115] (5) Update the estimated covariance matrix

[0116] M(t)=(IK(t)H)M(t|t-1).

[0117] After the above Kalman parameter update, KF has accumulated experience with the system and environment. The state and observation information used in the past do not need to be retained. Only the current state information, observation information and filter parameters need to be stored, and the node can obtain accurate motion state estimation values. Covered Locations and speed This data provides a reliable basis for calculating link quality evaluation indicators. For example, the relative position and speed of nodes can help predict the duration of a link, thereby reflecting the stability of the link. Their speed and position information can also be used to assess the reliability of signal transmission, which affects the effectiveness of the link.

[0118] Step 4: Calculation of link quality prediction indicators based on fuzzy logic.

[0119] The normalized predicted link lifetime (NPLL) and the normalized predicted expected transmission count (NPETC) are input into the fuzzy logic system. Through the steps of fuzzification, rule mapping and combination, and defuzzification, the predictive link quality parameters are generated. Specifically,

[0120] (1) Design the normalized predicted link lifetime as the core indicator for evaluating link stability.

[0121] For any communication link L ij , its link lifetime δ ij refers to node u j With node u i The period of time during which effective communication connection is maintained during the movement, i.e., during δ ij Within the time, the relative distance between the two nodes does not exceed the communication radius R, satisfying:

[0122] ||p i (t)-p j (t)+v i (t)-v j (t)·δ ij (t)||2-R≤0

[0123] Obviously, the longer the link lifetime is, the higher the stability is. To simplify the formula, define the relative distance d ij (t) = pi (t)-p j (t) and relative speed s ij (t) = v i (t)-v j (t), thus rewriting the constraint condition. Square both sides of the inequality and expand it to obtain a ij (t) is a quadratic inequality. Then when s ij When (t)2≠0, the link lifetime satisfies the inequality:

[0124]

[0125] Furthermore, according to the Cauchy-Schwarz inequality principle, we can deduce When the relative speed direction of the nodes is opposite to the distance direction and the relative distance reaches the communication radius, the link lifetime reaches the maximum value. In order to eliminate the dimension effect and obtain comparable indicators, we first determine the node u i Maximum lifetime of all neighbor links Then define the normalized predicted link survival time as:

[0126]

[0127] This indicator not only enables standardized comparison between different links, but also intuitively reflects the link stability level: when NPLL is close to 1, it indicates optimal link quality, while NPLL approaching 0 indicates that the link is about to be disconnected.

[0128] (2) Design a normalized predicted expected transmission number to evaluate the communication effectiveness of the link.

[0129] During communication, to ensure that the signal can be correctly received and decoded, the received signal power must exceed the receiver sensitivity threshold. Otherwise, communication will be interrupted due to poor signal quality, causing multiple data retransmissions until the data packet is successfully received. Therefore, the number of retransmissions directly reflects the link communication quality.

[0130] Based on the wireless propagation model adopted by the present invention, it can be obtained that the node u j Through link L ij The probability density function of the received signal amplitude r is Therefore, the link L ij The transmission interruption probability can be expressed as:

[0131]

[0132] Where γ(·,·) is the incomplete gamma function; P r (d ij (t)) represents the propagation distance dij The received power at time (t), p() represents the conditional probability density function of r under the condition of known received power.

[0133] Given the expression for the interruption probability, let the random variable X be the number of transmissions required for a successful communication, and its cumulative distribution function is:

[0134]

[0135] Link L ij The expected number of transmissions is calculated as The value range is [1, +∞). The more the link retransmission times are, the worse the communication quality is. To unify the dimensions, the normalized predicted expected transmission times κ is used. ij (t) Measure the effectiveness of link communication:

[0136] κ ij (t)=1-P out (d ij (t))

[0137] Link lifetime and expected number of packet transmissions characterize link stability and transmission efficiency, respectively. However, a single indicator is insufficient to comprehensively assess link quality. By defining reasonable fuzzy rules and membership functions, the fuzzy logic system (FLS) can organically integrate these two key indicators and effectively address the uncertainties introduced by factors such as channel state fluctuations and random interference in wireless communication environments, generating a more comprehensive and dynamic comprehensive evaluation of link quality. Therefore, the present invention introduces a fuzzy logic-based comprehensive link quality evaluation system, which processes:

[0138] 1) Fuzzification: The two link evaluation indicators are used as input variables and the numerical results are converted into fuzzy values ​​through fuzzy membership functions, such as Figure 3a 、 3b As shown, the present invention uses double triangles as fuzzy membership functions.

[0139] 2) Mapping and Combining IF / THEN Rules: The two fuzzy values ​​after the transformation in 1) are mapped using IF / THEN rules, categorizing the link mapping into four levels: Poor, Fair, Good, and Excellent, as shown in Table 1. During the rule matching process, multiple rules may be applied simultaneously. In such cases, the present invention uses a min-max method for rule selection, where the minimum value is used for the rule and the maximum value is used for the final result.

[0140] Table 1 IF / THEN fuzzy rules

[0141]

[0142] 3) Defuzzification: The process of converting the rule mapping result into a numerical result by taking the output membership function as the immediate utility function to obtain the quantitative predicted link quality (PLQ), such as Figure 4 shown.

[0143] Step 5: Route discovery phase based on node early warning mechanism.

[0144] Due to the lack of central or regional location servers in current FANETs, ​​UAV nodes cannot directly locate their destination nodes. Therefore, this paper adopts an on-demand routing paradigm—dynamically discovering network topology by flooding routing control packets on demand—to construct a link quality prediction routing protocol that integrates communication and perception.

[0145] ① Routing control packet format

[0146] In the on-demand routing paradigm, there are three types of routing control packets: RREQ packets, RREP packets, and RERR packets. To implement link quality prediction routing protocols, a special PLQ field is added to the route request packet. It represents the quality bottleneck value of all links on the current RREQ propagation path. This field records the minimum PLQ value of the path in real time during the RREQ forwarding process, providing a key criterion for subsequent path optimization.

[0147] ②RREQ forwarding mechanism based on node risk assessment

[0148] In on-demand routing protocols, the traditional flooding route discovery mechanism, while simple and effective, is prone to broadcast storms, which can cause network resources to be overwhelmed by a large number of invalid RREQ messages. Given the limited energy reserves and cache capacity of flying nodes, if nodes that are about to fail continue to participate in route discovery, this not only exacerbates network congestion but also significantly increases control overhead and routing latency, severely impacting route establishment efficiency.

[0149] This protocol proposes a lightweight node early warning mechanism, which monitors its own remaining energy E rem and cache queue length Q, quickly identify high-risk nodes, and selectively forward RREQ messages. Specifically, the system defines a low energy warning threshold E th and high congestion warning threshold Q th , when the remaining energy of the node is lower than E th Or the cache queue length exceeds Q th When the node is in a high-risk state, it will be marked as a high-risk state and will actively discard the RREQ message to avoid participating in an invalid route discovery process.

[0150] The thresholds of this early warning mechanism can be flexibly configured based on network characteristics and task requirements, either using a preset static value or adaptively adjusting based on network dynamics. Compared to traditional methods, this mechanism does not require expanding the RREQ message format, thus avoiding increased routing overhead. This significantly reduces computational complexity while maintaining assessment accuracy. By suppressing RREQ forwarding at high-risk nodes, it effectively reduces network congestion caused by flooding and improves routing stability and efficiency.

[0151] ③ On-demand link quality prediction routing establishment.

[0152] When the source node detects a communication demand but there is no valid path in the routing table, the routing search mechanism will be triggered and the RREQ message will be flooded. After receiving the RREQ, the intermediate node will first th and cache threshold Q th Perform risk assessment: If it is determined to be a low-risk node, further check the local routing table; if a valid path exists, directly unicast reply RREP, otherwise continue to forward or discard the request according to the redundancy suppression policy.

[0153] When the destination node receives the RREQ, it will dynamically update its PLQ and Hop fields and calculate its path utility (PU). The calculation of path utility not only needs to evaluate the overall quality of the link, but also needs to consider the impact of the number of hops in the entire route. p with u q The link weight parameter between the two nodes is denoted as ω(p,q), and its value range is [0,1]. A larger value indicates a more stable and reliable link. The more hops a route takes from the source node to the destination node, the longer the required delay and the higher the probability of packet loss, which in turn reduces the path quality. Therefore, for a route with m hops, the path utility is calculated as follows:

[0154]

[0155] The destination node sets a dynamic waiting time window when receiving RREQs to balance route discovery speed and path quality. If the waiting time is too short, route establishment can be accelerated, but a better path may be missed. Conversely, extending the waiting time increases the probability of finding a high-quality path, but increases initial latency. Therefore, this invention defines a method for adjusting the destination node's waiting time:

[0156]

[0157] Where, t L represents the waiting time, t F is the arrival delay of the first RREQ, and a represents the adjustable delay coefficient.

[0158] After the waiting time expires, the destination node selects the path with the highest PU value from all candidate paths and unicasts a RREP along that path. After receiving the RREP, the source node completes route establishment and initiates data transmission. This process follows the principles of a passive routing algorithm, synchronously building a bidirectional routing topology between the source and destination nodes. Relay nodes along the corresponding path store routing information in their routing tables. Once the route is established, the source node initiates data transmission.

[0159] Step 6: Route maintenance.

[0160] During the route maintenance phase, fault pre-detection and fast path recovery mechanisms are introduced to improve network robustness. Given the highly dynamic nature of FANET, routes established during the route discovery phase are at risk of failure. This phase aims to enable immediate detection and rapid replacement of failed routes, thereby ensuring continuity of network communications.

[0161] (1) Routing Failure Prevention

[0162] By monitoring node status (e.g., energy reserves, cache space, etc.) in real time, the system proactively identifies and corrects routing paths approaching failure thresholds to prevent data transmission interruptions. When a node detects that its resources have reached a preset threshold, it immediately scans its local routing table, identifies all potentially affected routes, and sends a RERR error message to the relevant previous hop. The system then automatically triggers the route discovery process, completing route updates before a path completely fails, ensuring communication continuity.

[0163] (2) Rapid route reconstruction

[0164] When an intermediate node detects a next-hop link interruption, it immediately sends a RERR error message to the source node, notifying it of the route failure. Simultaneously, the system triggers the route discovery process, recalculates the optimal path from the current node to the destination node, and quickly establishes an alternative route to resume data transmission.

[0165] The highly dynamic mobile ad hoc network interawareness integrated routing device based on the above method adopts a layered modular architecture design, supports distributed routing decisions and efficient communication, integrates various functional modules, and works together to cope with highly dynamic network environments, significantly improving network throughput, resource utilization, and stability. It is suitable for highly dynamic scenarios such as unmanned aerial platform formations, vehicle-mounted ad hoc networks, and emergency communication networks. Figure 5 As shown in the figure, the interactive relationship between modules and the information exchange between node devices are shown. In this layered modular architecture:

[0166] The upper layer is the application module, responsible for generating data transmission requests based on user or system task requirements, scheduling the middle-layer routing module to determine the optimal transmission path, and then initiating data transmission. The middle layer is the routing module, responsible for analyzing network topology changes and node dynamics, predicting link quality and stability, providing a basis for routing decisions, and performing node risk assessment and path utility calculation. Specifically, it includes a signal domain feature extraction module, a perception domain feature extraction module, a communication-perception domain information matching module, a Kalman filter, a link quality prediction module, and an on-demand routing module. The bottom layer is the communication module, responsible for basic data transmission, and accurately identifying neighbor node information through dual-domain information matching.

[0167] The upper, middle, and lower layers of the architecture are interdependent and collaborative: the upper-layer application module generates communication requests based on demand and dispatches the middle-layer routing module to determine the transmission path. The middle-layer routing module analyzes network status, makes routing decisions, and guides the lower-layer communication module for data transmission. The lower-layer communication module is responsible for basic data transmission and node information identification. This layered architecture supports distributed routing decisions and efficient communication, adapting to mobile and dynamic network environments.

[0168] In the above middle layer, the communication domain feature extraction module and the perception domain feature extraction module are used to extract dual-domain features of the communication domain and the perception domain, respectively; further, the communication-perception domain information matching module uses the beacon packet of the communication domain and the echo signal of the perception domain, combined with the probability measurement function of the improved cosine similarity and the bipartite graph model to achieve efficient feature matching and fusion, and realize the complete reconstruction and accurate identification of neighbor node information.

[0169] The Kalman filter is used to reduce noise and predict the motion state of neighboring nodes, thereby improving the estimation accuracy of the node motion state.

[0170] The Link Quality Prediction module dynamically assesses link quality, supporting efficient link quality assessment in dynamic network environments. This module normalizes the predicted link lifetime and the normalized predicted expected transmission times, inputting them into the fuzzy logic module to generate predictive link quality parameters that quantify the link's stability and effectiveness. These parameters are then output by the Link Quality Prediction module, providing precise prior knowledge for route selection and enhancing the robustness and agility of route discovery.

[0171] The on-demand routing module combines the on-demand routing paradigm with a routing discovery mechanism based on a node early warning mechanism. It significantly improves routing stability and network robustness through link quality prediction, fault pre-detection, and fast path recovery mechanisms.

[0172] The present invention also provides a computer-readable storage medium storing a computer program for implementing the aforementioned routing method. When executed by a processor, the program can implement any of the aforementioned steps in the synaesthesia routing method for flight ad hoc networks. The aforementioned routing method can be implemented as a software functional unit and sold or used as a standalone product. Users can implement the synaesthesia routing forwarding function by installing and executing the software.

[0173] The computer-readable storage medium includes but is not limited to the following forms:

[0174] (1) Physical storage media

[0175] ●USB flash drive: A portable storage device suitable for program distribution on a small network.

[0176] ●Mobile hard disk: A large-capacity storage device suitable for storing programs and data on large-scale networks.

[0177] ●Disk: Traditional storage medium, suitable for offline storage.

[0178] ●Optical discs: including CDs, DVDs, Blu-ray discs, etc., suitable for program distribution and long-term storage.

[0179] ●Computer memory: includes RAM and ROM, used for temporary storage and quick access to program code.

[0180] (2) Virtual storage media

[0181] ●Electrical carrier signal: Program code is transmitted via electrical signals.

[0182] ●Telecommunication signals: Transmitting program code via wireless or wired communication networks.

[0183] ●Software distribution media: Distribute program code via the Internet or local area network.

Claims

1. A highly dynamic mobile ad hoc network synaesthesia integrated routing method, characterized by: The steps are: Step 1: After all nodes initialize their state information, they obtain the location and velocity information of neighboring nodes. Furthermore, they use OFDM synaesthesia integrated waveform to realize neighboring node echo perception. Step 2: Calculate vector similarity by improving the probability measurement function of cosine similarity. Use the bipartite graph model and convex optimization method to perform feature matching and fusion between the communication domain feature matrix BF(t) and the perception domain feature matrix SF(t), achieving complete reconstruction and accurate identification of neighbor node information. Step 3: Each node uses the Kalman filter to reduce noise and predict the motion state of neighboring nodes to improve the accuracy of state estimation; Step 4: Calculate link quality prediction indicators based on fuzzy logic. Define the normalized predicted link lifetime as the core indicator for evaluating link stability. Define the normalized predicted expected transmission times to evaluate the communication effectiveness of the link. Furthermore, the normalized predicted link lifetime and the normalized predicted expected number of transmissions are input into the fuzzy logic system, and the predictive link quality parameters are generated through fuzzification, rule mapping and combination, and defuzzification. Step 5: In the routing discovery phase based on the node early warning mechanism, an on-demand routing paradigm is adopted. The network topology is dynamically discovered by flooding routing control packets on demand. A link quality prediction routing protocol integrating communication perception is constructed. At the same time, a lightweight node early warning mechanism is designed to monitor its own remaining energy E. rem and cache queue length Q, quickly identifying high-risk nodes and selectively forwarding RREQ messages; Step 6: During the routing maintenance phase, fault pre-detection and fast path recovery mechanisms are introduced to improve network robustness.

2. The highly dynamic mobile ad hoc network synaesthesia integrated routing method according to claim 1, characterized in that: In step 1, independent transmit and receive frequencies are allocated to each UFP node in combination with the frequency division duplex mode. The node is also equipped with three antenna systems, including a transmit antenna for signal transmission, a communication receive antenna for receiving beacon information, and a sensing receive antenna for processing echo signals.

3. The highly dynamic mobile ad hoc network synaesthesia integrated routing method according to claim 1, characterized in that: The specific method for step 2 is: Define a probability metric function based on improved cosine similarity to calculate vector similarity: Based on the above feature vector similarity measurement, the communication-perception dual-domain feature matrix matching optimization problem is constructed as follows: First, define the matching cost function as c t (i,j)=s t (i,j) -1 , used to measure the difference between two feature vectors; Subsequently, a dual-domain feature matrix matching model based on a bipartite graph is established. The communication domain and perception domain feature sets constitute the two vertex sets of the bipartite graph respectively, and the weight of the edge is the corresponding matching cost. The goal of optimal matching is to find the feature vector pair with the minimum total weight. Furthermore, to achieve the best match, an optimization matching problem is proposed, which includes: minimizing the total cost of all successful matching pairs; introducing a cost balance constraint to ensure that the matching cost distribution of each successful matching vector pair is balanced; and each communication domain feature vector matches at most one perception domain feature vector. The optimization problem is expressed as: Where, f1 represents sub-problem 1, which is obtained by summing the cost c of all matching pairs. t (i, j) and take the average value to calculate the overall matching cost; f2 represents sub-problem 2. This formula quantifies the distribution balance of the matching cost by calculating the root mean square of the sum of the square deviations of the matching cost of each matching pair and the overall matching cost average value f1; K = min(N b ,N s ), is the matching matrix, a t (i,j) indicates whether the i-th communication domain feature vector matches the j-th perception domain feature vector. When bf i (t) and sf j (t) matches, a t (i,j)=1; otherwise, a t (i,j)=0; The above matching problem can be transformed into a convex optimization problem and solved by convex optimization method; in order to deal with the binary constraint a t (i,j)∈{0,1}, introducing a slack variable set and impose constraints Adding the above constraint as a penalty factor to the original objective function, the corrected objective function is: Where, η is the penalty factor, which is used to balance the weight relationship between the original objective function (f1+f2) and the newly added constraint (C); Under the fixed {A} condition, the optimal B can be obtained by making full use of the first-order optimality condition: Finally, the CVX optimization toolkit is used to solve the modified objective function to obtain the optimal matching result.

4. The highly dynamic mobile ad hoc network synaesthesia integrated routing method according to claim 1, characterized in that: The specific method of step 3 is: The state transition model and observation model are established as follows: Where x(t) = [p(t), v(t), a(t)] T The state vector of the node movement at time t, x(t+1) is the predicted state of x(t) at the next moment, y(t) is the observation vector of the node, and p(t) = [p x (t),p y (t),p z (t)] represents the three-dimensional position; v(t) = [v x (t),v y (t),v z (t)] represents the velocity vector; and are the state transfer matrix and the observation matrix respectively, μ is the increment matrix; w(t) and z(t) are both zero-mean Gaussian distributed noise, and the covariance matrices are Q and R; According to the definition of the 3D GM motion model, the node acceleration is updated according to the following formula; Where α is the memory factor, and λ(t-1) is the Gaussian random variable added when the acceleration is updated; Therefore, the observation matrix is ​​H = [I 3×3 ,0 3×6 ], the state transfer matrix is: And the increment matrix is Where Δt is the time step of velocity update; According to the standard procedure of KF, the process of predicting and tracking the node motion state is as follows: (1) Prediction status (2) Prediction mean square error M(t|t-1)=FM(t-1)F T +Q. (3) Calculate KF gain K(t)=M(t|t-1)H T (R+HM(t|t-1)H T ) -1 . (4) Tracking status (5) Update the estimated covariance matrix M(t)=(IK(t)H)M(t|t-1).

5. The highly dynamic mobile ad hoc network synaesthesia integrated routing method according to claim 1, characterized in that: In step 4: The design method of normalized predicted link lifetime as the core indicator for evaluating link stability is as follows: For any communication link L ij , its link lifetime δ ij refers to node u j With node u i The period of time during which effective communication connection is maintained during the movement, i.e., during δ ij Within the time, the relative distance between the two nodes does not exceed the communication radius R, satisfying: ||p i (t)-p j (t)+v i (t)-v j (t)·δ ij (t)||2-R≤0. Define the relative distance d ij (t) = p i (t)-p j (t) and relative speed s ij (t) = v i (t)-v j (t), rewrite the constraints, square both sides of the inequality and expand them to obtain an equation about δ ij (t) of the quadratic inequality; then when s ij When (t)2≠0, the link lifetime satisfies the inequality: Furthermore, according to the Cauchy-Schwarz inequality principle, we can deduce When the relative speed direction of the nodes is opposite to the distance direction and the relative distance reaches the communication radius, the link lifetime reaches the maximum value. In order to eliminate the dimension effect and obtain comparable indicators, we first determine the node u i Maximum lifetime of all neighbor links Then define the normalized predicted link survival time as: The design method for normalized prediction of expected transmission times is: The wireless propagation model used can be obtained, node u j Through link L ij The probability density function of the received signal amplitude r is Therefore, the link L ij The transmission interruption probability is expressed as: Where γ(·,·) is the incomplete gamma function; P r (d ij (t)) represents the propagation distance d ij The received power at time (t), p() represents the conditional probability density function of r under the condition of known received power; Given the expression for the interruption probability, let the random variable X be the number of transmissions required for a successful communication, and its cumulative distribution function is: Link L ij The expected number of transmissions is calculated as The value range is [1, +∞). The more the link data retransmission times are, the worse the communication quality is. To unify the dimensions, the normalized predicted expected transmission times κ is used. ij (t) Measure the effectiveness of link communication: κ ij (t)=1-P out (d ij (t))。 6. The highly dynamic mobile ad hoc network synaesthesia integrated routing method according to claim 1, characterized in that: Step 4: The fuzzy logic system processing method is: 1) Fuzzification The two link evaluation indicators, normalized predicted link lifetime and normalized predicted expected transmission times, are used as input variables and the numerical results are converted into fuzzy values ​​through fuzzy membership function. 2) Mapping and combination of IF / THEN rules The two fuzzy values ​​after the transformation in step 1 are mapped using IF / THEN rules, and the link mapping is divided into four levels: Poor, Fair, Good, and Excellent. In the rule matching process, the min-max method is used for rule selection, where the minimum value is used for the rule and the maximum value is used for the final result. 3) Defuzzification The output membership function is used as the immediate utility function to convert the rule mapping results into numerical results to obtain the quantitative predicted link quality.

7. The highly dynamic mobile ad hoc network synaesthesia integrated routing method according to claim 1, characterized in that: The routing control packet in step 5 includes an RREQ packet, an RREP packet, and a RERR packet. A new PLQ field is added to the RREQ packet to represent the quality bottleneck value of all links on the current RREQ propagation path. This field records the minimum PLQ value of the path in real time during the RREQ forwarding process, providing a key criterion for subsequent path optimization.

8. The highly dynamic mobile ad hoc network synaesthesia integrated routing method according to claim 1, characterized in that: The specific method of step 5 is: When the source node detects a communication demand but there is no valid path in the routing table, the routing search mechanism will be triggered and the RREQ message will be flooded. After the intermediate node receives the RREQ, it will firstly th and cache threshold Q th Perform risk assessment: If the node is determined to be low-risk, further check the local routing table. If a valid path exists, a direct unicast RREP reply is sent. Otherwise, the request is forwarded or discarded based on the redundancy suppression policy. When the destination node receives the RREQ, it dynamically updates its PLQ and Hop fields and calculates its path utility; Node u p with u q The link weight parameter between them is expressed as ω(p,q), and its value range is [0,1]. The larger the value, the more stable and reliable the link. For a route with m hops, the path utility calculation formula is as follows: The destination node sets a dynamic waiting time window when receiving RREQ to balance the route discovery speed and path quality; The method for adjusting the destination node waiting time is: Where, t L represents the waiting time, t F is the arrival delay of the first RREQ, a represents the adjustable delay coefficient; After the waiting time expires, the destination node selects the path with the highest PU value from all candidate paths and unicasts the RREP back along that path. After receiving the RREP, the source node completes the route establishment and starts data transmission. This process follows the principles of the passive routing algorithm and simultaneously builds a bidirectional routing topology between the source and destination nodes. The relay nodes on the corresponding path store the routing information in the routing table. After the route is established, the source node starts data transmission.

9. The highly dynamic mobile ad hoc network synaesthesia integrated routing method according to claim 1, characterized in that: In step 6, by monitoring node status in real time, routing paths approaching the failure threshold are proactively identified and corrected to prevent data transmission interruptions. When a node detects that its resources have reached a preset threshold, it immediately scans the local routing table, identifies all potentially affected routes, and sends a RERR error message to the relevant previous hop node. The system then automatically triggers the route discovery process to complete the route update before the path completely fails, ensuring communication continuity. When an intermediate node detects that the next-hop link is interrupted, it will immediately send a RERR error message to the source node to notify the route failure; at the same time, the system synchronously triggers the route discovery process, recalculates the optimal path from the current node to the target node, and quickly establishes an alternative route to resume data transmission.

10. The device for implementing the highly dynamic mobile ad hoc network synaesthesia integrated routing method according to claim 1 is characterized in that: Adopting a layered modular architecture design, it supports distributed routing decisions and efficient communication, integrates various functional modules, and works together to cope with highly dynamic network environments. In the layered modular architecture: The upper layer is the application module, which is responsible for generating data transmission requests based on user or system task requirements, scheduling the middle-layer routing module to determine the optimal transmission path, and then initiating data transmission; the middle layer is the routing module, which is responsible for analyzing network topology changes and node dynamics, predicting the quality and stability of the link, providing a basis for routing decisions, and performing node risk assessment and path utility calculation. It specifically includes the signal domain feature extraction module, the perception domain feature extraction module, the communication-perception domain information matching module, the Kalman filter, the link quality prediction module, and the on-demand routing module; the bottom layer is the communication module, which is responsible for basic data transmission and realizes the accurate identification of neighbor node information through dual-domain information matching; The upper, middle, and lower layers are interdependent and collaborative: the upper application module generates communication requests based on demand and schedules the middle routing module to determine the transmission path; the middle routing module analyzes the network status, makes routing decisions, and guides the lower communication module for data transmission; the lower communication module is responsible for basic data transmission and node information identification; The communication domain feature extraction module and the perception domain feature extraction module are used to extract dual-domain features of the communication domain and the perception domain, respectively. Furthermore, the communication-perception domain information matching module uses the beacon packets of the communication domain and the echo signals of the perception domain, combined with the probability measurement function of the improved cosine similarity and the bipartite graph model, to achieve efficient feature matching and fusion, and realize the complete reconstruction and accurate identification of neighbor node information. The Kalman filter is used to reduce noise and predict the motion state of neighboring nodes, improving the estimation accuracy of node motion state; The link quality prediction module dynamically assesses link quality, supporting efficient link quality assessment in dynamic network environments. This module normalizes the predicted link lifetime and the normalized predicted expected transmission times, inputs them into the fuzzy logic module, and generates predictive link quality parameters to quantify the stability and effectiveness of the link. The output of the link quality prediction module quantifies the stability and effectiveness of the link, providing accurate prior knowledge for route selection and enhancing the robustness and agility of route discovery. The on-demand routing module combines the on-demand routing paradigm with a routing discovery mechanism based on a node early warning mechanism. It significantly improves routing stability and network robustness through link quality prediction, fault pre-detection, and fast path recovery mechanisms.

11. A device for implementing the highly dynamic mobile ad hoc network synaesthesia integrated routing method according to claim 1, characterized in that: A computer-readable storage medium storing a computer program implemented in the form of a software functional unit; when the program is executed by a processor, it can implement any of the steps in the aforementioned synaesthesia routing method for a flying ad hoc network; the computer-readable storage medium includes a physical storage medium and a virtual storage medium; wherein the virtual storage medium is an electric carrier signal, a telecommunication signal, or a software distribution medium.

Images