Mobile ad hoc network data routing method and industrial ad hoc network system

Abstract

The invention provides a mobile ad hoc network data routing method, which comprises the following steps that: S1, each node periodically broadcasts an EC message, and the EC message comprises the current position of the node which broadcasts the EC message, the position of a future prediction moment and an initial routing evaluation score; s2, when each node receives one EC message broadcasted by other nodes, executing the following steps: S21, calculating expected transmission times between the node and other nodes based on the received EC message broadcasted by other nodes; s22, calculating a route evaluation score between the node and other nodes according to a preset mode based on the expected transmission times of the node and other nodes, and storing the route evaluation score; s23, the route evaluation score in the EC message received in the step S21 is replaced by the route evaluation score in the step S22, and then the EC message is broadcasted; and S3, each node selects the node with the highest route evaluation score from all the single-hop neighbor nodes as the next-hop node.

Description

Technical Field

[0001] The present invention relates to the fields of wireless communication technology and computer technology, in particular to the field of network layer protocols for self-organizing networks, and more particularly to an automatic self-organizing network data routing method and an industrial self-organizing network system. Background Art

[0002] A mobile ad hoc network consists of mobile nodes. Each node can be either a source or destination node for data, and each node can help other nodes forward data. Because node transmission range is limited, data typically requires multiple hops to reach its destination. The process by which a data packet travels from the source node through several nodes to the destination is called data routing. The protocol that controls the selection of forwarding nodes by nodes holding data packets is called a routing method. Routing methods that ignore node mobility can result in packet loss when nodes move. This is because node mobility can cause wireless links to disconnect, making the link state information perceived by the node in the previous sensing cycle outdated. Nodes then send data to disconnected nodes, causing packet loss. Routing methods based on motion prediction anticipate node movement and, by introducing motion prediction, avoid transmitting data over inoperative wireless links, thus reducing packet loss. However, existing routing methods are ineffective in scenarios with obstructions. These methods only predict inter-node distances and use fixed node transmission ranges to estimate link lifespans. In scenarios with obstructions, the transmission range of nodes depends on inter-node channel conditions. For example, the transmission range of nodes in line-of-sight (LOS) conditions is much greater than that in non-line-of-sight (NLOS) conditions.

[0003] Existing mobile ad hoc network routing methods can be classified into two categories according to the link state perception method: on-demand perception routing method and periodic perception routing method. Among them, the on-demand perception routing method, such as the solution recorded in patent application CN119815458A, only starts sending and receiving signals to perceive the link state when it needs to send data. In the on-demand perception routing method, the data source node needs to wait for the link state information to be returned before sending data, which will cause a startup delay and increase the communication delay. The periodic perception routing method, such as the solution recorded in patent application CN119789180A, each node periodically sends and receives signals to maintain the link state, and there will be no startup delay. However, the periodic perception routing method will cause packet loss when the node moves. This is because the movement of the node will cause the wireless link to be disconnected, and the link state information perceived by the node in the previous perception cycle will be outdated. The node sends data to the disconnected node, resulting in packet loss. To address the packet loss problem of periodic perception routing methods, a periodic perception routing method based on motion prediction has emerged. As described in patent application CN114585042B, node movement is predicted and motion prediction is introduced to avoid using expired wireless links to transmit data, thereby reducing packet loss. However, in scenarios with obstructions, existing routing methods based on motion prediction are no longer applicable because these routing methods only predict the distance between nodes and use a fixed node transmission range value to estimate the link validity time. However, in scenarios with obstructions, the node transmission range depends on the channel conditions between nodes. For example, the node transmission range under line-of-sight conditions is much larger than the node transmission range under non-line-of-sight conditions. In this case, only considering the distance between nodes without considering the channel conditions between nodes will cause the link state information to become outdated, and packet loss or delay in data routing will increase.

[0004] The Industrial Internet is a common mobile ad hoc network scenario. Wirelessly controlled automated guided vehicles (AGVs) are a key component of the Industrial Internet. In specialized factories where explosive gases or liquids are present, wireless transmitter power is limited. Automated guided vehicles (AGVs) must be controlled to form low-power mobile ad hoc networks to complete industrial production tasks. In mobile ad hoc networks, there are no fixed control nodes to control data routing. Nodes holding data can only control data forwarding for the current hop, not subsequent data forwarding. Each node locally senses and estimates the optimal data routing path and then forwards data to the next hop on its estimated optimal data routing path. The actual data routing path is determined hop by hop by each node along the path. Existing routing methods that account for node mobility are no longer applicable in environments with obstructions. These methods only estimate inter-node distances and use fixed node transmission ranges to estimate link lifespans. However, in scenarios with obstructions, node transmission ranges depend on inter-node channel conditions. For example, node transmission ranges under line-of-sight conditions are much greater than those under non-line-of-sight conditions.

[0005] It should be noted that this background information is provided solely to introduce relevant information of the present invention to facilitate understanding of the technical solution of the present invention. It does not necessarily constitute prior art. In the absence of evidence demonstrating that the relevant information was disclosed prior to the filing date of the present invention, the relevant information should not be considered prior art. Summary of the Invention

[0006] Therefore, the purpose of the present invention is to overcome the above-mentioned defects of the prior art and provide a new mobile ad hoc network data routing method and an industrial ad hoc network system.

[0007] According to a first aspect of the present invention, a mobile ad hoc network data routing method is provided for a mobile node network, wherein the mobile node network includes a plurality of mobile nodes, each node being a device having a wireless communication function, wherein each mobile node is configured with a pre-trained channel condition prediction model, and the method comprises: S1, each node periodically broadcasts an EC message, wherein the EC message includes the current position of the node broadcasting the EC message, the position at a future predicted moment, and an initial routing evaluation score; S2, each node performs the following steps each time it receives an EC message broadcast from another node: S21, calculates its own current position based on the received EC message broadcast from the other node , receive the signal-to-noise ratio and the position at the predicted future time, and use the channel condition prediction model to predict the channel conditions between itself and other nodes and calculate the expected number of transmissions between itself and other nodes based on the channel conditions; S22, calculate the routing evaluation score between itself and other nodes in a preset manner based on the expected number of transmissions between itself and other nodes and save the routing evaluation score; S23, replace the routing evaluation score in the EC message received in step S21 with the routing evaluation score in step S22 and then broadcast the EC message; S3, each node selects the node with the highest routing evaluation score from all single-hop neighbor nodes as the next hop node based on the routing evaluation score between itself and other nodes saved by itself.

[0008] Preferably, the channel condition prediction model is pre-trained in the following manner: T1, collecting multiple original data sets with each node as a sending node, each original data set corresponds to a node, wherein the original data set corresponding to each node contains multiple original data, each original data contains the position of another node as a sending node, the position of the node as a receiving node, and the receiving signal-to-noise ratio; T2, for the original data set of each node, calculating whether the channel condition between the two nodes in each original data is a line-of-sight path or a non-line-of-sight path in a preset manner, and taking the position of the sending node and the position of the receiving node in each original data as samples and the channel condition between the two nodes as labels, constructing each The initial training set corresponding to each node is obtained; T3, the initial training set corresponding to each node is expanded, wherein the position of the receiving node in the sample labeled as non-line-of-sight path is expanded to the boundary of the site to obtain a new sample, and the label of the new sample is non-line-of-sight path; T4, the spatial coverage of the initial training set after expansion of each node in the site is evaluated, and the training set with the highest spatial coverage is selected as the final model training set; T5, based on the model training set selected in step T4, at the node where the training set is located, with the position of the sending node and the position of the receiving node as input and the channel condition between the two nodes as output, the fully connected neural network is trained until convergence to obtain a channel condition prediction model, and the model is distributed to other nodes.

[0009] Preferably, the preset method is: calculating the line-of-sight signal-to-noise ratio threshold and the non-line-of-sight signal-to-noise ratio threshold based on channel fading in the mobile ad hoc network scenario; calculating the channel condition between the sending node and the receiving node according to the received signal-to-noise ratio of the receiving node, wherein when the received signal-to-noise ratio is greater than or equal to the line-of-sight signal-to-noise ratio threshold, the channel condition is the line-of-sight path; when the received signal-to-noise ratio is less than or equal to the non-line-of-sight signal-to-noise ratio threshold, the channel condition is the non-line-of-sight path.

[0010] Preferably, in step T4, the spatial coverage of the expanded training set of each node in the venue is calculated in the following manner: T41, dividing the mobile ad hoc network scene into multiple grids of equal area; T42, for the initial training set after each node is expanded, sampling multiple points on the line connecting two nodes in each data to obtain a sampling point set corresponding to each node; T43, counting the number of grids covered by the sampling point set corresponding to each node, wherein if a grid has any number of sampling points in the sampling point set, the grid is determined to be covered by the sampling point set; T44, calculating the spatial coverage of the initial training set after the node is expanded based on the ratio of the number of grids covered by the sampling point set corresponding to each node to the total number of grids.

[0011] Preferably, the number of hidden layers of the fully connected neural network is 3, a binary cross entropy loss function is adopted, and the momentum gradient descent algorithm is used to optimize the parameters of the fully connected neural network through back propagation.

[0012] Preferably, in step S21, each node calculates the expected number of transmissions between itself and other nodes in the following manner:

[0013]

[0014] in, Representation node The current position at time t, Representation node future Time position, represents the prediction time domain, Indicates the EC message broadcast period, is the preset maximum transmission number threshold, The wireless link between node j and node i is The probability of interruption at time , represents the sampling period, is the prediction of node j The position of node i at time instant, is the prediction of node j The position of node j at time instant, where:

[0015]

[0016]

[0017]

[0018] in, represents the channel condition between node i and node j predicted by the channel condition prediction model.

[0019] Preferably, in step S22, each node calculates the route evaluation score in the following manner:

[0020]

[0021] in, represents the routing evaluation score between node j and node i, represents the initial route evaluation score in the EC message sent by node i, is a constant and .

[0022] According to a second aspect of the present invention, an industrial ad hoc network system is provided, comprising a plurality of mobile guide vehicles, each guide vehicle being a wireless node, and the industrial ad hoc network system is configured to perform data routing using the method described in the first aspect of the present invention.

[0023] Compared with the existing technology, the solution proposed in the present invention is based on the construction of the line-of-sight path map using a fully connected neural network. The line-of-sight path map is obtained by constructing a training data set through sending and receiving routing messages. When a node needs to send data to another node, the node applies a data routing protocol based on the line-of-sight path map to implement multi-hop forwarding and complete inter-vehicle communication. It is well suited for self-organizing network routing construction in obstructed environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The embodiments of the present invention are further described below with reference to the accompanying drawings, in which:

[0025] Figure 1 A schematic diagram of a process for constructing a channel condition prediction model according to an embodiment of the present invention;

[0026] Figure 2 Schematic diagram of a channel condition prediction model fitting process according to an embodiment of the present invention;

[0027] Figure 3 A schematic diagram of a process flow for constructing a routing protocol according to an embodiment of the present invention;

[0028] Figure 4 Schematic diagram of simulation results of data packet delivery rates of different routing methods under different numbers of obstructions according to an embodiment of the present invention;

[0029] Figure 5 Schematic diagram of average communication delay simulation results of different routing methods under different numbers of obstructions according to an embodiment of the present invention. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below through specific embodiments in conjunction with the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0031] As mentioned in the background, existing mobile ad hoc network routing strategies are not applicable in scenarios with obstructions. Therefore, the main technical problem to be solved by this invention is how to design a suitable mobile ad hoc network routing method for environments where obstructions block wireless signals. In particular, this invention aims to design a data routing method suitable for mobile ad hoc networks of automated guided vehicles (AGVs) in the Industrial Internet, enabling AGVs to achieve network communication without the assistance of fixed wireless devices.

[0032] To achieve the above objectives, the present invention provides a data routing method for a mobile ad hoc network, the method being applied to a mobile node network, the mobile node network comprising a plurality of mobile nodes, each of which is a device capable of wireless communication, wherein each mobile node is configured with a pre-trained channel condition prediction model. The method comprises: S1: each node periodically broadcasts an EC message, wherein the EC message includes the current location of the node broadcasting the EC message, its location at a predicted future time, and an initial route evaluation score; S2: each node, upon receiving an EC message broadcast from another node, performs the following steps: S21: calculating its current location, received signal-to-noise ratio (SNR), and its location at a predicted future time based on the EC message received from the other node, and using the channel condition prediction model to predict the channel conditions between itself and the other node, and calculating the expected number of transmissions between itself and the other node based on the channel conditions; S22: calculating a route evaluation score between itself and the other node based on the expected number of transmissions between itself and the other node according to a preset method, and storing the route evaluation score; and S23: replacing the route evaluation score in the EC message received in step S21 with the route evaluation score in step S22, and then broadcasting the EC message. S3. Each node selects the node with the highest routing evaluation score from all single-hop neighbor nodes as the next hop node based on the routing evaluation scores between itself and other nodes.

[0033] As can be seen, the present invention first leverages node mobility and the periodicity of node transmission and reception of routing messages to generate training data online. It then proposes a fully connected neural network-based method to detect obstruction locations and predict channel conditions, generating a line-of-sight path map. Then, based on this, a latency-aware data routing protocol is constructed. This protocol estimates the expected number of transmissions on wireless links between nodes, and calculates and updates the routing evaluation score based on this number. Because this number is monotonically increasing with communication latency, the routing evaluation score reflects routing latency, and nodes will select paths with high routing evaluation scores (corresponding to low latency) as their next-hop forwarding nodes.

[0034] In order to understand the present invention more intuitively, the present invention will be described in detail from two aspects: the construction of a channel condition prediction model and the construction of a routing protocol.

[0035] 1. Construction of Channel Condition Prediction Model

[0036] The present invention can predict the channel conditions between node pairs by constructing a channel condition prediction model, that is, whether there are any obstructions between the node pairs, so as to construct a line-of-sight map. Figure 1As shown, the construction of the channel condition prediction model of the present invention mainly includes the following steps: collecting the original data set, obtaining the training data set based on probability analysis, expanding the training data set based on prior knowledge, evaluating the training data set and disseminating the optimal training data set, and obtaining the channel condition prediction model based on supervised training. Each step is described in detail below.

[0037] The first step is to collect the original data set. As mentioned above, each node will periodically generate and broadcast EC (Enhanced-Chirp) messages. Each node will continue to receive EC messages from other nodes. The node that receives the EC message can estimate the signal-to-noise ratio of the signal at the physical layer to obtain the received signal-to-noise ratio of the signal. By locating and parsing EC messages, you can get the location where you received the message and the sending node location Because each node will continue to receive EC messages broadcast by other nodes, all AGVs can obtain a set of EC messages locally. The original data set is composed of, that is, each node as a receiving node can get multiple The data composed of.

[0038] The second step is to obtain a training data set based on probability analysis. In the present invention, LOS is used to indicate the occurrence of the event "channel condition is line-of-sight path", and NLOS is used to indicate the occurrence of the event "channel is non-line-of-sight path", and it is known that LOS and NLOS are mutually exclusive events. When the distance is known, the value of the instantaneous signal-to-interference-and-noise ratio at any position obeys the probability distribution known by the analytical expression. The line-of-sight path signal-to-noise ratio threshold and the non-line-of-sight path signal-to-noise ratio threshold can be calculated through the channel fading in the mobile ad hoc network scenario, and then the channel conditions between the sending node and the receiving node are calculated according to the receiving signal-to-noise ratio of the receiving node, wherein when the receiving signal-to-noise ratio is greater than or equal to the line-of-sight path signal-to-noise ratio threshold, the channel condition is line-of-sight path; when the receiving signal-to-noise ratio is less than or equal to the non-line-of-sight path signal-to-noise ratio threshold, the channel condition is non-line-of-sight path. The position of the sending node and the position of the receiving node in each piece of original data are used as samples, and the channel conditions between the two nodes are used as labels to construct the initial training set corresponding to each node. ,in: LOS or NLOS, indicating The channel conditions between.

[0039] The third step is to expand the training data set based on prior knowledge. In this step, the initial training set corresponding to each node is expanded in the same way. Assume that the sending node position of each data in each initial training set is To the receiving node location The coordinates of the intersection of the ray direction and the site boundary are ,when and When the channel condition between is NLOS, it indicates and There is occlusion between and The channel condition between is also NLOS, so the new data Add to the training dataset.

[0040] The fourth step is to evaluate the training data set. The evaluation here refers to evaluating the expanded initial training set corresponding to each node. Specifically, the scene is divided into grids of equal area, for data volume Each training data in the initial training set is sampled and Online points, merged to obtain Sampling points, count the number of grids with sampling points , calculate the spatial coverage of the training data set, the expression is For a mobile ad hoc network composed of multiple nodes, different nodes have different movement trajectories, and the resulting training datasets are also different. Because AGVs within a given area can only exchange data via the mobile ad hoc network, to reduce communication overhead, this method uses the node with the highest spatial coverage corresponding to the training dataset to locally train a channel condition prediction model, which is then distributed to other nodes.

[0041] The fifth step is to obtain the channel condition prediction model based on supervised training. Figure 2 As shown, at a given sending node location and receiving node location , uniform sampling on the line connecting the two points points, we can get a set of points , . Set the location When there is an occluder, the occluder position function , when the position When there is no occluder, the occluder position function , nonlinear function Since the fully connected neural network structure has excellent performance in nonlinear function fitting, the present invention designs a fully connected neural network with 3 hidden layers for fitting Supervised training is performed on the expanded training dataset. A binary cross-entropy loss function, which is sensitive to classification errors, is used as the loss function. A momentum gradient descent algorithm is used for optimization. The hidden layer weight matrix and bias terms of the fully connected neural network are adjusted using a backpropagation algorithm. After multiple rounds of training, a channel condition prediction model is obtained through fitting. Since the fully connected neural network is a model structure well known to those skilled in the art, this embodiment of the present invention does not provide a detailed description of the specific results of the fully connected neural network.

[0042] The channel addition prediction model obtained through supervised training is distributed to other nodes. When each node receives the EC message, it can predict the channel conditions based on the location of the sending node, its own location and the received signal-to-noise ratio.

[0043] 2. Routing Protocol Construction

[0044] In this invention, the routing protocol stipulates that each node periodically broadcasts EC messages to help other nodes perceive the routing path latency between them. Nodes that receive EC messages first estimate the expected number of transmissions on the inter-node wireless link over a period of time and then calculate a routing evaluation score based on this number. When forwarding application layer data, nodes compare the routing evaluation scores and select the node on the routing path with the highest score (and lowest latency) as the next hop.

[0045] like Figure 3 As shown, the routing method of the present invention includes the following steps: a node receiving an EC message calculates the expected number of transmissions of a wireless link between nodes, calculates a routing evaluation score, and selects a forwarding node. Each step is described in detail below.

[0046] The first step is to calculate the expected number of transmissions of wireless links between nodes. Time Node Generate and broadcast EC message, EC message contains node Current location and nodes future Time position ,in, Indicates the prediction time domain, which is generally the EC message broadcast period. Node j receives the Generate and broadcast EC messages and extract from EC messages and After that, node j predicts Time Node The position of is expressed as follows

[0047]

[0048] Simultaneous nodes Obtained through positioning and local path planning and , node j prediction The expression for your position at any moment is as follows:

[0049]

[0050] in, represents the sampling period, , Known and ,node Can be calculated The probability of wireless link interruption at time t is expressed as follows:

[0051] in, It is the channel condition predicted by the node through the pre-configured channel condition prediction model.

[0052] For all time, , according to the above formula, we can calculate the set , at this time node With node The expected number of transmissions on the wireless link between is:

[0053]

[0054] The above formula gives the future The expected number of transmissions required to transmit data between two nodes within a certain time period. If the expected number of transmissions is greater than the maximum number of transmissions, In the case of , the expected number of transmissions is positive infinity.

[0055] The second step is to calculate the routing evaluation score. Time Node Generate and broadcast EC message, which contains the route evaluation score with an initial value of 1 .node Received by the node After generating and broadcasting the EC message, extract the ,node Calculated After that, the node The new routing evaluation score is calculated according to the following formula :

[0056]

[0057] in is a constant and , the greater the expected number of transmissions, the smaller the value of the exponential function, and the smaller the value of the new routing evaluation score. Calculated Later General Save locally, and then replace the original EC message Replace with , change the original EC message and Replace with nodes respectively Current position and future The location after the time, and finally forward the EC message. Received Node Forwarded origin node After receiving the EC message, the node The new route evaluation score is calculated according to the following formula:

[0058]

[0059] Similarly, nodes Will Save locally, and then replace the original EC message Replace with , change the EC message and Replace with nodes respectively Current position and future The EC message is forwarded after the time. The subsequent routing evaluation score calculation and EC message forwarding process are the same as above. After a series of broadcasts, each node stores multiple routing evaluation scores with other nodes.

[0060] The third step is to select the forwarding node. Let the source node of the data flow in the network be node 1, and the destination node of the data flow be node N. Indicates the past Successfully sent to the node within the time The set of nodes that have sent EC messages is also called node After node N broadcasts the EC message, after the above EC message broadcast and forwarding process, node 1 stores the routing evaluation score set , where the nodes is the single-hop neighbor node of node 1. The communication delay and the expected number of transmissions are monotonically increasing, so node 1 will forward the application layer data sent to node N to the set The node with the largest routing evaluation score (minimum delay) Similarly, the node Stores a set of routing evaluation scores , where the nodes is a node The single-hop neighbor node of node Select Collection The node corresponding to the maximum routing evaluation score As the next hop forwarding node. And so on, after being forwarded by several nodes, node N receives the data from node 1.

[0061] In order to verify the performance of the mobile ad hoc network data routing method (LOS MAP based Delay Aware Routing, LM-DAR) of the present invention, the inventors simulated the mobile guided vehicle ad hoc network routing scenario in the industrial Internet by implementing a simulation program, and counted the data packet delivery rate and average communication delay of the access protocol. At the beginning of the simulation, all vehicle nodes are randomly distributed in the scene, and then all vehicle nodes will randomly select the end point in the blank area of ​​the scene. A path planning algorithm such as the A* path planning algorithm is used to obtain a path to the end point. The vehicle node will drive along the planned path and reach the end point. After reaching the end point, each vehicle node will randomly select the end point in the scene again and reach the end point along the planned path until the end of the simulation. The size of the site is set to 400 meters × 400 meters, and the size of the obstruction is 10 meters × 10 meters. The position of the obstruction is randomly generated. In order to simulate the situation where devices are clustered together in the actual scene, the distribution of the obstruction follows a cluster-based Poisson point process. The large-scale channel model adopts 3GPP Inf-SL. The broadcast period of the EC message Set to 1 second to predict the time domain length At the same time, the neighbor node timeout is set to 3 seconds. The maximum number of node transmissions is set to 4, and the timeout is set to 50 us. The simulation results are as follows Figure 4 and 5 As shown, Figure 4 The vertical axis is the data packet delivery rate result of the routing method, and the broken lines of different colors and types correspond to the results of different routing methods, among which the red solid line is the result of the routing method proposed in the present invention. Figure 5 The horizontal axis represents the number of obstructions, and the vertical axis represents the average communication latency of the routing methods. Different colors and types correspond to the results of different routing methods, with the solid red line representing the results of the routing method proposed in this paper. Simulation results demonstrate that the proposed routing method is suitable for automated guided vehicle mobile ad hoc networks in the presence of obstructions, and that the proposed routing method outperforms the comparative methods in both packet delivery rate and communication latency.

[0062] Compared with the existing technology, in an industrial production environment with obstructions, the solution proposed in the present invention constructs a line-of-sight map based on a fully connected neural network. The line-of-sight map is obtained by constructing a training data set through sending and receiving routing messages. When a node needs to send data to another node, the node applies a data routing protocol based on the line-of-sight map to implement multi-hop forwarding and complete vehicle-to-vehicle communication. It is well suited for self-organizing network routing construction in an environment with obstructions.

[0063] It should be noted that although the above describes the various steps in a specific order, it does not mean that the steps must be performed in the above specific order. In fact, some of these steps can be executed concurrently or even in a different order as long as the required functions can be achieved.

[0064] The present invention may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present invention.

[0065] A computer-readable storage medium may be a tangible device that holds and stores instructions used by an instruction execution device. Computer-readable storage media may include, for example, but is not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disc (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove having instructions stored thereon, and any suitable combination thereof.

[0066] While various embodiments of the present invention have been described above, the above descriptions are intended to be illustrative, non-exhaustive, and not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is selected to best explain the principles of the embodiments, their practical applications, or technological improvements in the marketplace, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A data routing method for a mobile ad hoc network, used in a mobile node network, wherein the mobile node network includes a plurality of mobile nodes, each of which is a device with wireless communication function, wherein: Each mobile node is configured with a pre-trained channel condition prediction model, wherein the method comprises: S1. Each node periodically broadcasts an EC message, which contains the current location of the node broadcasting the EC message, the location at the predicted future time, and the initial routing evaluation score; S2. Each node performs the following steps when it receives an EC message broadcast from another node: S21. Calculate its current position, received signal-to-noise ratio, and future predicted position based on the EC messages broadcast from other nodes, and use a channel condition prediction model to predict the channel conditions between itself and other nodes and calculate the expected number of transmissions between itself and other nodes based on the channel conditions. S22, calculating the routing evaluation score between itself and other nodes in a preset manner based on the expected number of transmissions between itself and other nodes and saving the routing evaluation score; S23, replacing the route evaluation score in the EC message received in step S21 with the route evaluation score in step S22 and then broadcasting the EC message; S3. Each node selects the node with the highest routing evaluation score from all single-hop neighbor nodes as the next hop node based on the routing evaluation scores between itself and other nodes.

2. The mobile ad hoc network data routing method according to claim 1, characterized in that: The channel condition prediction model is pre-trained in the following manner: T1. Collect multiple raw data sets with each node as a sending node. Each raw data set corresponds to a node. The raw data set corresponding to each node contains multiple raw data. Each raw data set includes the position of another node as the sending node, the position of the node as the receiving node, and the received signal-to-noise ratio. T2. For each node's original data set, calculate whether the channel condition between the two nodes in each piece of original data is line-of-sight or non-line-of-sight according to a preset method. Use the sending node's position and the receiving node's position in each piece of original data as samples and the channel condition between the two nodes as labels to construct an initial training set corresponding to each node. T3. Expand the initial training set corresponding to each node, wherein the receiving node position in the sample labeled as non-line-of-sight is expanded to the site boundary to obtain a new sample, and the new sample is labeled as non-line-of-sight; T4. Evaluate the spatial coverage of the initial training set after each node expansion in the site, and select the training set with the highest spatial coverage as the final model training set; T5. Based on the model training set selected in step T4, at the node where the training set is located, with the position of the sending node and the position of the receiving node as input and the channel condition between the two nodes as output, train the fully connected neural network until convergence to obtain a channel condition prediction model, and distribute the model to other nodes.

3. The mobile ad hoc network data routing method according to claim 2, characterized in that: The preset method is: Calculate the line-of-sight signal-to-noise ratio threshold and non-line-of-sight signal-to-noise ratio threshold based on channel fading in mobile ad hoc network scenarios; The channel condition between the sending node and the receiving node is calculated based on the received signal-to-noise ratio (SNR) of the receiving node. When the SNR is greater than or equal to the line-of-sight SNR threshold, the channel condition is line-of-sight. When the SNR is less than or equal to the non-line-of-sight SNR threshold, the channel condition is non-line-of-sight.

4. The mobile ad hoc network data routing method according to claim 2, characterized in that: In step T4, the spatial coverage of the expanded training set of each node in the site is calculated as follows: T41. Divide the mobile ad hoc network scene into multiple grids of equal area; T42. For each node in the expanded initial training set, sample multiple points on the line connecting two nodes in each data to obtain a set of sampling points corresponding to each node; T43. Count the number of grids covered by the sampling point set corresponding to each node. If a grid contains any number of sampling points in the sampling point set, the grid is considered to be covered by the sampling point set. T44. Calculate the spatial coverage of the initial training set after the node is expanded based on the ratio of the number of grids covered by the sampling point set corresponding to each node to the total number of grids.

5. The mobile ad hoc network data routing method according to claim 4, characterized in that: The number of hidden layers of the fully connected neural network is 3, a binary cross entropy loss function is adopted, and the momentum gradient descent algorithm is used to optimize the parameters of the fully connected neural network through back propagation.

6. The mobile ad hoc network data routing method according to claim 1, characterized in that: In step S21, each node calculates the expected number of transmissions between itself and other nodes in the following manner: in, Representation node The current position at time t, Representation node future Time position, represents the prediction time domain, Indicates the EC message broadcast period, is the preset maximum transmission number threshold, The wireless link between node j and node i is The probability of interruption at time , represents the sampling period, is the prediction of node j The position of node i at time instant, is the prediction of node j The position of node j at time instant, where: in, represents the channel condition between node i and node j predicted by the channel condition prediction model.

7. The mobile ad hoc network data routing method according to claim 6, characterized in that: In step S22, each node calculates the route evaluation score in the following manner: in, represents the routing evaluation score between node j and node i, represents the initial route evaluation score in the EC message sent by node i, is a constant and .

8. An industrial ad hoc network system, comprising a plurality of mobile guide vehicles, each guide vehicle being a wireless node, characterized in that: The industrial ad hoc network system is configured to perform data routing using the method according to any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that A computer program is stored thereon, and the computer program can be executed by a processor to implement the steps of any one of the methods of claims 1-7.

10. An electronic device, characterized in that: include: One or more processors, and memory, wherein the memory is configured to store executable instructions; The one or more processors are configured to implement the steps of the method according to any one of claims 1 to 7 by executing the executable instructions.

Images