Off-network gateway dynamic election method and system based on edge AI and satellite ephemeris

Abstract

The application discloses an off-grid gateway dynamic election method and system based on edge AI and satellite ephemeris, relates to the technical field of off-grid communication networks, and comprises the following steps: in an off-grid ad hoc network, a candidate gateway node set is formed, multi-source running state information of each candidate gateway node in the candidate gateway node set is collected, and a node state feature vector is constructed according to a field arrangement order; based on the node state feature vector, intelligent reasoning is performed on the edge side to generate a gateway evaluation index, the gateway evaluation index is subjected to a gateway availability constraint in combination with satellite ephemeris information to form a constrained gateway evaluation index; dynamic off-grid gateway election is performed according to the constrained gateway evaluation index, an off-grid gateway node is determined, gateway role announcement information is broadcast by the off-grid gateway node, and non-gateway nodes are controlled to collect data to be sent to the off-grid gateway node through a local multi-hop link. The application realizes gateway collaborative optimization in an off-grid communication scenario, reduces energy consumption and improves backhaul reliability.

Description

Technical Field

[0001] This invention relates to the field of off-grid communication network technology, and in particular to a method and system for dynamic election of off-grid gateways based on edge AI and satellite ephemeris. Background Technology

[0002] In applications such as forestry, agriculture, marine monitoring, mineral exploration, wildlife protection, and disaster emergency response, a large number of smart terminals are deployed in off-grid environments lacking cellular communication infrastructure coverage. These terminals typically rely on low-power wireless communication technologies to build self-organizing networks, enabling data collection and local collaborative transmission between nodes. To achieve cross-regional data backhaul, existing systems generally introduce nodes with satellite communication capabilities to act as gateways, transmitting aggregated data to remote servers via satellite communication links, thus forming an "end-to-end-to-cloud" communication system to support information acquisition and remote management needs in off-grid scenarios.

[0003] However, existing methods still have two limitations: First, the selection of gateway nodes often relies on static rules or finite state indicators, making it difficult to comprehensively characterize the changing characteristics of node communication conditions, energy status, and load in a dynamic environment; Second, satellite communication has significant time-varying and intermittent characteristics, and traditional methods usually passively respond to changes in link availability after the gateway is selected, lacking the forward-looking utilization of satellite orbits and communication time windows, resulting in limited coordination between gateway decisions and actual satellite communication conditions. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a dynamic election method for off-grid gateways based on edge AI and satellite ephemeris, which solves the problem of insufficient coordination between gateway election decision-making, node operating status, and satellite communication availability in off-grid environments.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a dynamic election method for off-grid gateways based on edge AI and satellite ephemeris, which includes forming a set of candidate gateway nodes in an off-grid self-organizing network, collecting multi-source operating status information for each candidate gateway node in the set of candidate gateway nodes, and constructing a node status feature vector according to the order of field arrangement.

[0008] Based on node state feature vectors, intelligent reasoning is performed at the edge to generate gateway evaluation metrics, and gateway availability constraints are applied to the gateway evaluation metrics in combination with satellite ephemeris information to form constrained gateway evaluation metrics.

[0009] Dynamic off-network gateway election is performed based on constrained gateway evaluation metrics to determine off-network gateway nodes. The off-network gateway nodes then broadcast gateway role announcement information to control non-gateway nodes to aggregate data to be sent to the off-network gateway nodes via local multi-hop links.

[0010] After selecting the off-network gateway node, the satellite communication time window of the off-network gateway node is calculated based on the satellite ephemeris information. Within the satellite communication time window, the off-network gateway node is triggered to establish a satellite communication link and report data. Outside the satellite communication time window, the off-network gateway node is controlled to enter hibernation and wake up according to the preheating time, thus completing the data reporting timing decision and hibernation wake-up control.

[0011] As a preferred embodiment of the off-grid gateway dynamic election method based on edge AI and satellite ephemeris described in this invention, the formation of the candidate gateway node set includes, during the operation of the off-grid ad hoc network, performing gateway capability screening based on the node communication capabilities, online status, and power status in the off-grid ad hoc network to form the candidate gateway node set.

[0012] As a preferred embodiment of the off-grid gateway dynamic election method based on edge AI and satellite ephemeris described in this invention, the multi-source operational status information collected includes satellite communication observation information, off-grid self-organizing network status information, energy status information, and operational load information. Among these, the satellite communication observation information includes the satellite signal-to-noise ratio and the number of visible satellites; the off-grid self-organizing network status information includes the hierarchical position of the candidate gateway node in the off-grid self-organizing network and the channel occupancy status; the energy status information includes the remaining power and energy acquisition rate; and the operational load information includes the length of the data queue to be sent.

[0013] After completing the collection of multi-source operational status information, the various operational status information are numerically processed and arranged in the order of fields such as satellite signal-to-noise ratio, number of visible satellites, hierarchical location, channel occupancy, remaining power, energy acquisition rate, and length of the data queue to be transmitted, forming the node status feature vector of the candidate gateway node.

[0014] As a preferred embodiment of the off-network gateway dynamic election method based on edge AI and satellite ephemeris described in this invention, wherein: the step of generating gateway evaluation indicators by performing intelligent inference on the edge side includes, during the election period, inputting the node state feature vector corresponding to each candidate gateway node in the candidate gateway node set into a lightweight intelligent model deployed locally on the node, and performing forward inference operations;

[0015] The lightweight intelligent model is constructed using a predictive decision-making model based on reinforcement learning, and consists of a state space, a discrete action space, and a reward function.

[0016] The lightweight intelligent model constructs a state space based on node state feature vectors and calculates action value estimates for various operating modes in the discrete action space by combining a reward function.

[0017] The action value estimate of the active gateway mode is extracted from the action value estimate and used as the gateway evaluation index for candidate gateway nodes in the current election cycle.

[0018] As a preferred embodiment of the off-network gateway dynamic election method based on edge AI and satellite ephemeris described in this invention, the step of applying gateway availability constraints to gateway evaluation indicators by combining satellite ephemeris information to form constrained gateway evaluation indicators includes, after generating gateway evaluation indicators, candidate gateway nodes combine the node's local geographical location and the node's local real-time clock, calculate the satellite's spatial position vector relative to the candidate gateway node during the election period based on satellite ephemeris information, and calculate the satellite's elevation angle at any moment during the election period from the satellite's spatial position vector;

[0019] Based on the satellite elevation angle and the minimum elevation angle requirement for satellite communication, calculate the remaining visible window time during the election cycle to ensure communication.

[0020] When the remaining visible window time is less than the duration required to complete a data collection, satellite communication link establishment and data reporting process, or when the remaining power of the candidate gateway node in the latest sample is lower than the safe minimum power required for the gateway role to operate, the gateway evaluation index will be marked as unavailable.

[0021] If the remaining visible window time is not less than the duration required to complete one data collection, satellite communication link establishment and data reporting process, and the remaining power of the candidate gateway node in the latest sampling is not less than the safe lower limit power required for the gateway role to operate, the original gateway evaluation index value is retained, and an available status flag is added to the gateway evaluation index to obtain the constrained gateway evaluation index.

[0022] As a preferred embodiment of the off-network gateway dynamic election method based on edge AI and satellite ephemeris described in this invention, the step of performing dynamic off-network gateway election according to constrained gateway evaluation indicators and determining off-network gateway nodes includes: selecting the candidate gateway node with the highest value of the constrained gateway evaluation indicators as the off-network gateway node by comparing the values ​​of the constrained gateway evaluation indicators in the available state.

[0023] When multiple candidate gateway nodes have the same constrained gateway evaluation metric value, the candidate gateway node whose hierarchical position is closer to the off-network self-organizing network center is selected as the off-network gateway node.

[0024] As a preferred embodiment of the off-network gateway dynamic election method based on edge AI and satellite ephemeris described in this invention, wherein: controlling non-gateway nodes to collect data to be sent to the off-network gateway node via local multi-hop links includes: after receiving gateway role notification information, the non-gateway nodes in the off-network self-organizing network update their node role status, do not perform satellite communication link establishment and data reporting operations during the current election period, and forward the data to be sent generated or cached by their nodes to the off-network gateway node via local multi-hop communication links in the off-network self-organizing network;

[0025] Off-network gateway nodes receive data to be sent forwarded from non-gateway nodes and perform aggregation processing on the received data to be sent, forming data to be reported.

[0026] As a preferred embodiment of the off-network gateway dynamic election method based on edge AI and satellite ephemeris described in this invention, the satellite communication time window of the off-network gateway node calculated based on satellite ephemeris information includes taking the moment when the satellite elevation angle first meets the minimum elevation angle requirement for allowing satellite communication as the start time of the satellite communication time window, and taking the moment when the satellite elevation angle no longer meets the minimum elevation angle requirement for allowing satellite communication as the end time of the satellite communication time window, and determining the satellite communication time window by the start time and the end time.

[0027] As a preferred embodiment of the off-grid gateway dynamic election method based on edge AI and satellite ephemeris described in this invention, the sleep-wake control includes: the off-grid gateway node predicting the start time of the next satellite communication time window based on satellite ephemeris information, and calculating the warm-up time in combination with the startup time required for the off-grid gateway node to recover from a low-power operation state to have satellite communication capabilities;

[0028] Before the warm-up period arrives, the off-grid gateway node continues to operate in a low-power state. After the warm-up period arrives, the off-grid gateway node exits the low-power state and enters the communication preparation state.

[0029] Secondly, the present invention provides an off-grid gateway dynamic election system based on edge AI and satellite ephemeris, including a candidate screening module, used to screen nodes with gateway capabilities in an off-grid self-organizing network to form a candidate gateway node set, and to construct a node state feature vector of the candidate gateway nodes.

[0030] The intelligent evaluation module is used to perform lightweight intelligent inference on the edge side based on node state feature vectors, generate gateway evaluation metrics, and apply gateway availability constraints.

[0031] The gateway election module is used to determine the off-network gateway node based on constrained gateway evaluation metrics, and to control non-gateway nodes to gather data to be sent to the off-network gateway node.

[0032] The timing control module is used to calculate the satellite communication time window based on satellite ephemeris information and to execute data reporting timing decisions and sleep / wake-up control.

[0033] The beneficial effects of this invention are as follows: By using a lightweight intelligent model on the edge side to infer and evaluate the multi-source operating status of nodes, a dynamic characterization of the comprehensive capabilities of the gateway role in an off-grid ad hoc network is achieved, enabling the gateway election results to adaptively adjust according to changes in communication conditions, energy status, and load; by using satellite ephemeris information to predict communication availability time and implementing time window and sleep-wake control, precise matching between gateway node communication behavior and satellite availability is achieved, ensuring that data reporting is executed in an orderly manner within the effective time period and reducing energy consumption during non-communication periods, thereby improving the overall efficiency of satellite communication utilization and the reliability of node operation in the off-grid environment. Attached Figure Description

[0034] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly explained below.

[0035] As is obvious, the accompanying drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without any creative effort.

[0036] Figure 1 This is a flowchart of an off-grid gateway dynamic election method based on edge AI and satellite ephemeris.

[0037] Figure 2 This is a schematic diagram of an off-grid gateway dynamic election system based on edge AI and satellite ephemeris.

[0038] Figure 3 A flowchart for generating and applying constraints to gateway evaluation metrics.

[0039] Figure 4 This is a flowchart for satellite communication time window calculation and hibernation / wake-up control. Detailed Implementation

[0040] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0041] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0042] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0043] Reference Figures 1-4 This is one embodiment of the present invention, which provides a dynamic election method for off-network gateways based on edge AI and satellite ephemeris, including the following steps:

[0044] S1. In the off-network self-organizing network, a set of candidate gateway nodes is formed, and multi-source operating status information is collected for each candidate gateway node in the set. The node status feature vector is constructed according to the field arrangement order.

[0045] Furthermore, during the operation of the off-grid ad hoc network, gateway capability screening is performed based on the node communication capabilities, online status, and power status in the off-grid ad hoc network to form a set of candidate gateway nodes.

[0046] During the initialization phase or the start of the election cycle of the off-grid ad hoc network, communication capability checks are performed on the nodes in the off-grid ad hoc network. Only nodes with satellite communication interfaces and the ability to assume the gateway role are included in the candidate gateway node set. During each election cycle, the online status of the nodes in the off-grid ad hoc network is checked. If none of the nodes in the off-grid ad hoc network participate in local communication or respond to neighboring nodes' broadcast information during the election cycle, the checked nodes will not be included in the candidate gateway node set for the current election cycle. During each election cycle, the remaining power of the nodes in the off-grid ad hoc network is assessed. If the remaining power of a node in the off-grid ad hoc network is lower than the safety threshold required for the gateway role to operate, the checked nodes will not be included in the candidate gateway node set for the current election cycle. The remaining power assessment is based on the remaining power reported by the nodes at the start of the election cycle, which is used to narrow down the size of the candidate gateway node set.

[0047] By comprehensively screening communication capabilities, online status, and power status, a set of candidate gateway nodes that meet the conditions for participating in the gateway election in the current election cycle is obtained.

[0048] It should be noted that an off-grid ad hoc network refers to a network structure that does not rely on cellular communication infrastructure and consists of multiple nodes forming multi-hop connections through local wireless communication.

[0049] The election cycle refers to the time interval used in an off-grid ad hoc network to perform node status judgment and gateway election. During each election cycle, a comprehensive judgment is made on the communication capability, online status, and power status of the nodes in the off-grid ad hoc network to determine whether the nodes meet the conditions for participating in the gateway election. The election cycle can be triggered by a time interval or by events such as changes in the gateway node status, communication failure, or completion of data reporting, in order to ensure that the gateway role in the off-grid ad hoc network can be dynamically adjusted according to changes in the node's operating status.

[0050] The gateway role refers to the node role in an off-grid ad hoc network that is responsible for data aggregation and satellite communication. Nodes that assume the gateway role are used to receive data forwarded by other nodes in the off-grid ad hoc network via local multi-hop communication links, and establish satellite communication links to send the data to be reported to the remote receiving end via satellite communication links to complete data reporting. Nodes that do not assume the gateway role only participate in local communication and data forwarding in the off-grid ad hoc network, and do not perform satellite communication link establishment and data reporting operations.

[0051] The minimum power requirement for gateway operation refers to the minimum remaining power required to ensure that nodes in an off-grid ad hoc network can complete a data aggregation, satellite communication link establishment, and data transmission process. This minimum remaining power requirement consists of power consumption during the satellite communication preparation phase, satellite communication link establishment phase, data transmission phase, and local processing phase. Candidate gateway nodes record the average power of each phase locally and time the phase duration. The power consumption for each phase is obtained by multiplying the average power by the phase duration, and then the power consumption of each phase is summed to obtain the total power consumption for a complete process. Candidate gateway nodes in the process... Record the remaining power before starting and record the remaining power again after the process ends. Verify the consistency of the total power consumption calculation results by comparing the difference between the two remaining power values. After the candidate gateway node has completed multiple satellite communication link establishments and data transmissions, select the average of the total power consumption that has passed the consistency verification of several times (e.g., 10 times). Add a safety redundancy (e.g., 10%) to the average of the total power consumption of the candidate gateway node to form the minimum remaining power requirement, which serves as the safety lower limit power required for the gateway role to operate. The remaining power and the safety lower limit power are both expressed in terms of the available energy metering of the node power supply and are kept in the same metering caliber as the stage power consumption.

[0052] Furthermore, using the local real-time clock of the nodes in the off-grid ad hoc network as a unified time reference, multi-source operational status information is collected from each candidate gateway node in the candidate gateway node set during the election cycle. The multi-source operational status information includes satellite communication observation information, off-grid ad hoc network status information, energy status information, and operational load information.

[0053] Among them, satellite communication observation information includes at least the satellite signal-to-noise ratio and the number of visible satellites; off-grid ad hoc network status information includes at least the hierarchical position of the candidate gateway node in the off-grid ad hoc network and the channel occupancy status; energy status information includes at least the remaining power and energy acquisition rate; and operating load information includes at least the length of the data queue to be sent.

[0054] Furthermore, after collecting multi-source operational status information, the multi-source operational status information is subjected to numerical processing and field arrangement processing.

[0055] During numerical processing, normalization is performed on satellite signal-to-noise ratio, number of visible satellites, hierarchical position, channel occupancy, remaining power, energy acquisition rate, and length of the data queue to be transmitted, so that the values ​​of each field fall within a unified numerical range of zero to one. Normalization adopts a linear mapping based on the minimum and maximum values ​​of the fields. The minimum and maximum values ​​of the fields are given by the node's local parameter table or obtained from historical data collection and updated during operation. When a candidate gateway node fails to collect a certain multi-source operating status information during the election period, a placeholder identifier is written in the corresponding field position. The placeholder identifier is negative one, which is used to indicate the uncollected state and to distinguish valid values ​​within the range of zero to one.

[0056] After numerical processing is completed, the multi-source operating status information is processed by field arrangement. The fields are arranged in the following order: satellite signal-to-noise ratio, number of visible satellites, hierarchical position, channel occupancy, remaining power, energy acquisition rate, and length of the data queue to be transmitted. The node status feature vector is formed according to the field arrangement order.

[0057] S2. Based on the node state feature vector, perform intelligent inference on the edge side to generate gateway evaluation indicators, and combine satellite ephemeris information to apply gateway availability constraints to the gateway evaluation indicators, forming constrained gateway evaluation indicators.

[0058] Furthermore, after obtaining the node state feature vector of each candidate gateway node in the candidate gateway node set, edge-side intelligent inference processing is performed on the candidate gateway node side.

[0059] Edge-side intelligent inference processing is based on a lightweight intelligent model deployed locally on the candidate gateway node. The lightweight intelligent model runs on the local computing unit of the candidate gateway node. After offline deployment, the candidate gateway node directly calls and executes inference calculations during the election cycle. The lightweight intelligent model takes the node state feature vector as input and performs forward inference operations through the parameters of the lightweight intelligent model to generate gateway evaluation indicators that characterize the candidate gateway node's ability to perform gateway role behavior in the current running state.

[0060] Furthermore, in this embodiment, the lightweight intelligent model is constructed using a predictive decision-making model based on reinforcement learning. The gateway election process is modeled as a partially observable Markov decision process. The lightweight intelligent model consists of a state space, a discrete action space, and a reward function.

[0061] The state space is constructed using node state feature vectors as input. Fields from these vectors are input into the lightweight intelligent model in the order they are arranged, characterizing the satellite communication state, off-grid ad hoc network state, and energy state of candidate gateway nodes during the election cycle. The state features in the state space consist of multiple sub-features, including communication-related features, equipment operation features, and network coordination features. Communication-related features, composed of satellite signal-to-noise ratio and the number of visible satellites, represent the satellite communication observation state of candidate gateway nodes. Equipment operation features, composed of remaining battery power and energy acquisition rate, represent the energy state of candidate gateway nodes. Network coordination features, composed of hierarchical position, data queue length, and channel occupancy, represent the cooperative position of candidate gateway nodes in the off-grid ad hoc network. All state features are converted into numerical representations within a unified numerical range before being input into the lightweight intelligent model.

[0062] The discrete action space is used to describe the operating modes that candidate gateway nodes can execute during the election cycle. The operating modes include at least deep sleep mode, local forwarding mode, gateway standby mode, and active gateway mode.

[0063] Among them, the deep sleep mode disables radio frequency communication capabilities, only maintaining the local timing and status maintenance functions of the node; the local forwarding mode enables the node's local wireless communication capabilities to perform data reception and multi-hop forwarding within the off-network self-organizing network; the gateway standby mode enables the node's satellite reception capabilities to perform satellite search and tracking operations, but does not perform satellite data transmission; the active gateway mode enables the node's satellite transmission capabilities to establish satellite communication links after data aggregation is completed, and to complete the satellite communication link transmission of data to be reported, so as to realize data reporting.

[0064] The reward function is constructed in a composite form to characterize the rewards generated by different operating modes during the election period. The reward function comprehensively considers task completion rewards, energy consumption, risk penalties, and rotation adjustments. Task completion rewards are positive (e.g., 1) when the active gateway mode is executed during the election period and successfully completes one data aggregation, satellite communication link establishment, and satellite communication link transmission of data to be reported; negative (e.g., -1) when the active gateway mode is executed during the election period but any of the following occurs: satellite communication link establishment failure, satellite communication link interruption, or failure to transmit data to be reported; and zero when the active gateway mode is not executed during the election period. Energy consumption is obtained by the candidate gateway node recording its operating power and timing the election period duration locally. Operating power is the sum of transmit power, receive power, and processing power; energy consumption is the product of operating power and the election period duration, normalized using the minimum safe power requirement for gateway operation as the normalization benchmark, ensuring the energy consumption value is between zero and one. Risk penalties are calculated when the candidate gateway node is in gateway standby mode. In the active gateway mode, the difference between the current remaining power of the candidate gateway node and the total power consumption of a complete process is calculated to obtain the remaining power after completing a complete process. When the remaining power after completing a complete process is lower than the safety lower limit power required for the gateway role to operate, it is taken as a non-zero value. The risk penalty is the difference between the safety lower limit power required for the gateway role to operate and the remaining power after completing a complete process, and is normalized using the safety lower limit power required for the gateway role to operate as the normalization benchmark, so that the risk penalty value is between zero and one. The risk penalty is zero when the remaining power after completing a complete process is not lower than the safety lower limit power required for the gateway role to operate, and the risk penalty is zero when the operating mode is deep sleep mode or local forwarding mode. The rotation adjustment is determined by the number of consecutive times the candidate gateway node has served as the gateway role. The number of consecutive times the candidate gateway node has served as the gateway role is the number of election cycles in which the candidate gateway node has been continuously selected as the off-network gateway node. The rotation adjustment is one minus the normalized value of the number of consecutive times the candidate gateway node has served as the gateway role and the lower limit is truncated, so that the rotation adjustment is monotonically reduced as the number of consecutive times the candidate gateway node has served as the gateway role increases, and the value of the rotation adjustment is between zero and one.

[0065] The reward function is expressed as:

[0066] ;

[0067] in, For the reward function, As a reward for completing the task, Energy consumption As a risk penalty amount, This is the rotation adjustment amount.

[0068] The lightweight intelligent model is constructed using a lightweight fully connected neural network structure. The neural network consists of an input layer, several hidden layers, and an output layer. The dimension of the input layer is consistent with the dimension of the node state feature vector, and the dimension of the output layer is consistent with the number of operating modes in the discrete action space. The output value is used to represent the value estimate of different operating modes in the current state. The training phase of the lightweight intelligent model is completed on the server side. The training data is generated from satellite channel simulation data and historical operating data. After training, the neural network parameters are quantized and deployed to the local storage space of the candidate gateway node.

[0069] During the operation of the candidate gateway node, the lightweight intelligent model supports an online fine-tuning mechanism. It maintains an experience data storage area locally on the candidate gateway node to record the status, operation mode and reward data in the recent election cycle. When the update conditions are met, it performs a small update on the neural network parameters to adapt to changes in satellite occlusion, energy acquisition and network load in the specific deployment environment.

[0070] After the lightweight intelligent model completes training and is deployed to the candidate gateway nodes, during the election period, the candidate gateway nodes input the node state feature vector into the lightweight intelligent model to perform forward inference calculation. During the forward inference calculation, the lightweight intelligent model calculates the action value estimate corresponding to each running mode in the discrete action space based on the current node state feature vector.

[0071] The action value estimate output by the lightweight intelligent model is used to reflect the expected return level of candidate gateway nodes in the current state when executing different operating modes. During the inference phase, the action value estimate corresponding to the active gateway mode is extracted from the action value estimate, and the selected action value estimate is used as the gateway evaluation index of the candidate gateway node in the current election cycle.

[0072] Among them, the gateway evaluation index is obtained by reasoning and calculation of the node state feature vector through a lightweight intelligent model. The gateway evaluation index represents the comprehensive ability level of the candidate gateway node to assume the gateway role and complete data collection and satellite communication behavior under the current state feature vector conditions in a single numerical form.

[0073] Furthermore, after completing edge-side intelligent inference processing and generating gateway evaluation metrics, gateway availability constraint processing is performed on the gateway evaluation metrics using satellite ephemeris information.

[0074] Satellite ephemeris information records the state data of satellite orbital parameters changing over time. Candidate gateway nodes, combining their local geographical location and local real-time clock, calculate the satellite's spatial position vector relative to the candidate gateway node during the election period based on the satellite ephemeris information. Then, they calculate the satellite's elevation angle at any given moment during the election period from the satellite's spatial position vector, expressed as:

[0075] ;

[0076] in, For a moment Satellite elevation angle, Let be the spatial location vector of the candidate gateway node. For a moment The satellite's spatial position vector, Let be the local zenith direction unit vector at the location of the candidate gateway node. For vector magnitude operations, The arcsine function is used to map sine values ​​to angle values. This represents a point in time within the election cycle, with values ​​ranging from the start to the end of the current election cycle.

[0077] Candidate gateway nodes calculate the remaining visibility window time during the election period for the satellite to maintain communication conditions, based on the satellite elevation angle and the minimum elevation angle requirement allowed for satellite communication. This window is expressed as:

[0078] ;

[0079] ;

[0080] in, This refers to the current moment when the candidate gateway node performs an availability assessment during the election cycle. From The moment when the satellite elevation angle first falls below the minimum elevation angle required for satellite communication. To allow for the minimum elevation angle required for satellite communications, This marks the end of the current election cycle. For only containing A collection of single elements For the set union operation, This is a minimum value operation, used to select the minimum time value from the set within parentheses as the output. It is applied when no such value exists within the current election cycle. At that moment, take ; From The satellite maintains the remaining visible window time that meets communication requirements during the election cycle.

[0081] The minimum elevation angle requirement for satellite communication is a parameter determined based on the spatial geometry of the satellite communication methods supported by the candidate gateway node. It limits the candidate gateway node to performing satellite communication operations only when the satellite elevation angle meets the communication requirements. The minimum elevation angle is determined by the antenna pattern corresponding to the antenna type and installation attitude of the candidate gateway node, the minimum signal-to-noise ratio requirement for stable demodulation at the receiver, and the obstruction conditions of the deployment environment. An example range of minimum elevation angle values ​​is provided. The minimum elevation angle is not less than To avoid link instability caused by near-horizon multipath and occlusion, the minimum elevation angle is no greater than [value missing]. This is to avoid overly conservative practices that could compress the satellite communication time window.

[0082] When the satellite elevation angle is lower than the minimum elevation angle required to perform satellite communication, or when the remaining visible window time for the satellite to meet the communication conditions during the election cycle is less than the duration required to complete a data collection, satellite communication link establishment, and data reporting process, the current satellite visibility status is determined to not meet the communication conditions.

[0083] The duration required to complete a data aggregation, satellite communication link establishment, and data reporting process consists of communication preparation time, link establishment time, and data transmission time. Communication preparation time represents the time it takes for the off-network gateway node to transition from a communication preparation state to a transmittable state. Link establishment time represents the time it takes to complete satellite search, synchronization, handshake, and link parameter negotiation. Both communication preparation time and link establishment time are obtained by the off-network gateway node through local timing during historical reporting processes and are stored locally on the node as statistical durations under the current environment. Data transmission time is calculated by the ratio of the amount of data to be reported to the effective transmission rate. The amount of data to be reported is determined by the length of the data queue to be sent combined with the number of bytes in a single data record. The effective transmission rate is jointly determined by the rate level corresponding to the satellite modulation and coding scheme and the available rate corresponding to the link quality. The actual value is the smaller of the two available rates. To cover the additional time caused by retransmissions and short-term jitter, a safety redundancy time is added to the calculated data transmission time. The safety redundancy time can be obtained by multiplying the historical retransmission count and the average retransmission time.

[0084] If the satellite visibility status does not meet the communication conditions, the candidate gateway node marks the gateway evaluation index as unavailable. If the satellite visibility status meets the communication conditions, the remaining power of the candidate gateway node is considered. If the remaining power of the candidate gateway node in the latest sample is lower than the safe minimum power required for the gateway role to operate, the gateway evaluation index is marked as unavailable.

[0085] When the satellite visibility condition meets the communication requirements and the remaining power of the candidate gateway node in the latest sample meets the safety minimum power requirement for the gateway role to operate, the gateway evaluation index remains in an available state.

[0086] After determining the satellite visibility status and remaining power, the original gateway evaluation index values ​​are retained, and an availability status flag is added to the gateway evaluation index, thus obtaining the constrained gateway evaluation index.

[0087] S3. Perform dynamic off-network gateway election based on constrained gateway evaluation indicators, determine the off-network gateway node, and have the off-network gateway node broadcast gateway role announcement information to control non-gateway nodes to aggregate the data to be sent to the off-network gateway node through local multi-hop links.

[0088] Furthermore, during the election cycle, the off-network self-organizing network performs validity screening on the constrained gateway evaluation metrics of the candidate gateway node set that are in an available state.

[0089] The validity screening is only performed on the constrained gateway evaluation metrics marked as available. Constrained gateway evaluation metrics marked as unavailable do not participate in the off-network gateway election in the current election cycle.

[0090] By comparing the values ​​of the constrained gateway evaluation metrics in the available state, the candidate gateway node with the highest value of the constrained gateway evaluation metric is selected as the off-network gateway node.

[0091] When multiple candidate gateway nodes have the same constrained gateway evaluation metric value, the candidate gateway node whose hierarchical position is closer to the off-network self-organizing network center is selected as the off-network gateway node.

[0092] Furthermore, after the off-network gateway node is determined, the off-network gateway node broadcasts gateway role announcement information within the off-network self-organizing network, announcing the gateway role allocation status for the current election cycle to other nodes in the off-network self-organizing network.

[0093] After receiving the gateway role announcement, non-gateway nodes in an off-grid self-organizing network update their local node role status to confirm that they will not assume the gateway role during the current election cycle.

[0094] Furthermore, after completing the gateway role announcement, the off-network self-organizing network enters the data aggregation control phase; non-gateway nodes forward the data to be sent generated or cached by their own nodes to the off-network gateway node via the local multi-hop communication link in the off-network self-organizing network.

[0095] During data forwarding, non-gateway nodes only perform local communication and data forwarding operations, and do not perform satellite communication link establishment and data reporting operations; off-network gateway nodes receive data from multiple non-gateway nodes and perform aggregation processing on the received data to form data to be reported.

[0096] It should be noted that during the election cycle, when the constrained gateway evaluation index status of the off-network gateway node changes, or when the off-network gateway node is unable to maintain the gateway role operation status, the off-network self-organizing network triggers a new election cycle, re-executes the process of comparing the constrained gateway evaluation index and determining the off-network gateway node, and realizes the dynamic adjustment of the off-network gateway node in the off-network self-organizing network.

[0097] S4. After selecting the off-network gateway node, calculate the satellite communication time window of the off-network gateway node based on the satellite ephemeris information, and trigger the off-network gateway node to perform satellite communication link establishment and data reporting within the satellite communication time window. Outside the satellite communication time window, control the off-network gateway node to enter hibernation and wake up according to the preheating time, thus completing the data reporting timing decision and hibernation wake-up control.

[0098] Furthermore, after the off-network gateway node is determined, the off-network gateway node combines the node's local geographical location, the node's local real-time clock, and satellite ephemeris information to calculate the continuous time interval in which the satellite maintains communication conditions relative to the off-network gateway node in the time dimension, and determines the continuous time interval as the satellite communication time window.

[0099] The satellite communication time window is represented by a start time and an end time. The start time corresponds to the moment when the satellite elevation angle first meets the communication conditions, and the end time corresponds to the moment when the satellite elevation angle no longer meets the communication conditions. The satellite communication time window is used to limit the time range within which the off-network gateway node is allowed to perform satellite communication link establishment and data reporting operations.

[0100] The off-network gateway node uses the relationship between the current time and the satellite communication time window as the basis for data reporting timing decisions. If the current time is within the satellite communication time window, the data reporting operation is executed; if the current time is not within the satellite communication time window, the data reporting operation is not executed.

[0101] Furthermore, within the satellite communication time window, the off-network gateway node performs satellite communication link establishment operations and sends the collected data to be reported to the remote receiving end through the satellite communication link; outside the satellite communication time window, the off-network gateway node does not perform satellite communication link establishment and data reporting operations, and enters a low-power operation state, only maintaining local timing and status monitoring functions.

[0102] When not performing data reporting operations, the off-network gateway node maintains a low-power operating state and remains in a low-power operating state until the warm-up time arrives. After the warm-up time arrives, the off-network gateway node exits the low-power operating state and enters the communication preparation state.

[0103] The off-grid gateway node uses the satellite ephemeris information to predict the moment when the satellite elevation angle first meets the communication conditions, and takes this moment as the start time of the next satellite communication time window. It also calculates the warm-up time based on the startup time required for the off-grid gateway node to recover from low-power operation to satellite communication capability, expressed as:

[0104] ;

[0105] in, The warm-up time indicates the trigger moment when the off-network gateway node enters the communication preparation state. This is the start time of the next satellite communication time window. The startup time required for an off-network gateway node to recover from a low-power operating state to satellite communication capability.

[0106] The startup time required for an off-grid gateway node to recover from a low-power operating state to satellite communication capability refers to the time required for the off-grid gateway node to complete the restoration of communication functions, satellite reception preparation, and communication parameter initialization after ending the low-power operating state. The startup time is determined by the hardware configuration and communication startup process of the off-grid gateway node. It is recorded or measured locally during node operation and stored locally on the node.

[0107] This embodiment also provides an off-network gateway dynamic election system based on edge AI and satellite ephemeris, including:

[0108] The candidate filtering module is used to filter nodes with gateway capabilities in an off-grid ad hoc network to form a set of candidate gateway nodes and to construct the node state feature vector of the candidate gateway nodes.

[0109] The intelligent evaluation module is used to perform lightweight intelligent inference on the edge side based on node state feature vectors, generate gateway evaluation metrics, and apply gateway availability constraints.

[0110] The gateway election module is used to determine the off-network gateway node based on constrained gateway evaluation metrics, and to control non-gateway nodes to gather data to be sent to the off-network gateway node.

[0111] The timing control module is used to calculate the satellite communication time window based on satellite ephemeris information and to execute data reporting timing decisions and sleep / wake-up control.

[0112] In summary, this invention achieves dynamic characterization of the comprehensive capabilities of the gateway role in off-grid ad hoc networks by using a lightweight intelligent model at the edge to infer and evaluate the multi-source operating status of nodes. This enables the gateway election results to adaptively adjust according to changes in communication conditions, energy status, and load. By using satellite ephemeris information to predict communication availability time and implementing time window and sleep / wake-up control, it achieves precise matching between gateway node communication behavior and satellite availability. This ensures that data reporting is executed in an orderly manner during effective periods and reduces energy consumption during non-communication periods, thereby improving the overall efficiency of satellite communication utilization and the reliability of node operation in off-grid environments.

[0113] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A dynamic election method for off-network gateways based on edge AI and satellite ephemeris, characterized by: include, In an off-grid self-organizing network, a set of candidate gateway nodes is formed, and multi-source operating status information is collected for each candidate gateway node in the set. The node status feature vector is constructed according to the order of the fields. The collected multi-source operational status information includes satellite communication observation information, off-grid self-organizing network status information, energy status information, and operational load information. Among them, satellite communication observation information includes satellite signal-to-noise ratio and the number of visible satellites; off-grid self-organizing network status information includes the hierarchical position of candidate gateway nodes in the off-grid self-organizing network and channel occupancy; energy status information includes remaining power and energy acquisition rate; and operational load information includes the length of the data queue to be sent. Based on node state feature vectors, intelligent reasoning is performed at the edge to generate gateway evaluation metrics, and gateway availability constraints are applied to the gateway evaluation metrics in combination with satellite ephemeris information to form constrained gateway evaluation metrics. The step of generating gateway evaluation metrics by performing intelligent inference at the edge includes, during the election period, inputting the node state feature vector corresponding to each candidate gateway node in the candidate gateway node set into a lightweight intelligent model deployed locally on the node, and performing forward inference operations. The lightweight intelligent model is constructed using a lightweight fully connected neural network structure. The fully connected neural network consists of an input layer, several hidden layers, and an output layer. The dimension of the input layer is consistent with the dimension of the node state feature vector, and the dimension of the output layer is consistent with the number of operating modes in the discrete action space. During the forward inference process, a state space is constructed based on the node state feature vector. The value of each operation mode in the discrete action space is evaluated by combining the reward function to obtain the action value estimate of each operation mode. The action value estimate of the active gateway mode is extracted from the action value estimate of each operation mode and used as the gateway evaluation index of the candidate gateway node in the current election cycle. Dynamic off-network gateway election is performed based on constrained gateway evaluation metrics to determine off-network gateway nodes. The off-network gateway nodes then broadcast gateway role announcement information to control non-gateway nodes to aggregate data to be sent to the off-network gateway nodes via local multi-hop links. The step of performing dynamic off-network gateway election based on constrained gateway evaluation indicators to determine the off-network gateway node includes selecting the candidate gateway node with the highest value of the constrained gateway evaluation indicator as the off-network gateway node by comparing the values ​​of the constrained gateway evaluation indicators in the available state. After selecting the off-network gateway node, the satellite communication time window of the off-network gateway node is calculated based on the satellite ephemeris information. Within the satellite communication time window, the off-network gateway node is triggered to establish a satellite communication link and report data. Outside the satellite communication time window, the off-network gateway node is controlled to enter hibernation and wake up according to the preheating time, thus completing the data reporting timing decision and hibernation wake-up control.

2. The off-network gateway dynamic election method based on edge AI and satellite ephemeris as described in claim 1, characterized in that: The formation of the candidate gateway node set includes, during the operation of the off-grid ad hoc network, performing gateway capability screening based on the node communication capabilities, online status, and power status in the off-grid ad hoc network to form a candidate gateway node set.

3. The off-network gateway dynamic election method based on edge AI and satellite ephemeris as described in claim 2, characterized in that: The process of applying gateway availability constraints to gateway evaluation metrics by combining satellite ephemeris information to form constrained gateway evaluation metrics includes the following steps: after generating gateway evaluation metrics, candidate gateway nodes combine their local geographic location and local real-time clock to calculate the spatial position vector of the satellite relative to the candidate gateway node during the election period based on satellite ephemeris information, and calculate the satellite elevation angle of the satellite at any time during the election period from the satellite spatial position vector. Based on the satellite elevation angle and the minimum elevation angle requirement for satellite communication, calculate the remaining visible window time during the election cycle to ensure communication. When the remaining visible window time is less than the duration required to complete a data collection, satellite communication link establishment and data reporting process, or when the remaining power of the candidate gateway node in the latest sample is lower than the safe minimum power required for the gateway role to operate, the gateway evaluation index will be marked as unavailable. If the remaining visible window time is not less than the duration required to complete one data collection, satellite communication link establishment and data reporting process, and the remaining power of the candidate gateway node in the latest sampling is not less than the safe lower limit power required for the gateway role to operate, the original gateway evaluation index value is retained, and an available status flag is added to the gateway evaluation index to obtain the constrained gateway evaluation index.

4. The off-network gateway dynamic election method based on edge AI and satellite ephemeris as described in claim 3, characterized in that: The control of non-gateway nodes to aggregate data to be sent to off-net gateway nodes via local multi-hop links includes: after receiving gateway role notification information, non-gateway nodes in the off-net ad hoc network update node role status, do not perform satellite communication link establishment and data reporting operations during the current election cycle, and forward the data to be sent generated or cached by this node to the off-net gateway node via local multi-hop communication links in the off-net ad hoc network. Off-network gateway nodes receive data to be sent forwarded from non-gateway nodes and perform aggregation processing on the received data to be sent, forming data to be reported.

5. The off-network gateway dynamic election method based on edge AI and satellite ephemeris as described in claim 4, characterized in that: The satellite communication time window for calculating the off-network gateway node based on satellite ephemeris information includes taking the moment when the satellite elevation angle first meets the minimum elevation angle requirement for satellite communication as the start time of the satellite communication time window, and taking the moment when the satellite elevation angle no longer meets the minimum elevation angle requirement for satellite communication as the end time of the satellite communication time window. The satellite communication time window is determined by the start time and the end time.

6. A dynamic election system for off-grid gateways based on edge AI and satellite ephemeris, based on the dynamic election method for off-grid gateways based on edge AI and satellite ephemeris as described in any one of claims 1 to 5, characterized in that: include, The candidate filtering module is used to filter nodes with gateway capabilities in an off-grid ad hoc network to form a set of candidate gateway nodes and to construct the node state feature vector of the candidate gateway nodes. The intelligent evaluation module is used to perform lightweight intelligent inference on the edge side based on node state feature vectors, generate gateway evaluation metrics, and apply gateway availability constraints. The gateway election module is used to determine the off-network gateway node based on constrained gateway evaluation metrics, and to control non-gateway nodes to gather data to be sent to the off-network gateway node. The timing control module is used to calculate the satellite communication time window based on satellite ephemeris information and to execute data reporting timing decisions and sleep / wake-up control.

Images