Underwater optical network routing optimization method based on distance perception and motion prediction

Abstract

The invention discloses an underwater optical network routing optimization method based on distance perception and motion prediction, relates to the technical field of underwater wireless optical communication networks, and aims to solve the problems in the prior art. The method realizes intelligent routing optimization through signal attenuation distance perception, multi-sensor trajectory prediction and opportunistic routing decision. Over-water and under-water nodes establish optical communication links through LED emitters and photoelectric detectors, and track sensing is achieved by configuring various motion sensors. Comprising six core steps of physical connection establishment, signal attenuation distance perception, multi-sensor fusion trajectory prediction, hierarchical network state convergence, multi-target optimization opportunity routing decision and adaptive parameter optimization. Physical layer distance sensing and prediction driving routing are innovatively combined, the key technical problems that a traditional underwater optical communication routing algorithm cannot adapt to node dynamic drift, lacks accurate distance sensing and cannot predict network evolution are solved, and the communication performance and energy consumption efficiency of the dynamic underwater optical network are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of underwater wireless optical communication network technology, and specifically to an underwater optical network routing optimization method based on distance sensing and motion prediction. Background Technology

[0002] With the deepening development of marine resource development and marine scientific research, underwater wireless optical communication technology has become a key supporting technology in fields such as marine engineering, environmental monitoring, and underwater robotics. Underwater wireless optical communication (UOWC) is considered one of the most promising underwater communication technologies due to its advantages such as high bandwidth, low latency, and strong security. Currently, underwater optical communication systems mainly use LEDs or laser diodes as light sources, transmitting data in the blue-green light band (450-550nm). Regarding routing protocols, existing technologies mainly include depth routing protocols (DBR) and energy-efficient depth routing protocols (EEDBR). The DBR protocol uses node depth information as a forwarding decision parameter, prioritizing forwarding data packets to nodes with shallower depths. EEDBR, based on this, considers the remaining energy of nodes, selecting nodes with lower depths and sufficient energy as the next hop. In addition, there are location-based routing protocols that utilize node GPS coordinates or acoustic positioning information for routing decisions.

[0003] However, existing underwater optical communication routing technologies have significant drawbacks: First, most existing routing algorithms are based on the assumption that node locations are fixed, making them unsuitable for dynamic network environments where nodes drift freely with ocean currents. Traditional protocols struggle to cope with frequent changes in network topology due to ocean currents and tides. Second, they lack real-time and accurate inter-node distance measurement mechanisms, relying primarily on hop count or fixed power models for routing decisions, which fail to reflect actual transmission costs and reliability. Third, existing protocols employ a passive response strategy, lacking the ability to predict future network conditions and only adjusting routes after link failures, resulting in high data loss rates. Finally, they lack sufficient adaptability to changes in the optical characteristics of the underwater environment and lack a multi-objective optimization mechanism that comprehensively considers communication quality, transmission delay, and energy consumption.

[0004] Chinese patent publication number "CN 119520362 A" is titled "Design Method of Underwater Routing Protocol Based on Link Awareness." This technology divides underwater wireless sensors into families based on the assigned roles of node control data packets, establishes a data transmission mechanism between master and slave nodes, and introduces a reward function to optimize routing decisions. While this method considers inter-node cooperation and routing optimization to some extent, it still has certain shortcomings: it primarily targets relatively fixed family network structures and cannot effectively handle dynamic topology changes in scenarios where nodes are freely drifting; it lacks accurate real-time distance perception capabilities between nodes, still relying on preset network structures and static parameters; it does not consider node trajectory prediction, making it unable to anticipate network topology evolution trends; and it lacks a multi-sensor fusion environmental perception mechanism, limiting its adaptability to complex underwater environments. Therefore, there is an urgent need to develop an intelligent underwater optical communication routing optimization method capable of real-time node distance perception, trajectory prediction, and dynamic adaptation to network topology changes to meet the growing demands of dynamic marine networking applications.

[0005] To address the problems of routing failure caused by dynamic changes in network topology under node-free drifting environments, lack of real-time accurate distance perception between nodes, inability to predict future network state changes, and poor adaptability to complex underwater environments, this invention provides an underwater optical network routing optimization method based on distance perception and motion prediction. This method can solve key technical challenges such as intelligent routing decision-making, real-time distance perception and trajectory prediction, network topology evolution prediction, and multi-objective routing optimization in dynamic underwater environments, meeting the intelligent networking needs of dynamic underwater optical communication networks for marine environmental monitoring, underwater robot collaboration, and marine ranching management.

[0006] The technical solution of this invention to solve the technical problem is as follows: An underwater optical network routing optimization method based on distance perception and motion prediction includes the following hardware configuration and steps.

[0007] The method is based on an optical communication network composed of underwater freely drifting nodes. Each node is equipped with: a blue-green LED emitter, a photodetector, a three-axis accelerometer, a three-axis gyroscope, a three-axis magnetometer, a pressure sensor, a microprocessor, and a wireless communication module. Nodes communicate point-to-point via optical signals. The method includes the following steps: S1: Physical connection establishment: The node sends an optical signal through the LED transmitter, and the photodetectors of the adjacent nodes receive the optical signal and demodulate the data to establish a physical communication link between the nodes.

[0008] S2: Signal attenuation distance sensing: The transmitting node inserts a standardized distance probe pulse sequence before the data frame, and the receiving node calculates the distance between nodes in real time by analyzing the degree of pulse power attenuation.

[0009] S3: Node trajectory prediction: Each node collects its own motion state data through sensors, and integrates information from multiple sensors to establish a trajectory prediction model to predict the node's position and motion trend in the future time period.

[0010] S4: Layered network state aggregation: Adopting a four-layer architecture of individual layer - neighborhood layer - region layer - whole network layer, the trajectory information of individual nodes is aggregated layer by layer to form a full network motion situation awareness.

[0011] S5: Opportunistic Routing Decision: Based on real-time distance information, trajectory prediction results, and network status aggregation information, a comprehensive routing score is calculated to select the optimal transmission path, realizing prediction-driven opportunistic routing.

[0012] S6: Adaptive parameter optimization: Dynamically adjust algorithm parameters based on network performance feedback, including routing weight coefficients, prediction model parameters, and transmission power parameters.

[0013] The beneficial effects of this invention are: 1. Significantly improves communication performance and network reliability: Accurate real-time distance perception is achieved through multi-frequency component pulse attenuation analysis algorithm; the trajectory prediction algorithm based on multi-sensor data fusion improves prediction accuracy, realizing a fundamental shift from passive response routing to active predictive routing; through opportunistic routing decision and multi-objective optimization algorithm, the data packet delivery rate is effectively improved, and the end-to-end transmission delay is reduced, effectively solving the communication reliability problem in dynamic underwater environments.

[0014] 2. Significantly optimized energy efficiency and network survivability: Through adaptive parameter optimization algorithms and intelligent power control mechanisms, the overall network energy consumption is reduced by 25-45%, and the network survival time is extended by 60-80%, meeting the needs of long-term unattended underwater deployment; the hierarchical network status aggregation mechanism can perceive changes in the water environment in real time and automatically adjust algorithm parameters, maintaining stable communication performance under different sea areas, depths, and time conditions, and possessing strong environmental adaptability.

[0015] 3. Achieve system-level collaborative optimization and wide applicability: The algorithm has low complexity and high computational efficiency, supporting large-scale network deployment with hundreds of nodes; it organically combines distance perception at the physical layer, trajectory prediction at the network layer, and routing decision at the application layer, achieving cross-layer collaborative optimization and system-level performance optimization; it is particularly suitable for dynamic underwater optical communication network application scenarios such as marine environmental monitoring, underwater robot swarm collaboration, marine ranch management, and underwater archaeological exploration, providing key technical support for the intelligent development of underwater IoT. Attached Figure Description

[0016] Figure 1This is a block diagram of the hardware structure of an underwater optical communication node for an underwater optical network routing optimization method based on distance perception and motion prediction, as described in this invention.

[0017] Figure 2 This is a diagram illustrating the topology and working principle of the underwater optical communication network described in this invention.

[0018] Figure 3 This is a flowchart of an underwater optical network routing optimization method based on distance perception and motion prediction, as described in this invention.

[0019] Figure 4 This is a flowchart of a multi-sensor data fusion algorithm for an underwater optical network routing optimization method based on distance perception and motion prediction, as described in this invention.

[0020] Figure 5 This is a flowchart of the opportunistic routing decision algorithm for an underwater optical network routing optimization method based on distance perception and motion prediction, as described in this invention. Detailed Implementation

[0021] The present invention will now be described in further detail with reference to the accompanying drawings.

[0022] like Figure 1 The diagram shows the hardware structure of a single underwater node, which includes an LED transmitter array 1 responsible for transmitting light signals, a photodetector 2 for receiving light signals, a sensor module 3 for real-time acquisition of the node's motion status, a communication module 4 for processing signal modulation and demodulation, a microprocessor 5 for executing distance sensing and trajectory prediction algorithms, a data storage module 6 for caching routing information, and a power system 7 for supplying power to all modules. Figure 2 The diagram illustrates the overall topology and working principle of the underwater optical communication network. 8 represents all rectangular nodes on the surface, 9 represents all circular nodes underwater (which are freely drifting communication nodes), and 10 represents the direction of the main ocean current influencing node movement. The red dashed lines show the historical and predicted movement paths of the nodes, the dashed frame represents the four-layer architecture of hierarchical network state convergence, and the green arrows show the movement trends of local nodes. Both figures demonstrate that this method, through real-time distance sensing, multi-sensor trajectory prediction, and hierarchical network state convergence, achieves intelligent opportunistic routing decisions in dynamic underwater environments, effectively solving the technical challenge of traditional underwater optical communication routing algorithms being unable to adapt to freely drifting nodes.

[0023] like Figure 3 As shown, an underwater optical network routing optimization method based on distance awareness and motion prediction is presented, and the process is as follows: S1: Physical Connection Establishment: Nodes modulate digital signals into light intensity variations using high-efficiency blue-green LED transmitters. The light signals propagate in the water at wavelengths of 450-550nm. High-sensitivity photodetectors at adjacent nodes capture the light signals and convert them into electrical signals. After amplification, filtering, and demodulation, the original data is recovered, thus establishing a point-to-point physical communication link between nodes. During this process, the transmission power can be dynamically adjusted between multiple levels, and the communication distance can reach hundreds of meters in clear seawater. The system employs differential coding and forward error correction techniques to improve transmission reliability, ensuring stable data exchange in complex underwater optical environments.

[0024] S2: Signal Attenuation Distance Sensing: The transmitting node inserts a standardized distance detection pulse sequence containing multiple different frequency components before each data frame, with each frequency component transmitted at a preset standard power. The photodetector of the receiving node measures the received power of each frequency component, and calculates the signal attenuation by comparing the difference between the transmitted and received power. Based on Beer-Lambert's law of light propagation in water, combined with the absorption and scattering coefficients of water, the precise distance between nodes is calculated in real time. The system uses the least squares method to fuse the ranging results of multiple frequency components and performs dynamic calibration using node pairs with known distances to achieve high-precision distance sensing, providing an accurate spatial information basis for subsequent routing decisions.

[0025] S3: Node Trajectory Prediction: Each node synchronously collects its own motion state data through a configured sensor suite, including motion acceleration measured by a three-axis accelerometer, angular velocity information acquired by a three-axis gyroscope, motion direction determined by a three-axis magnetometer, and depth changes monitored by a pressure sensor. An adaptive weighted fusion algorithm is used to integrate multi-sensor information, dynamically adjusting the weight coefficients of each sensor to adapt to different environmental conditions. By analyzing the motion correlation of neighboring nodes, an ocean current influence model is established to identify the passive drift characteristics of nodes driven by ocean currents. Based on historical motion data and current state, a comprehensive prediction model incorporating position, velocity, acceleration, and ocean current influence is established to predict the node positions and motion trends at multiple future time points, providing forward-looking information support for network topology evolution.

[0026] S4: Hierarchical Network State Aggregation: A four-layer distributed architecture—Individual Layer, Neighborhood Layer, Regional Layer, and Global Layer—is adopted to achieve hierarchical aggregation and abstraction of network state information. The Individual Layer maintains detailed motion states, historical trajectories, and prediction information for each node. The Neighborhood Layer aggregates trajectory information from direct neighbors using a weighted average algorithm, forming local motion situation awareness. The Regional Layer uses clustering algorithms to identify collective motion patterns of nodes over a large area, including typical ocean current characteristics such as overall drift, rotational motion, and diffusion motion. The Global Layer uses a distributed consensus algorithm to fuse information from each region into a global network state model, generating network topology evolution predictions and a global motion trend map. A differential information transmission mechanism is used between layers, transmitting only state changes to reduce communication overhead. Simultaneously, an anomaly detection mechanism is established to identify equipment failures or special ocean phenomena, ensuring the accuracy and timeliness of network state information.

[0027] S5: Opportunistic Routing Decision: Based on real-time distance information, trajectory prediction results, and hierarchical network state aggregation information, a multi-objective optimization routing scoring system is constructed. First, all candidate transmission paths are discovered. Then, a comprehensive scoring function is calculated, including distance score, trajectory matching score, link stability score, and energy efficiency score. The weights of each score are dynamically adjusted according to data priority and network state. The distance score evaluates transmission cost based on an exponential decay function; the trajectory matching score calculates the consistency of node movement direction using a cosine function; the link stability score predicts the proportion of available link time in the future; and the energy efficiency score balances transmission performance with the remaining energy of nodes. An opportunity waiting strategy is implemented, determining the transmission timing by comparing current transmission benefits with expected future benefits. When predictions indicate a better path will be available later, delayed transmission is chosen. Finally, the optimal path or transmission timing is selected based on the comprehensive score, achieving prediction-driven intelligent opportunistic routing and significantly improving packet delivery rate.

[0028] S6: Adaptive Parameter Optimization: A closed-loop optimization mechanism based on network performance feedback is established to monitor key performance indicators such as packet delivery rate, end-to-end latency, energy efficiency, and prediction accuracy in real time. When performance indicators deviate from preset target values, gradient descent and reinforcement learning algorithms are used to dynamically adjust system parameters. For routing weight coefficients, the weight allocation of each scoring function is adjusted according to the delivery rate deviation; for trajectory prediction models, sensor fusion weights and prediction time windows are optimized based on prediction accuracy feedback; for distance sensing algorithms, attenuation model parameters are adjusted using ranging error statistics; for transmit power control, dynamic power adjustment is achieved based on network load and node remaining energy. The network state transition probability matrix is ​​used to predict the future evolution trend of the network topology, providing forward-looking guidance for parameter optimization. The entire optimization process uses adaptive step size control to ensure the stability of parameter convergence, enabling the system to continuously adapt to changes in the marine environment and dynamic network evolution, achieving the optimization effect of extending network lifetime.

[0029] In S1, the physical connection establishment process includes: the LED transmitter of the sending node modulates the digital signal into a change in light intensity. The LED operates in the 450-550nm blue-green light band and supports dynamic power adjustment. The light signal propagates in the water and is captured by the photodetector of the receiving node. The photodetector uses a PIN photodiode or an APD avalanche photodiode. The photodetector converts the light signal into an electrical signal and recovers the original data through demodulation, completing one communication process. The communication distance between nodes is affected by the optical properties of the water.

[0030] In S2, distance sensing is based on the attenuation law of light signals propagating in water. A multi-frequency component pulse attenuation analysis method is used, detecting pulse sequences containing 3-5 different frequency components. The distance calculation formula is as follows: in: D The distance between nodes; α(λ) The water absorption coefficient; β(λ) The water scattering coefficient; Prx For received power; Ptx This refers to the transmission power. Gtx For transmit gain; Grx For receiving gain; Δ env Environmental correction factor; λ λ is the wavelength of light.

[0031] In S3, node trajectory prediction employs a multi-sensor data fusion algorithm, such as... Figure 4 As shown, the specific algorithm steps include: S31: Multi-sensor synchronous data acquisition: Simultaneously acquires raw data from a triaxial accelerometer, triaxial gyroscope, triaxial magnetometer, and pressure sensor. All sensors use a unified clock reference for synchronous sampling to ensure consistent data timestamps. The accelerometer detects linear motion changes at the node, the gyroscope monitors angular velocity and rotational state, the magnetometer provides orientation reference and attitude information, and the pressure sensor reflects the node's vertical displacement and depth changes. Sensor data undergoes low-pass filtering to eliminate high-frequency noise and median filtering to remove outlier interference.

[0032] S32: Sensor Data Fusion: Integrating multi-sensor information using an adaptive weighted fusion algorithm. in: for t The motion state vector fused at each moment; For the first j Sensor measurement vectors; For the first j The dynamic weighting coefficients of the sensor satisfy ;M This represents the total number of sensor types.

[0033] The weighting coefficients are dynamically adjusted based on the real-time accuracy, environmental adaptability, and historical reliability of each sensor. In calm waters, the accelerometer has a higher weight; in turbulent environments, the gyroscope and magnetometer have increased weights; and in deep water, the pressure sensor has a higher weight. The fusion algorithm employs a Kalman filter framework, continuously optimizing the state estimation through a prediction-update loop. Simultaneously, a sensor fault detection mechanism is established, automatically reducing the weight of a sensor or removing its data when an anomaly occurs.

[0034] S33: Ocean Current Impact Modeling: Establishing an ocean current impact model by analyzing the motion correlation of neighboring nodes: in: K This represents the number of neighboring nodes; γk For the first k The influence weight of each neighboring node; Let k be the velocity vector of the kth neighbor node; This is the velocity vector of this node.

[0035] Ocean current modeling algorithms identify common environmental driving forces by comparing the velocity vector differences of neighboring nodes. When multiple neighboring nodes exhibit similar motion patterns, it indicates the presence of a significant ocean current influence. Influence weights. γk The contribution of nodes to current estimation is inversely proportional to the distance between them; closer nodes contribute more to the estimation. The algorithm also considers the time delay effect, applying appropriate time weights to nodes at different distances to reflect the physical characteristics of current propagation. Furthermore, a current persistence model is established to predict future current trends using historical current data.

[0036] S34: Trajectory Prediction Calculation: Establishing a motion prediction model to predict future trajectories. in: for t Time position vector; Let be the velocity vector at time t; Let t be the acceleration vector. This represents the vector of ocean current influence. It is a random error vector; For the predicted time interval.

[0037] The trajectory prediction model combines classical kinematic equations with marine environmental characteristics. Prediction time interval. The time interval is dynamically adjusted based on the stability of the node motion; a longer interval (5-10 minutes) is used when the motion is stable, and shortened to 1-3 minutes when the motion is intense. Random error vector. A Gaussian white noise model is employed, with its variance determined statistically based on historical prediction accuracy. The algorithm also incorporates a prediction confidence assessment mechanism, calculating the confidence level of each prediction based on sensor data quality, ocean current model reliability, and historical prediction accuracy, providing uncertainty quantification information for subsequent routing decisions. Prediction results include location point predictions and trajectory probability distributions, supporting various prediction accuracy requirements.

[0038] In S4, the hierarchical network state convergence process includes: Individual-level state maintenance: Each node maintains its own motion state information, including position history, velocity changes, and predicted trajectory. A circular buffer is used to store the most recent historical data. The data structure includes timestamps, 3D position coordinates, velocity vectors, acceleration information, and motion confidence scores. Nodes periodically perform data compression and feature extraction, retaining key motion pattern information to reduce storage overhead.

[0039] Neighborhood information aggregation: Nodes calculate local movement trends by aggregating neighborhood information. in: This represents the neighborhood motion trend vector. For the first n The trajectory vectors of the neighboring nodes; For the first n The weights of each neighboring node; Nneighbor This represents the number of neighboring nodes.

[0040] Weight The algorithm is determined based on a comprehensive consideration of inter-node communication reliability, the reciprocal of distance, and historical data consistency. It employs an exponential decay mechanism, giving higher weight to newer trajectory information. To avoid information redundancy and oscillations, a neighborhood information update threshold is set; convergence updates are only triggered when changes in the motion trend exceed this preset threshold.

[0041] Regional pattern recognition: This function identifies the collective motion patterns of nodes within a region, including typical ocean current patterns such as overall drift, rotational motion, and diffusion. K-means clustering and principal component analysis (PCA) are used for pattern classification. The algorithm can identify eight typical motion patterns: stationary pattern, linear drift, circular flow, divergence-diffusion, convergence-convergence, spiral motion, chaotic diffusion, and combined patterns. Automatic pattern classification and labeling are achieved by calculating the motion variance, correlation coefficient, and spectral characteristics of the node group.

[0042] Full-network situational fusion: Regional information is fused into a full-network state model using a distributed algorithm to generate network topology evolution predictions. A distributed consensus algorithm is employed to ensure the synchronization and consistency of information across regions. The fusion process considers the credibility weights of different regions and dynamically adjusts them based on historical prediction accuracy and data quality. The generated full-network state model includes a network connectivity matrix, a node distribution density map, and topology evolution trend predictions.

[0043] In S5, the opportunistic routing decision uses a multi-objective optimization algorithm, such as... Figure 5 As shown, the specific algorithm steps include: S51: Candidate Path Discovery: Based on the current network topology and trajectory prediction results, all possible data transmission paths are identified. The algorithm employs a breadth-first search strategy, expanding layer by layer from the source node, considering transmission paths of 1-4 hops. During path search, trajectory prediction information is incorporated, selecting only candidate paths that remain connected within the prediction time window, avoiding invalid paths that are about to be disconnected.

[0044] S52: Multidimensional scoring calculation: Calculate a comprehensive score for each candidate path. in: Overall route score; For dynamic weighting coefficients, satisfying ; Each sub-scoring function is defined as follows: Distance rating: ,in For distance attenuation parameters, Normalized distance; Trajectory matching score: ,in The angle between the node's direction of motion and the target direction; Link stability score: ,in To predict link stabilization time, Total forecast time; Energy efficiency rating: ,in To minimize transmission power consumption, This represents the energy required for the current path.

[0045] The weighting coefficients are dynamically adjusted based on data type and network status: urgent data receives increased distance weighting to reduce latency, ordinary data receives increased energy consumption weighting to conserve energy, and network congestion is mitigated by increasing link stability weighting to improve reliability. The scoring calculation employs parallel processing, simultaneously calculating scores for all candidate paths to enhance decision-making efficiency.

[0046] S53: Opportunity Waiting Strategy: Evaluate the benefits of current transmission versus waiting for a better opportunity. When the expected future benefit is significantly higher than the current benefit, delay transmission is chosen. The waiting strategy uses a benefit prediction model, calculating the probability of a better path appearing within a future time window based on trajectory prediction and historical statistical data. When the predicted benefit increases by more than 20% and the waiting time does not exceed the maximum latency tolerance of the data packet, delay transmission is chosen to wait for a better opportunity.

[0047] S54: Optimal Path Selection: Based on comprehensive scoring and opportunity analysis, the optimal transmission path or timing is selected. The decision algorithm supports multiple transmission strategies: single-path transmission, multi-path parallel transmission, and load-balanced transmission. For critical data, a redundant transmission strategy is adopted, using multiple independent paths to send data simultaneously to improve reliability.

[0048] In S6, the adaptive parameter optimization process includes: Performance Monitoring: The system monitors key performance indicators such as packet delivery rate, end-to-end latency, energy efficiency, and prediction accuracy in real time. It employs a sliding window mechanism to collect performance data, with the window size adaptively adjusting based on network dynamics. The monitoring module establishes a multi-layered performance evaluation system, encompassing instantaneous performance, short-term trends, and long-term stability. When performance anomalies are detected, a rapid response mechanism is triggered to shorten the monitoring cycle and obtain more refined performance change information. Simultaneously, a performance baseline model is established, and statistical analysis is used to identify normal fluctuations and abnormal deviations, avoiding erroneous adjustments due to natural changes in the marine environment.

[0049] Dynamic adjustment of routing weights: When performance metrics deviate from target values, the routing weights are adjusted. in: and These are the old and new weighting coefficients, respectively. To adjust the step size parameter; The target performance value; This represents the current performance value.

[0050] The weight adjustment algorithm employs a gradient descent optimization strategy, determining the adjustment direction based on the direction and magnitude of performance deviations. To prevent weight oscillations, a momentum term is introduced to record historical adjustment trends; the adjustment step size is increased when consecutive adjustment directions are consistent, and decreased when the direction changes. A weight constraint mechanism ensures that all weight coefficients always satisfy the normalization condition and sets reasonable upper and lower bounds to prevent excessive weighting of one target from causing other targets to be ignored. A soft update strategy is used during the adjustment process; new weights are merged with old weights through an exponential moving average, improving system stability.

[0051] Prediction model parameter adjustment: The trajectory prediction model parameters are adjusted based on prediction accuracy feedback. Prediction accuracy is evaluated using a multi-dimensional index system, including position prediction error, velocity prediction deviation, and direction prediction accuracy. When prediction accuracy continues to decline, a model recalibration process is initiated to re-estimate sensor fusion weights, ocean current influence coefficients, and random error variance. Parameter adjustment employs an online learning algorithm, combining least squares and Kalman filtering techniques to achieve incremental updates of model parameters. A multi-scale architecture for the prediction model is established: the short-term prediction model focuses on rapid response, while the long-term prediction model emphasizes trend capture, dynamically selecting an appropriate prediction time scale based on the application scenario.

[0052] Network topology evolution prediction: Predicting future network topology changes using the network state transition probability matrix. in: for t The network topology state probability vector at time 1; T This is the state transition matrix; Representing the transition matrix Power of 1.

[0053] The topology evolution prediction algorithm is based on Markov chain theory, abstracting the network topology into a finite set of states, including four typical states: high connectivity, medium connectivity, low connectivity, and disconnected connectivity. The state transition matrix is ​​obtained through statistical learning of historical topology data and updated in real time based on current ocean current patterns, node density, and communication quality. The prediction process considers the temporal and spatial correlations of topology changes and uses node group movement patterns to infer future connectivity changes. A topology stability assessment mechanism is established to identify critical links and bottleneck nodes, providing forward-looking network structure information for routing decisions.

[0054] Dynamic power control: LED transmission power is dynamically adjusted based on network load and node remaining energy. A distributed coordination mechanism is employed for power control, with nodes autonomously adjusting their transmission power according to neighbor density, channel quality, and energy status. A power-distance-bit error rate mapping model is established to ensure minimal energy consumption while meeting communication quality requirements. A tiered power adjustment strategy is used, setting multiple power levels for precise control, with the adjustment step size dynamically determined based on the current power level and remaining energy percentage. A load-balanced power allocation mechanism is implemented, where high-load nodes appropriately reduce power to divert data to other paths, avoiding localized overload. An energy early warning mechanism is established; when a node's energy falls below a threshold, it enters energy-saving mode, prioritizing critical data transmission and extending the overall network lifetime.

Claims

1. A routing optimization method for underwater optical networks based on distance sensing and motion prediction, characterized in that, Includes the following steps: S1: Physical connection establishment: The node sends light signals through the LED transmitter, and the photodetectors of the adjacent nodes receive the light signals and demodulate the data to establish a physical communication link between the nodes; S2: Signal attenuation distance sensing: The transmitting node inserts a standardized distance probe pulse sequence before the data frame, and the receiving node calculates the distance between nodes in real time by analyzing the degree of pulse power attenuation. S3: Node trajectory prediction: Each node collects its own motion state data through sensors, and integrates information from multiple sensors to establish a trajectory prediction model to predict the node's position and motion trend in the future time period. S4: Layered network state aggregation: A four-layer architecture of individual layer - neighborhood layer - region layer - whole network layer is adopted to aggregate the trajectory information of individual nodes layer by layer to form a full network motion situation awareness; S5: Opportunistic Routing Decision: Based on real-time distance information, trajectory prediction results and network status convergence information, calculate the comprehensive routing score to select the optimal transmission path and realize prediction-driven opportunistic routing. S6: Adaptive parameter optimization: Dynamically adjust algorithm parameters based on network performance feedback, including routing weight coefficients, prediction model parameters, and transmission power parameters.

2. The underwater optical network routing optimization method based on distance sensing and motion prediction according to claim 1, characterized in that, In step S1, the physical connection establishment process includes: the LED transmitter of the sending node modulates the digital signal into a change in light intensity; the light signal propagates in the water and is captured by the photodetector of the receiving node; the photodetector converts the light signal into an electrical signal and recovers the original data through demodulation, completing one communication process; in step S2, distance sensing is based on the attenuation law of light signal propagation in water, and the distance calculation formula is: in: D The distance between nodes; α(λ) The water absorption coefficient; β(λ) The water scattering coefficient; Prx For received power; Ptx This refers to the transmission power. Gtx For transmit gain; Grx For receiving gain; Δ env Environmental correction factor; λ λ is the wavelength of light.

3. The underwater optical network routing optimization method based on distance sensing and motion prediction according to claim 1, characterized in that, In step S3, the node trajectory prediction employs a multi-sensor data fusion algorithm, the specific algorithm steps of which include: S31: Multi-sensor data synchronous acquisition: Simultaneously acquires raw data from a triaxial accelerometer, triaxial gyroscope, triaxial magnetometer, and pressure sensor; S32: Sensor Data Fusion: Integrating multi-sensor information using an adaptive weighted fusion algorithm. in: for t The motion state vector fused at each moment; For the first j Sensor measurement vectors; For the first j The dynamic weighting coefficients of the sensor satisfy ; M This represents the total number of sensor types. S33: Ocean Current Impact Modeling: Establishing an ocean current impact model by analyzing the motion correlation of neighboring nodes: in: K This represents the number of neighboring nodes; γk For the first k The influence weight of each neighboring node; Let k be the velocity vector of the kth neighbor node; This is the velocity vector of this node; S34: Trajectory Prediction Calculation: Establishing a motion prediction model to predict future trajectories. in: for t Time position vector; Let be the velocity vector at time t; Let t be the acceleration vector at time t; This represents the vector of ocean current influence. It is a random error vector; For the predicted time interval.

4. The underwater optical network routing optimization method based on distance sensing and motion prediction according to claim 1, characterized in that, In step S4, the hierarchical network state aggregation process includes: Individual-level state maintenance: Each node maintains its own motion state information, including position history, velocity changes, and predicted trajectory; Neighborhood information aggregation: Nodes calculate local movement trends by aggregating neighborhood information. in: This represents the neighborhood motion trend vector; For the first n The trajectory vectors of the neighboring nodes; For the first n The weights of each neighboring node; Nneighbor This represents the number of neighboring nodes; Regional layer pattern recognition: Identify the collective motion patterns of nodes within a region, including typical ocean current patterns such as overall drift, rotational motion, and diffusion motion; Full network-level situational fusion: Regional information is fused into a full network state model through distributed algorithms to generate network topology evolution predictions.

5. The underwater optical network routing optimization method based on distance sensing and motion prediction according to claim 1, characterized in that, In step S5, the opportunistic routing decision employs a multi-objective optimization algorithm, the specific algorithm steps of which include: S51: Candidate Path Discovery: Identify all possible data transmission paths based on the current network topology and trajectory prediction results; S52: Multidimensional scoring calculation: Calculate a comprehensive score for each candidate path. in: Overall route score; For dynamic weighting coefficients, satisfying ; Each sub-scoring function is defined as follows: Distance rating: ,in For distance attenuation parameters, Normalized distance; Trajectory matching score: ,in The angle between the node's direction of motion and the target direction; Link stability score: ,in To predict link stabilization time, Total forecast time; Energy efficiency rating: ,in To minimize transmission power consumption, This represents the energy required for the current path. S53: Opportunity Waiting Strategy: Evaluate the benefits of current transmission versus waiting for a better opportunity, and choose to delay transmission when the expected future benefit is significantly higher than the current benefit; S54: Optimal Path Selection: Select the optimal transmission path or timing based on comprehensive scoring and opportunity analysis.

6. The underwater optical network routing optimization method based on distance sensing and motion prediction according to claim 1, characterized in that, In step S6, the adaptive parameter optimization process includes: Performance metrics monitoring: Real-time monitoring of key performance metrics such as packet delivery rate, end-to-end latency, energy efficiency, and prediction accuracy; Dynamic adjustment of routing weights: When performance metrics deviate from target values, the routing weights are adjusted. in: and These are the old and new weighting coefficients, respectively. To adjust the step size parameter; The target performance value; This represents the current performance value.

7. The underwater optical network routing optimization method based on distance sensing and motion prediction according to claim 6, further comprising: Prediction model parameter adjustment: Adjust trajectory prediction model parameters based on prediction accuracy feedback to improve prediction accuracy; Network topology evolution prediction: Predicting future network topology changes using the network state transition probability matrix. in: for t The network topology state probability vector at time 1; T This is the state transition matrix; Representing the transition matrix Power of; Dynamic power control: The LED emission power is dynamically adjusted according to the network load and the remaining energy of the node to balance communication performance and energy efficiency.

Images