A Multi-Link Transmission Strategy Optimization Method and System Based on Swarm Intelligence

Abstract

This invention belongs to the field of data transmission technology, specifically relating to a method and system for optimizing multi-link transmission strategies based on swarm intelligence. The method includes: collecting state data of each physical link and buffer state data at the receiving end, and calculating the weighted average transmission delay of all active links; determining the out-of-order blocking potential of each physical link based on the predicted one-way transmission delay, the weighted average transmission delay, and the reassembly buffer occupancy at the receiving end; obtaining the corrected heuristic function value of each physical link using the out-of-order blocking potential; calculating the state transition probability based on the pheromone concentration of each physical link and the corrected heuristic function value; selecting a target link for data packet transmission based on the state transition probability; and updating the pheromone concentration based on the transmission result. This invention can effectively detect and suppress link selection that leads to out-of-order blocking at the receiving end, ensuring bandwidth utilization while reducing the degree of transmission out-of-order, and improving transmission orderliness and reliability.

Description

Technical Field

[0001] This invention relates to the field of data transmission technology. More specifically, this invention relates to a method and system for optimizing multi-link transmission strategies based on swarm intelligence. Background Technology

[0002] With the iteration of wireless communication technologies and the growth of mobile internet services, multi-link heterogeneous network environments have become the infrastructure supporting critical services such as high-definition video and the industrial internet. By aggregating heterogeneous network resources such as WiFi, 5G, and LTE for parallel transmission, the bandwidth bottleneck of a single link can be significantly overcome and connection robustness improved. However, heterogeneous network environments are highly dynamic, with parameters such as bandwidth and latency of each physical link constantly fluctuating dramatically. This makes how to efficiently distribute data streams based on the real-time status of the links a core challenge that determines system performance.

[0003] Current mainstream multi-link transmission scheduling technologies typically make decisions based on round-trip time or instantaneous throughput, often employing a greedy logic that prioritizes filling low-latency links. However, this strategy, based on single-moment measurement data, has limitations: firstly, instantaneous measurements often contain random noise, making it difficult to accurately reflect long-term link trends and easily leading to scheduling oscillations; secondly, existing technologies generally only focus on the transmitter's detection metrics, severely neglecting the receiver's data reassembly status and buffer capacity. This one-sided perspective is particularly problematic when link differences are significant.

[0004] In practical applications, when there are large differences in transmission latency between heterogeneous links, blindly greedy scheduling can lead to serious disorder in the arrival order of data packets. This causes a large number of gaps in the receiver's reassembly buffer due to discontinuous sequence numbers, resulting in head-of-line blocking. As the out-of-order situation worsens, the buffer resources are quickly exhausted, forcing the receiver to announce a smaller receiving window, which in turn limits the sending rate, resulting in a huge waste of bandwidth resources.

[0005] Therefore, there is an urgent need for a method and system for optimizing multi-link transmission strategies based on swarm intelligence. Summary of the Invention

[0006] To address the technical problem that existing transmission strategies fail to effectively measure the interference intensity caused by link delay differences on system orderliness, resulting in severe out-of-order blocking at the receiving end and thus affecting transmission performance, this invention provides solutions in the following aspects.

[0007] In a first aspect, the present invention provides a multi-link transmission strategy optimization method based on swarm intelligence, including:

[0008] The instantaneous one-way delay, instantaneous throughput, and receiver reassembly buffer occupancy of each physical link are collected. The collected instantaneous one-way delay and instantaneous throughput are smoothed to obtain the predicted one-way transmission delay and the estimated available bandwidth. The weighted average transmission delay of all active links is calculated. Based on the predicted one-way transmission delay, the weighted average transmission delay, and the receiver reassembly buffer occupancy, the out-of-order blocking potential of each physical link is determined. Based on the estimated available bandwidth, the predicted one-way transmission delay, and the out-of-order blocking potential, the modified heuristic function value of each physical link is determined. Based on the pheromone concentration of each physical link and the modified heuristic function value, the state transition probability of each physical link is determined. The target link is selected for data packet transmission based on the state transition probability.

[0009] This invention smooths the collected instantaneous one-way delay and instantaneous throughput, and constructs a dynamic feature reflecting the long-term trend of the link and the actual carrying capacity of the receiver by combining the reassembly buffer occupancy at the receiving end. This effectively filters out measurement noise. By introducing out-of-order blocking potential energy, this invention couples the delay deviation of the link with the buffer state at the receiving end, measuring the receiver blocking cost that link selection may cause. It also uses out-of-order blocking potential energy to correct the heuristic function value, achieving automatic suppression of links with high out-of-order risk without introducing additional adjustment factors. This invention combines pheromone concentration and corrected heuristic function value to calculate the state transition probability, enabling the system to adaptively converge to the optimal path that balances bandwidth utilization and transmission order, effectively eliminating head-of-line blocking at the receiving end and improving the overall efficiency and stability of multi-link transmission.

[0010] Preferably, the step of calculating the weighted average transmission delay of all active links includes: using the proportion of the available bandwidth estimate of each physical link in the total available bandwidth as a weight, and performing a weighted summation of the predicted one-way transmission delay of each physical link to obtain the weighted average transmission delay.

[0011] This invention uses the available bandwidth ratio of each physical link as a weight to perform a weighted summation of the predicted one-way transmission delay, giving high-bandwidth links a greater say in the benchmark calculation. This allows the calculated weighted average transmission delay to more accurately reflect the actual arrival time of most data packets at the receiving end, thereby effectively avoiding the unreasonable deviation of the benchmark value by the extreme delay values ​​of low-bandwidth edge links and ensuring the accuracy of the system's judgment on the overall transmission order.

[0012] Preferably, the out-of-order blocking potential energy satisfies the expression: ;in, Indicates link In the Disordered blocking potential energy at any given moment; Indicates link In the The predicted one-way transmission delay at any given time; Indicates the first The weighted average transmission delay of all active links at any given time; Indicates the first The minimum transmission delay among all available links at any given time; To prevent tiny constants with a denominator of zero; Indicates the first The amount of reassembly buffer usage at the receiving end at any given time; This indicates the maximum capacity of the reassembly buffer at the receiving end; Represents an exponential function with the natural constant as its base; The potential energy sensitivity coefficient.

[0013] This invention constructs out-of-order blocking potential energy by nonlinearly coupling the latency deviation of the link with the buffer load rate of the receiver. It can dynamically measure the blocking cost that choosing a certain link at the current moment may cause to the receiver. When the buffer load rate of the receiver increases, it can drastically amplify the contribution of latency deviation to potential energy, so that a small timing deviation is transformed into a huge potential energy resistance, thereby keenly sensing and warning of potential head-of-line blocking risks at the receiver.

[0014] Preferably, the modified heuristic function value satisfies the expression: ;in, Indicates link In the The value of the heuristic function is adjusted at any given time. Indicates link In the The estimated available bandwidth at any given time; Indicates link In the The predicted one-way transmission delay at any given time; Indicates link In the Disorder blocking potential energy at any given moment.

[0015] This invention uses out-of-order blocking potential energy to modify the heuristic function, so that the value of the heuristic function decreases inversely with the increase of potential energy. This achieves natural suppression of high-risk links without the need to introduce additional adjustment factors. While ensuring the basic transmission performance of the link, it forcibly reduces the attractiveness of high-blocking-risk links and prevents data packets from accumulating and overflowing at the receiving end.

[0016] Preferably, the state transition probability satisfies the expression: ;in, Indicates the first Select link at any time The probability of transmission; Indicates link In the The pheromone concentration at any given time; Indicates link In the The value of the heuristic function is adjusted at any given time. This is the set of currently available links; Link index in the set of available links; For pheromone importance factors; This is a heuristic information importance factor.

[0017] Preferably, the step of selecting a target link for data packet transmission based on the state transition probability includes: the sending end selecting a target link for sending data packets using a roulette wheel method based on the state transition probability.

[0018] Preferably, the step of smoothing the collected instantaneous one-way delay and instantaneous throughput to obtain the one-way transmission delay prediction value and the available bandwidth estimate value includes: smoothing the instantaneous one-way delay sequence using an exponentially weighted moving average algorithm to obtain the one-way transmission delay prediction value at the current moment; and smoothing the instantaneous throughput using an exponentially weighted moving average algorithm to obtain the available bandwidth estimate value.

[0019] This invention utilizes an exponentially weighted moving average algorithm to smooth instantaneous one-way latency and instantaneous throughput. This effectively filters out high-frequency measurement noise in the network environment and eliminates random fluctuations in measurement data at a single moment. As a result, the predicted values ​​obtained by the system can more accurately reflect the long-term capacity trend and stability of the link, and avoid the algorithm making incorrect scheduling decisions due to instantaneous data jumps.

[0020] Preferably, it further includes updating the pheromone concentration of the target link based on the data packet transmission results.

[0021] Preferably, updating the pheromone concentration of the target link based on the data packet transmission result includes: in response to the data packets arriving at the receiving end in sequence, updating the pheromone concentration of the target link using a positive incentive strategy, wherein the updated pheromone concentration satisfies the expression: In response to the data packet causing out-of-order delivery at the receiver, the pheromone concentration of the target link is updated using a negative suppression strategy. The updated pheromone concentration satisfies the expression: ;in, Indicates the updated number Time Link pheromone concentration; The pheromone evaporation coefficient; Indicates the first Time Link pheromone concentration; Update the step size for pheromones.

[0022] This invention updates pheromone concentration by implementing a strategy of positive incentive or negative inhibition based on whether data packets arrive at the receiving end in order. It establishes a feedback learning mechanism based on transmission results, which enhances the possibility of high-performing links being selected again, while reducing the probability of selecting out-of-order links. This enables the system to automatically converge to a set of optimal transmission schemes through continuous iteration, which maximizes bandwidth utilization and controls the degree of out-of-order transmission within the tolerance range of the receiving end.

[0023] Secondly, the present invention provides a multi-link transmission strategy optimization system based on swarm intelligence, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned multi-link transmission strategy optimization method based on swarm intelligence is implemented.

[0024] By adopting the above technical solution, the above-mentioned multi-link transmission strategy optimization method based on swarm intelligence is generated into a computer program and stored in a memory so that it can be loaded and executed by a processor. In this way, a terminal device can be made based on the memory and the processor for convenient use.

[0025] The beneficial effects of this invention are as follows: By smoothing the original sampled data and combining it with the buffer information fed back from the receiving end, this invention constructs a dynamic feature set that includes link-side performance indicators and receiving-side state indicators. This solves the problems of noise in single-moment measurement data and the inability to perceive out-of-order congestion. This invention introduces out-of-order congestion potential energy, nonlinearly couples the link's delay deviation with the receiving end's buffer load rate, and corrects the heuristic function through a damping suppression model, thereby achieving natural suppression of high-risk links. This invention adopts a state transition and pheromone update mechanism based on swarm intelligence, which can adaptively optimize multi-link transmission strategies. While ensuring high throughput, it reduces the degree of out-of-order congestion at the receiving end, eliminates head-of-line congestion, and improves data transmission performance and user experience in heterogeneous network environments. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating the multi-link transmission strategy optimization method based on swarm intelligence in this invention;

[0027] Figure 2 This diagram illustrates the dynamic coupling relationship between out-of-order blocking potential and the reorganization cache usage.

[0028] Figure 3 This is a schematic diagram comparing the reassembly buffer usage at the receiving end of the present invention and existing technologies;

[0029] Figure 4 This is a schematic diagram comparing the distribution of link selection strategies between the present invention and existing technologies;

[0030] Figure 5 This is a schematic diagram comparing the end-to-end delay distribution of data packets between the present invention and existing technologies. Detailed Implementation

[0031] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0033] This invention discloses a multi-link transmission strategy optimization method based on swarm intelligence, referring to... Figure 1 This includes steps S1-S4:

[0034] S1. Collect the instantaneous one-way delay, instantaneous throughput, and reassembly buffer usage of each physical link. Smooth the collected instantaneous one-way delay and instantaneous throughput to obtain the predicted one-way transmission delay and the estimated available bandwidth. Calculate the weighted average transmission delay of all active links.

[0035] It should be noted that in a multi-link heterogeneous network environment, the transmission performance of each physical link fluctuates dynamically in real time. Measurement data at a single moment often contains noise and is difficult to reflect long-term trends. If only the sending end status is considered while ignoring the receiving end buffer status, the sending end will not be able to perceive the receiving end congestion caused by out-of-order delivery, ultimately leading to a mismatch between the transmission strategy and the actual network carrying capacity. Therefore, this invention constructs a dynamic feature set that includes link-side performance indicators and receiving-side status indicators by smoothing the original sampled data and combining it with the buffer information fed back by the receiving end.

[0036] Specifically, the following status data is collected and updated periodically:

[0037] For any physical link, the instantaneous one-way delay of the link is measured by sending a probe packet with a timestamp or by using the transmission time of the data packet; the instantaneous one-way delay sequence is smoothed by using an exponentially weighted moving average algorithm to filter out high-frequency measurement noise, thereby obtaining the predicted value of the one-way transmission delay at the current moment.

[0038] The instantaneous throughput is calculated by counting the amount of data successfully sent and received within a unit of time; the instantaneous throughput is smoothed using an exponentially weighted moving average algorithm to reflect the long-term capacity trend of the link, thereby obtaining an estimate of the available bandwidth.

[0039] The difference between the instantaneous one-way delay and the predicted one-way transmission delay of the previous moment is calculated to obtain the delay deviation; the square of the delay deviation is processed by exponential weighted moving average to obtain the delay jitter variance that reflects the stability of the link.

[0040] By parsing the acknowledgment characters or selective acknowledgment message header information returned by the receiving end, the reassembly buffer usage and maximum reassembly buffer capacity of the receiving end can be extracted. The reassembly buffer usage refers to the total number of bytes of data packets that have been successfully received in the receiving end's reassembly buffer but cannot be submitted to the application layer temporarily due to discontinuous sequence numbers.

[0041] Calculate the weighted average transmission delay at the current moment based on the predicted one-way transmission delay values ​​for all active links:

[0042]

[0043] in, Indicates the first The weighted average transmission delay of all active links at any given time; Indicates the current sampling time; This is the set of currently available links; For link index; For link In the The estimated available bandwidth at any given time; For link In the The predicted one-way transmission delay at any given time.

[0044] It should be noted that in multi-link aggregation transmission, the total throughput of the system is mainly contributed by high-bandwidth links. This means that the vast majority of data packets are transmitted through high-bandwidth links. Therefore, the latency characteristics of high-bandwidth links objectively represent the mainstream rhythm of data flow arriving at the receiving end. If an arithmetic average is simply used, the extreme latency values ​​of low-bandwidth links will unreasonably skew the benchmark value, causing the algorithm to misjudge the overall state of the system. Therefore, this invention uses the bandwidth ratio of each link as a weighting coefficient to perform a weighted summation of the latency of each link, giving high-bandwidth links greater weight. This allows the resulting average latency to more accurately reflect the actual arrival time of most data packets. Using this as a benchmark for subsequent deviation calculations can effectively avoid the fluctuations of edge links interfering with the system's judgment of the overall transmission order.

[0045] S2. Determine the out-of-order blocking potential of each physical link based on the predicted one-way transmission delay, the weighted average transmission delay, and the reassembly buffer usage at the receiving end.

[0046] It should be noted that traditional transmission strategies fail to measure the interference intensity caused by link delay differences on system orderliness. When the delay of a link deviates significantly from the average level, it leads to the discretization of the arrival time of data packets at the receiver, thereby exacerbating the reassembly pressure at the receiver. If data continues to be sent to links with large deviations when the receiver's buffer is about to saturate, it will directly cause overflow and head-of-line blocking. Therefore, this invention introduces the concept of potential energy, nonlinearly coupling the link delay deviation with the receiver's buffer load rate to measure the blocking cost that choosing a certain link at the current moment may cause to the receiver.

[0047] Specifically, the out-of-order blocking potential of each link satisfies the expression:

[0048]

[0049] in, Indicates link In the Disordered blocking potential energy at any given moment; Indicates link In the The predicted one-way transmission delay at any given time; Indicates the first The weighted average transmission delay of all active links at any given time; Indicates link The delay deviation is the larger the value, the greater the difference between the transmission rhythm of the link and the overall system rhythm, which leads to a higher probability of sequence number holes at the receiving end. Indicates the first The minimum transmission delay among all available links at any given time, used to eliminate the dimension of delay deviation; To prevent the use of tiny constants with a denominator of zero, in this embodiment, the value is taken as... Second; Indicates the first The amount of reassembly buffer usage at the receiving end at any given time; This indicates the maximum capacity of the reassembly buffer at the receiving end; This indicates the current buffer load rate at the receiving end. The larger the value, the weaker the receiving end's ability to tolerate out-of-order data. Represents an exponential function with the natural constant as its base; The potential energy sensitivity coefficient is used to adjust the weight of the impact of cache load rate on out-of-order blocking potential energy. The empirical value is 3.0. In other embodiments, implementers can set the potential energy sensitivity coefficient according to the actual implementation situation.

[0050] When the receiver's buffer load rate At lower levels, The value is relatively small, and the out-of-order blocking potential energy mainly depends on the latency deviation of the link. When the buffer load rate at the receiving end increases, The value of increases dramatically, amplifying the contribution of time delay deviation to out-of-order blocking potential energy, so that any tiny time deviation will be transformed into huge potential energy resistance.

[0051] For example, Figure 2 This is a schematic diagram illustrating the dynamic coupling relationship between out-of-order blocking potential and reorganization cache occupancy in this invention. When the reorganization cache occupancy increases slightly, the out-of-order blocking potential will respond quickly and fluctuate significantly. This high out-of-order blocking potential is transformed into a huge resistance, which in turn inhibits the link selection that leads to out-of-order, thereby pushing the reorganization cache occupancy back to a low level.

[0052] S3. Determine the corrected heuristic function value for each physical link based on the estimated available bandwidth, predicted one-way transmission delay, and out-of-order blocking potential.

[0053] It should be noted that existing swarm intelligence algorithms typically employ greedy heuristics, which tend to overuse low-latency links, leading to packet congestion at the receiver. Therefore, this invention uses a damped suppression model, which uses out-of-order blocking potential energy to modify the heuristic function. When the potential energy increases, the heuristic value decreases inversely, thereby achieving natural suppression of high-risk links without introducing additional adjustment factors. This ensures transmission performance while forcibly reducing the attractiveness of high-blocking-risk links.

[0054] Specifically, the correction heuristic function for each link satisfies the expression:

[0055]

[0056] in, Indicates link In the The value of the heuristic function at each moment is used as an evaluation index for relative merit. Indicates link In the The estimated available bandwidth at any given time; Indicates link In the The predicted one-way transmission delay at any given time; This represents the transmission capacity of a link per unit of latency, i.e., the basic performance score. Indicates link In the Disordered blocking potential energy at any given moment; The damping coefficient is based on potential energy, and its value ranges from [value range missing]. .

[0057] When disordered blocking potential energy When the value approaches 0, the damping coefficient approaches 1, and the modified heuristic function value reverts to the basic performance score of the link, with the algorithm prioritizing performance; when the out-of-order blocking potential energy As the denominator increases, the value of the corrected heuristic function is affected. The probability of this link being selected is significantly reduced.

[0058] S4. Determine the state transition probability of each physical link based on the pheromone concentration of each physical link and the value of the modified heuristic function, and select the target link for data packet transmission based on the state transition probability.

[0059] It should be noted that this invention uses the Ant Colony Optimization (ACO) algorithm for path decision-making, treating the data packets to be sent as ants and each physical link as a path. It calculates the state transition probability of each link by combining the pheromone concentration reflecting the long-term performance of the link with the value of the correction heuristic function, and distributes data packets according to the state transition probability. At the same time, it updates the pheromone according to the transmission results to achieve adaptive iteration of the strategy.

[0060] Specifically, before transmission begins, the pheromone concentration of each physical link is initialized to the same constant. This ensures that each link has an equal chance of being selected in the initial stage.

[0061] Furthermore, the state transition probabilities of each link satisfy the expression:

[0062]

[0063] in, Indicates the first Select link at any time The probability of transmission; Indicates link In the The pheromone concentration at any given moment reflects the historical transmission quality of the link. Indicates link In the The value of the heuristic function is adjusted at any given time. This is the set of currently available links; Link index in the set of available links; This is a pheromone importance factor used to adjust the weight of historical experience on decision-making. The empirical value is set to 1.0. If this value is too large, the algorithm is prone to stalling. This is a heuristic information importance factor used to adjust the weight of the current link state on the decision. The empirical value is set to 2.0. If this value is too large, the algorithm is prone to degenerate into a greedy search. In other embodiments, implementers can adjust the pheromone importance factor and the heuristic information importance factor according to their emphasis on system stability and response speed.

[0064] The sending end selects the target link to send data packets based on the state transition probability using a roulette wheel method. Upon receiving confirmation from the receiving end, it updates the pheromone concentration of the target link according to whether the transmission caused out-of-order delivery. Specifically:

[0065] If the data packets arrive in order, the expression for updating the pheromone concentration is:

[0066]

[0067] If data packets are out of order, the expression for updating pheromone concentration is:

[0068]

[0069] in, Indicates the updated number Time Link pheromone concentration; This is the pheromone evaporation coefficient, used to simulate the process of pheromone dissipating over time in nature, preventing the algorithm from getting trapped in a local optimum due to the infinite accumulation of pheromones. The range of values ​​is The empirical value is 0.1; Indicates the first Time Link pheromone concentration; The pheromone update step size is used to reward or punish transmission behavior, with an empirical value set to 0.1. This indicates the amount of pheromones remaining. This indicates a positive incentive for sequential arrival behavior, increasing the likelihood that the link will be selected again; This indicates a negative suppression of out-of-order behavior, reducing the likelihood of the link being selected again. If the updated... If the value falls below the preset minimum, it will be reset to the minimum value to maintain the algorithm's exploratory capabilities.

[0070] Through the iterative updates based on the improved ant colony algorithm, the system can continuously learn and optimize the link selection strategy, eventually converging to a set of optimal transmission schemes that maximize bandwidth utilization while keeping out-of-order delivery within the receiver's buffer tolerance range, thus achieving multi-link transmission strategy optimization based on swarm intelligence.

[0071] For example, Figure 3 This is a schematic diagram comparing the receiver reassembly buffer usage of the present invention and existing technologies. Figure 3 The gray dashed line represents the existing technology using a greedy algorithm, whose reassembly buffer usage fluctuates wildly in a sawtooth pattern, with peak values ​​frequently exceeding high levels, indicating severe head-of-line blocking. The red solid line represents the present invention, whose reassembly buffer usage, although occasionally fluctuating momentarily, quickly falls back and remains at a low level, indicating that out-of-order blocking is effectively suppressed and the arrival rhythm of data packets is highly consistent.

[0072] Figure 4 This is a schematic diagram comparing the distribution of link selection strategies between the present invention and existing technologies. Figure 4 The medium gray bars show that existing technologies tend to use a large number of low-latency links (Link 0) and high-bandwidth links (Link 1) simultaneously, resulting in a mix of fast and slow data packets; the blue bars show that this invention automatically identifies the high-bandwidth backbone link (Link 1) and makes adaptive adjustments for overall order, significantly reducing the number of times other links that may disrupt the transmission pattern are selected.

[0073] Figure 5 This is a schematic diagram comparing the end-to-end delay distribution of data packets between the present invention and existing technologies. Figure 5 Existing technologies exhibit a discrete bimodal distribution, meaning that data packets arrive at different times and exhibit a significant long-tail effect. This invention presents a highly concentrated unimodal distribution, indicating that data packets from each link arrive synchronously, greatly reducing the waiting time at the receiving end.

[0074] This invention also discloses a multi-link transmission strategy optimization system based on swarm intelligence, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the multi-link transmission strategy optimization method based on swarm intelligence according to this invention.

[0075] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

Claims

1. A multi-link transmission strategy optimization method based on swarm intelligence, characterized in that, include: The instantaneous one-way delay, instantaneous throughput, and reassembly buffer usage at the receiving end of each physical link are collected. The collected instantaneous one-way delay and instantaneous throughput are smoothed to obtain the predicted value of one-way transmission delay and the estimated value of available bandwidth. The weighted average transmission delay of all active links is calculated. Based on the predicted one-way transmission delay, the weighted average transmission delay, and the reassembly buffer occupancy at the receiver, the out-of-order blocking potential of each physical link is determined, satisfying the following: ； , Indicates link In the Out-of-order blocking potential energy at any given moment, and predicted one-way transmission delay; Indicates the first The weighted average transmission delay of all active links at any given time; Indicates the first The minimum transmission delay among all available links at any given time; To prevent tiny constants with a denominator of zero; Indicates the first The amount of reassembly buffer usage at the receiving end at any given time; This indicates the maximum capacity of the reassembly buffer at the receiving end; Represents an exponential function with the natural constant as its base; The potential energy sensitivity coefficient; Based on the available bandwidth estimate, the predicted one-way transmission delay, and the out-of-order blocking potential, determine the corrected heuristic function value for each physical link, satisfying: ； , , Indicates link In the The corrected heuristic function value at time, the estimated available bandwidth value, and the predicted one-way transmission delay value; Based on the pheromone concentration of each physical link and the value of the modified heuristic function, the state transition probability of each physical link is determined, and the target link is selected for data packet transmission according to the state transition probability.

2. The multi-link transmission strategy optimization method based on swarm intelligence according to claim 1, characterized in that, The calculation of the weighted average transmission delay of all active links includes: The weighted average transmission delay is obtained by using the proportion of the available bandwidth estimate of each physical link in the total available bandwidth as a weight and then summing the predicted one-way transmission delay values ​​of each physical link.

3. The multi-link transmission strategy optimization method based on swarm intelligence according to claim 1, characterized in that, The state transition probability satisfies the expression: ； in, Indicates the first Select link at any time The probability of transmission; Indicates link In the The pheromone concentration at any given time; Indicates link In the The value of the heuristic function is adjusted at any given time. This refers to the set of currently available links. Link index in the set of available links; For pheromone importance factors; This is a heuristic information importance factor.

4. The multi-link transmission strategy optimization method based on swarm intelligence according to claim 1, characterized in that, The step of selecting a target link for data packet transmission based on state transition probability includes: The sending end selects the target link to send data packets based on the state transition probability using a roulette wheel betting method.

5. The multi-link transmission strategy optimization method based on swarm intelligence according to claim 1, characterized in that, The process of smoothing the collected instantaneous one-way delay and instantaneous throughput to obtain the predicted one-way transmission delay and the estimated available bandwidth includes: The instantaneous one-way delay sequence is smoothed using an exponentially weighted moving average algorithm to obtain the predicted one-way transmission delay value at the current moment; the instantaneous throughput is smoothed using an exponentially weighted moving average algorithm to obtain the estimated available bandwidth value.

6. The multi-link transmission strategy optimization method based on swarm intelligence according to claim 1, characterized in that, Also includes: The pheromone concentration of the target link is updated based on the data packet transmission results.

7. The multi-link transmission strategy optimization method based on swarm intelligence according to claim 6, characterized in that, Also includes: In response to the data packets arriving at the receiving end in order, the pheromone concentration of the target link is updated using a positive incentive strategy. The updated pheromone concentration satisfies the expression: ; In response to the out-of-order delivery of the data packets at the receiving end, the pheromone concentration of the target link is updated using a negative suppression strategy. The updated pheromone concentration satisfies the expression: ;in, Indicates the updated number Time Link pheromone concentration; The pheromone evaporation coefficient; Indicates the first Time Link pheromone concentration; Update the step size for pheromones.

8. A multi-link transmission strategy optimization system based on swarm intelligence, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement the multi-link transmission strategy optimization method based on swarm intelligence according to any one of claims 1-7.

Images