CN115278811A - MPTCP connection path selection method based on decision tree model - Google Patents
MPTCP connection path selection method based on decision tree model Download PDFInfo
- Publication number
- CN115278811A CN115278811A CN202210899601.5A CN202210899601A CN115278811A CN 115278811 A CN115278811 A CN 115278811A CN 202210899601 A CN202210899601 A CN 202210899601A CN 115278811 A CN115278811 A CN 115278811A
- Authority
- CN
- China
- Prior art keywords
- path
- network
- data
- decision tree
- tree model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000003066 decision tree Methods 0.000 title claims abstract description 82
- 238000010187 selection method Methods 0.000 title claims description 6
- 230000005540 biological transmission Effects 0.000 claims abstract description 57
- 238000012549 training Methods 0.000 claims abstract description 40
- 238000000034 method Methods 0.000 claims abstract description 29
- 230000006854 communication Effects 0.000 claims description 10
- 238000004891 communication Methods 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 4
- 230000009286 beneficial effect Effects 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 238000010276 construction Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 230000001934 delay Effects 0.000 description 2
- 230000003111 delayed effect Effects 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000015556 catabolic process Effects 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W40/00—Communication routing or communication path finding
- H04W40/02—Communication route or path selection, e.g. power-based or shortest path routing
- H04W40/12—Communication route or path selection, e.g. power-based or shortest path routing based on transmission quality or channel quality
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W40/00—Communication routing or communication path finding
- H04W40/02—Communication route or path selection, e.g. power-based or shortest path routing
- H04W40/12—Communication route or path selection, e.g. power-based or shortest path routing based on transmission quality or channel quality
- H04W40/125—Communication route or path selection, e.g. power-based or shortest path routing based on transmission quality or channel quality using a measured number of retransmissions as a link metric
Abstract
The embodiment of the invention provides a decision tree model generation method for determining a preferred path in MPTCP connection, which comprises the following steps: acquiring a training set consisting of a plurality of training samples, wherein each training sample comprises network parameter characteristics and a label, the network parameter characteristics comprise MAC layer network parameters and transmission layer network parameters corresponding to at least two network paths in MPTCP connection established based on a multi-path transmission control protocol, and the label indicates a preferred path in the at least two network paths under the corresponding network parameter characteristics; generating a decision tree model by utilizing a preset decision tree learning rule based on the training set so as to make the decision tree model make a decision according to the input network parameter characteristics to output a result of a preferred path in at least two network paths corresponding to the MPTCP connection; the invention can reduce the prediction lag and improve the accuracy of the result of the predicted optimal path.
Description
Technical Field
The invention relates to the field of wireless communication, in particular to the technical field of data transmission, and more particularly relates to a decision tree model-based MPTCP connection path selection method.
Background
The rapid development of wireless networks has greatly changed the production and life of humans. Next generation mobile networks are expected to present a large number of vertical domain use cases, such as: large-scale internet of things (IoT), remote machinery, autonomous driving, and Virtual Reality (VR). However, wireless networks are more variable and unstable than wired transmissions due to environmental constraints. Especially in extreme cases, a mobile application relying on a single path interface suffers from a significant performance degradation. Therefore, some researchers have designed a multi-path scheme (such as multi path Transmission Control Protocol, MPTCP for short) in an attempt to compensate for this performance gap by integrating multiple radio access technologies.
In consideration of the case of low-speed movement, a good MPTCP path selector should actively detect the wireless environment and dynamically return to the best quality network path to transmit the incoming packets of the application layer. This will result in frequent network quality handovers as the mobile device may span multiple network coverage areas in a short time. By selecting a suitable network path, the transmit buffer of MPTCP can significantly reduce the (In-flight) packet queues In the transmission of the sub-streams with poor quality, thereby improving the overall performance of the MPTCP connection. Generally, the MPTCP path selector determines the priority of each network path based only on network parameters in the transport layer (TCP layer); for example, the priority of the network path is determined by a congestion window value or a minimum round trip time (MinRTT). However, when the network path quality is predicted according to the network parameters of the transport layer and the network path is switched, a large switching delay often exists, data transmission congestion is easily caused due to the delay of the network path quality prediction, and the data transmission efficiency is affected.
Disclosure of Invention
Therefore, an object of the present invention is to overcome the above-mentioned drawbacks of the prior art and to provide a method for selecting an MPTCP connection path based on a decision tree model.
The purpose of the invention is realized by the following technical scheme:
according to a first aspect of the present invention, there is provided a decision tree model generation method for determining a preferred path in an MPTCP connection, comprising: acquiring a training set consisting of a plurality of training samples, wherein each training sample comprises network parameter characteristics and a label, the network parameter characteristics comprise MAC layer network parameters and transmission layer network parameters corresponding to at least two network paths in MPTCP connection established based on a multi-path transmission control protocol, and the label indicates a preferred path in the at least two network paths under the corresponding network parameter characteristics; and generating a decision tree model by utilizing a preset decision tree learning rule based on the training set so that the decision tree model makes a decision according to the input network parameter characteristics to output a result of a preferred path in at least two network paths corresponding to the MPTCP connection.
In some embodiments of the invention, the MAC layer network parameters include a signal strength and a signal-to-noise ratio of each of the at least two network paths corresponding to the MPTCP connection.
In some embodiments of the invention, the at least two network paths include a mobile network path and a WiFi path, wherein the MAC layer network parameters further include: the reference signal received power and reference signal received quality of the mobile network path, and the transmit data rate and receive data rate of the WiFi path.
In some embodiments of the invention, the transport layer network parameters include: a round trip delay, a congestion window value, a packet delivery rate, and a packet loss rate, or a combination thereof, for each of at least two network paths corresponding to an MPTCP connection.
In some embodiments of the invention, the method comprises: when two devices establishing the MPTCP connection communicate, collecting a network path adopted by data transmission and corresponding network parameter characteristics to form initial acquisition data; when the deviation of each corresponding parameter in the network parameter characteristics of the two pieces of initial acquired data is smaller than or equal to a preset threshold value and different network paths are adopted, determining an optimal path according to the data transmission quality of the network paths corresponding to the two pieces of initial acquired data; and adding labels to the two initial acquisition data according to the determined preferred path to form a sample, and storing the sample into a data set, wherein the training set is a subset of the data set.
In some embodiments of the invention, the quality of data transmission comprises the goodput rate and/or application latency of the application sending the data.
According to a second aspect of the present invention, there is provided a decision tree model-based MPTCP connection path selection method for selecting a path for transmitting data in an MPTCP connection between a first device and a second device, the method including: acquiring network parameter characteristics corresponding to MPTCP connection established between first equipment and second equipment based on a multi-path transmission control protocol as characteristics to be predicted, wherein the MPTCP connection comprises at least two network paths; inputting the feature to be predicted into a decision tree model obtained by a decision tree model generation method which is deployed in first equipment and used for determining a preferred path in the MPTCP connection according to the first aspect, and outputting a result of the preferred path; and determining the priority of each path in at least two network paths according to the result of the preferred path, and determining the transmitted path according to the priority of each path when the first equipment transmits data to the second equipment.
In some embodiments of the present invention, the decision tree model generation method for determining a preferred path in an MPTCP connection further includes: when an MPTCP connection is established between first equipment and second equipment for communication, collecting a network path adopted by the first equipment for sending data and corresponding network parameter characteristics to form operation collected data; when the deviation of each corresponding parameter in the network parameter characteristics of the two operation collected data is smaller than or equal to a preset threshold value and different network paths are adopted, determining an optimal path according to the data transmission quality of the network paths corresponding to the two operation collected data; adding labels to the two running collected data according to the determined preferred path to form a newly added sample, and storing the newly added sample into a data set to obtain an updated data set; a training set is obtained from the updated data set to regenerate the decision tree model, and the decision tree model deployed on the first device is updated with the regenerated decision tree model.
According to a third aspect of the present invention, there is provided an electronic apparatus comprising: one or more processors; and a memory, wherein the memory is to store executable instructions; the one or more processors are configured to implement the steps of the method of the first aspect and/or the second aspect via execution of the executable instructions.
Compared with the prior art, the invention has the advantages that:
the invention improves the prior art, and trains a decision tree model to make a better optimal path decision by introducing MAC layer network parameters and combining cross-layer network parameters (MAC layer network parameters and transmission layer network parameters). The quality change of a wireless link is sensed in time by using cross-layer parameters (namely parameters of an MAC layer and a transmission layer) of network transmission, and a path selection strategy is updated, so that the data packet is prevented from being blocked in the network transmission, and the transmission performance of the MPTCP is improved; in addition, mutual dependency relations may exist among parameters in the network parameter characteristics, linear or nonlinear relations may also exist between the parameters and the labels in the network parameter characteristics, and the decision tree model does not need to assume independence among the characteristics, so the method selects the decision tree model to predict the optimal path, avoids the influence of the linear and nonlinear coupling relations among different parameters on prediction precision, and improves the accuracy of the result of the predicted optimal path. By improving these two points, the prediction lag can be reduced, and the prediction accuracy can be improved.
Drawings
Embodiments of the invention are further described below with reference to the accompanying drawings, in which:
fig. 1 is a diagram for explaining the hysteresis for predicting the quality of a network path and switching the network path based on the minimum round trip delay of a transport layer according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of an application scenario according to an embodiment of the present invention;
FIG. 3 is a flowchart illustrating a method for selecting an MPTCP connection path based on a decision tree model according to an embodiment of the present invention;
fig. 4 is a schematic block diagram of a method for selecting an MPTCP connection path based on a decision tree model according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be further described in detail by embodiments with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the invention.
As mentioned in the background section, predicting the quality of a network path according to network parameters of a transport layer and switching the network path often has a large switching delay, which is likely to cause data transmission congestion due to the delay of network path quality prediction and affect the data transmission efficiency. Therefore, the invention obtains the network parameters of the MAC layer closer to the physical layer, constructs a training set, wherein each training sample comprises network parameter characteristics and labels, wherein the network parameter characteristics comprise the MAC layer network parameters corresponding to at least two network paths in the MPTCP connection established based on the multi-path transmission control protocol, the labels indicate the preferred path of the at least two network paths under the corresponding network parameter characteristics, and predicts the result of the preferred path by means of the MAC layer network parameters in a mode of training a decision tree model and selects the network path (some places are called paths for short) in the MPTCP connection; therefore, the probability of data transmission congestion caused by prediction lag is reduced, and the data transmission efficiency is improved.
For a better understanding of the present invention, the significance of the MAC parameters for MPTCP connection path selection is illustrated by the accompanying drawings. As shown in fig. 1, it is visually demonstrated that the network path quality is predicted based on the minimum round trip delay (MinRTT) of the transport layer and the hysteresis of the network path switching is shown to illustrate the reason for selecting the MAC layer network parameters according to the present invention. In the present invention, schematically, the MPTCP connection includes a mobile network path (taking an LTE path as an example) and a WiFi path, fig. 1a shows a variation curve of Received Signal Strength (RSSI) in MAC layer network parameters corresponding to the LTE path and the WiFi path in each measurement time slot, and fig. 1b shows a variation curve of signal-to-noise ratio (SINR) in MAC layer network parameters corresponding to the LTE path and the WiFi path in each measurement time slot. In fig. 1a and 1b, the bold vertical line is to determine the switching point according to the minimum round trip delay (MinRTT) of the transmission layer, and the circle is an ideal switching point, and it can be seen that the MinRTT switching point always misses the ideal switching point observed according to the physical layer network parameters. This is due to the delayed acknowledgement mechanism of the transport layer (TCP layer), the RTT-based path selector must wait at least one RTT for a corrupted TCP sub-flow to switch. In addition, as can be seen from fig. 1c and 1d, the delayed switching results, and as the MAC layer attribute in fig. 1a and 1b changes, the poor quality path in fig. 1c and 1d accumulates more data packets in transmission than the good quality path, thereby causing data transmission congestion. Therefore, the invention improves the prior art, and trains the decision tree model to make a better optimal path decision by introducing MAC layer network parameters and combining cross-layer network parameters (MAC layer network parameters and transmission layer network parameters). The method comprises the steps that the quality change of a wireless link is sensed in time by using cross-layer parameters (namely parameters of an MAC layer and a transmission layer) of network transmission, and a path selection strategy is updated, so that the data packet is prevented from being blocked in the network transmission, and the transmission performance of the MPTCP is improved; in addition, mutual dependency relations may exist among parameters in the network parameter characteristics, linear or nonlinear relations may also exist between the parameters and the labels in the network parameter characteristics, and the decision tree model does not need to make assumptions on the independence among the characteristics. By improving these two points, the prediction lag can be reduced and the prediction accuracy can be improved.
The invention is explained in three aspects of sample construction, model training and application scenarios.
1. Construction of the sample
In order to reduce prediction lag and improve prediction accuracy, the method trains a decision tree model by means of a sample of MAC layer network parameter construction closer to a physical layer to predict a preferred path. Meanwhile, although the path prediction priority by using the network parameters of the transport layer only lags behind the path prediction priority according to the network parameters of the MAC layer, the transport layer also contains abundant information, which is beneficial to determining the path priority to a certain extent. According to one embodiment of the invention, a data set is firstly constructed, wherein the data set comprises a plurality of samples, each sample comprises network parameter characteristics and a label, the network parameter characteristics comprise MAC layer network parameters corresponding to at least two network paths in the MPTCP connection established based on the multi-path transmission control protocol and transport layer network parameters, and the label indicates a preferred path in the at least two network paths under the corresponding network parameter characteristics. The technical scheme of the embodiment can at least realize the following beneficial technical effects: the embodiment has the advantages that the quality change of the wireless link can be sensed in time by using the cross-layer parameters (namely the parameters of the MAC layer and the transmission layer) of the network transmission, the path selection strategy is updated, the data packet is prevented from being blocked in the network transmission, and the transmission performance of the MPTCP is improved. According to an embodiment of the present invention, assuming that there are two paths in the MPTCP connection, the value of the label corresponding to path 0 is set to 0, and the value of the label corresponding to path 1 is set to 1, if it is better to transmit data by using path 1 under the network parameter characteristics of a training sample, the value of the label in the training sample is 1. Similarly, if there are more paths in the MPTCP connection in the actual application scenario, the option of increasing the value of more labels may be added.
For various network paths, some common parameters of the MAC layer are corresponding, and the common parameters can be obtained to construct samples. According to an embodiment of the present invention, the MAC layer network parameters include a signal strength and a signal-to-noise ratio of each of at least two network paths corresponding to the MPTCP connection. The technical scheme of the embodiment can at least realize the following beneficial technical effects: the network parameter characteristics are simply and efficiently constructed according to the signal strength and the signal-to-noise ratio corresponding to the MAC layer, and the trained decision tree model can more accurately and quickly acquire the priority of the path according to the signal strength and the signal-to-noise ratio.
For different network paths, some specific parameters of the MAC layer may be included, and further, besides general parameters of the MAC layer, it may also be considered to add these specific parameters of the MAC layer to the constructed network parameter characteristics, so as to avoid that the accuracy is not high due to less information referred by the decision tree model. According to an embodiment of the present invention, the at least two network paths are wireless network paths, including a mobile network path and a WiFi path, wherein the MAC layer network parameters further include: the reference signal received power and reference signal received quality of the mobile network path, and the transmit data rate and receive data rate of the WiFi path. The explanation of the corresponding MAC layer network parameters of this embodiment is shown in table 1:
TABLE 1
According to one embodiment of the invention, the transport layer network parameters include: round Trip Time (RTT), a Congestion window value (CWnd value), a Packet Delivery Rate (PDR), and a Packet Loss Rate (PLR), or a combination thereof, of each of at least two network paths corresponding to an MPTCP connection. The PDR is a ratio of the number of data packets received by the data receiver to the number of data packets sent by the application layer of the data sender in a certain period. PLR refers to the ratio of the number of packets lost in a transmission to the number of packets sent.
For the way of obtaining the data set, the communication may be simulated and the data may be measured to construct the sample in a simulation way, or the communication may be simulated and the data may be measured to construct the sample in an actual scene, which is not limited in the present invention. According to one embodiment of the invention, when two devices establishing MPTCP connection communicate, a network path adopted by data transmission and corresponding network parameter characteristics are collected to form initial acquisition data; when the deviation of each corresponding parameter in the network parameter characteristics of the two pieces of initial acquisition data is smaller than or equal to a preset threshold value and different network paths are adopted, determining a preferred path according to the data transmission quality of the network paths corresponding to the two pieces of initial acquisition data; and adding labels to the two initial acquisition data according to the determined preferred path to form a sample, and storing the sample into a data set, wherein the training set is a subset of the data set. According to an embodiment of the present invention, the data transmission quality is determined according to an application goodput (abbreviated ag) and an application delay (abbreviated ad) of an application that transmits data. Determining a preferred path by comparing the goodput rate with the magnitude of the value of the application delay; among the two network paths, the network path with higher effective throughput rate is more optimal; network paths with smaller application delays are preferred for the same goodput. Because the values of ag and ad are required to be recorded when the application program transmits data, the data transmission quality can be determined simply and efficiently by adopting the values of ag and ad, and the additional overhead of recording is avoided. It should be understood that other ways to determine a more optimal network path may be used, and the present invention is not limited in this respect, such as: comparing the application delays and then the goodput rates; for another example: and scoring the quality of the network paths according to the effective throughput rate and the application program delay in the two network paths, and determining the preferred path according to the path quality scores obtained by scoring. To illustrate an exemplary manner of constructing a sample, the inventor selects the latter manner to measure and construct a sample, but because wireless communication has many influencing factors, it is difficult to keep the network parameters measured when data is transmitted in different paths consistent, so that a label of a sample can be determined based on two initial collected data which are different in adopted network paths but similar in network parameter characteristics, which is described below with reference to table 2:
TABLE 2
In the exemplary sample given in table 2, the mobile network path is the LTE path and the other path is the WiFi path. When constructing the sample, the sample may be constructed by using parameters measured when the first device and the second device perform communication, and in order to compare data transmission quality of different paths, a sample having similar network parameter characteristics (a deviation of each corresponding parameter in the network parameter characteristics of the two pieces of initially acquired data is less than or equal to a preset threshold, each parameter may be independently set with the preset threshold, or one preset threshold is uniformly set for all the parameters, and the preset threshold may be determined according to needs of an implementer, it should be understood that a value of the preset threshold is set to be smaller, a label of the sample is more accurate, but difficulty in obtaining the sample is increased, and therefore, it needs to be set according to actual needs of the implementer, which is not limited by the present invention)) may be constructed by using two pieces of initially acquired data having different network paths for transmitting data (it should be understood that paired associated samples are only used for comparing advantages and disadvantages of paths to determine a label value of the pair of samples, and each sample is separate and is not input in pair during training). For example, row 3 and row 4 in table 2 represent a pair of associated samples, where the ag value in row 4 is greater, which indicates that the path taken is a WiFi path, and is more preferable, the labels corresponding to the samples in row 3 and row 4 are both set to 1 (indicating that the preferred path is a WiFi path); the 5 th row and the 6 th row represent a pair of associated samples, wherein the ag value of the 6 th row is larger, which indicates that the adopted path is better when being the LTE path, and the labels corresponding to the samples of the 5 th row and the 6 th row are both set to 0 (which indicates that the preferred path is the LTE path).
2. Model training
Using the data set constructed in any of the above embodiments, the model can be trained to predict the outcome of the preferred path. It should be understood that, in the field, when training a model, in order to evaluate the accuracy of the model, all samples in a data set are not used for training, but are divided into multiple seed sets according to a predetermined ratio, such as dividing the data set into a training set and a test set, wherein the data set is trained by the samples in the training set (corresponding to the training samples), and the test set is used for evaluating the accuracy of the model after training. According to an embodiment of the present invention, there is provided a method for generating a decision tree model, including: acquiring a training set consisting of a plurality of training samples; and generating a decision tree model by utilizing a preset decision tree learning rule based on the training set so that the decision tree model can make a decision according to the input network parameter characteristics and output a result of a preferred path in at least two network paths corresponding to the MPTCP connection. According to an embodiment of the present invention, the predetermined decision tree learning rule adopts any one of the learning rules ID3, C4.5, CART. According to one embodiment of the invention, 10-fold cross-validation is used to evaluate the accuracy of the decision tree model, i.e. the whole classification dataset is randomly divided into two parts: wherein 90% of the data is used as a training set for training the decision tree model, and the remaining 10% of the data is used as a test set for evaluation of the decision tree model. The technical scheme of the embodiment can at least realize the following beneficial technical effects: considering the complex relationship between the network parameter characteristics and the performance metrics of each network path (TCP subflow) in the MPTCP connection, a Machine Learning algorithm (ML algorithm, such as a neural network model, a decision tree model, etc.) is selected as a tool for solving the above problems. However, the selection of a proper ML algorithm is crucial to the accuracy of the selection of the predicted path, mutual dependency relations may exist among parameters in the network parameter characteristics in the data set, linear or nonlinear relations may also exist between each parameter and the label in the network parameter characteristics, and the decision tree model does not need to make assumptions on the independence among the characteristics, so that the method selects the decision tree model to predict the preferred path, so as to avoid the influence of the linear and nonlinear coupling relations among different parameters on the prediction accuracy, and improve the accuracy of the result of the predicted preferred path.
3. Application scenarios
According to the decision tree model obtained in the foregoing embodiment, referring to fig. 2, assuming that the mobile device establishes an MPTCP connection including a WiFi path and an LTE path, the decision tree model may be used to predict a preferred path result according to a network parameter characteristic (for example, related parameters of a network protocol stack, including MAC layer network parameters and/or transport layer network parameters included in the network parameter characteristic, are acquired through an operating system interface of the mobile device; for example, basic parameters of a physical interface are monitored through an application program interface API provided by an Android system in fig. 2 to serve as network parameters of MAC layers and/or transport layers of the WiFi and LTE paths), so as to determine a path for sending data from the MPTCP connection, so as to reduce a probability of data congestion. According to an embodiment of the present invention, referring to fig. 3, a method for selecting an MPTCP connection path based on a decision tree model is provided, for selecting a path for sending data in an MPTCP connection between a first device and a second device, including: s1, acquiring network parameter characteristics (refer to the embodiment for acquiring the network parameter characteristics, which is not described herein) corresponding to an MPTCP connection established between first equipment and second equipment based on a multi-path transmission control protocol as characteristics to be predicted, wherein the MPTCP connection comprises at least two network paths; s2, inputting the feature to be predicted into a decision tree model which is deployed in first equipment and obtained according to the method for generating the decision tree model of the embodiment, and outputting a result of an optimal path; and S3, determining the priority of each path in the at least two network paths according to the result of the preferred path, and determining the path to be transmitted according to the priority of each path when the first equipment transmits data to the second equipment. For example, assume that there are two network paths: the LTE path (the value corresponding to the tag is 0) and the WiFi path (the value corresponding to the tag is 1), and the output result of the preferred path is 1 (WiFi path is preferred), the priority of the WiFi path is set to 0 (assuming that the smaller the value is, the higher the priority is), and the priority of the LTE path is set to 1 (for example, a path selector in MPTCP version 0.95 is modified to determine the priority of the path based on the result of the preferred path of the decision tree model of the present invention), and when the first device sends data to the second device, the path with the higher priority is preferentially used to send data, that is, the WiFi path is preferentially used to send data. If more than two network paths are adopted, if the decision tree model still outputs a label corresponding to a preferred path, the priority of the paths except the preferred path can be determined in a random mode. Alternatively, the generated decision tree model (e.g., using a random forest) comprises a plurality of decision trees, and the priority of each path is determined by the result of the predicted preferred path of all the decision trees. For example, it is assumed that the decision tree model includes 10 decision trees, and there are an LTE path (corresponding to a label with a value of 0), a WiFi path (corresponding to a label with a value of 1), and a bluetooth path (corresponding to a label with a value of 2), where 5 decision tree predicted preferred paths are 0,3 decision tree predicted preferred paths are 1, and 2 decision tree predicted preferred paths are 2; then, according to the result of the preferred path, the priority of the LTE path is set to 0, the priority of the wifi path is set to 1, and the priority of the bluetooth path is set to 2 (assuming that the smaller the value, the higher the priority).
In the communication process of the two devices, the relationship between the communication network parameter characteristics and the corresponding preferred path may be influenced by various factors, so that a sample update data set may be continuously collected to ensure the prediction accuracy of the model, and according to an embodiment of the present invention, the MPTCP connection path selection method based on the decision tree model further includes: when an MPTCP connection is established between a first device (which may be a mobile device, such as a mobile phone, a tablet computer, a smart watch, and the like) and a second device (which may be a mobile device or a non-mobile device, such as a server, for example) for communication, a network path and corresponding network parameter characteristics, which are adopted by the first device to send data, are collected to form operation collection data; when the deviation of each corresponding parameter in the network parameter characteristics of the two operation collected data is smaller than or equal to a preset threshold value and different network paths are adopted, determining an optimal path according to the data transmission quality of the network paths corresponding to the two operation collected data; adding labels to the two running acquired data according to the determined preferred path to form a newly added sample and storing the newly added sample into a data set to obtain an updated data set; a training set is obtained from the updated data set to regenerate the decision tree model, and the decision tree model deployed on the first device is updated with the regenerated decision tree model. Referring to fig. 4, in order to obtain priority on the mobile network path and WiFi path (two TCP sub-flows), two decision tree models are maintained. And the trained offline decision tree is used for outputting the result of the preferred path according to the measured network parameter characteristics. Meanwhile, new training samples are continuously collected on the first equipment, and an online decision tree model is constructed by using the new training samples. And periodically copying the on-line decision tree model to replace the original off-line decision tree model so as to adapt to the continuously changing wireless environment. Taking an example of an application 1 on a first device sending data to an application 2 on a second device, after a decision tree model is deployed, the relevant work scene can be divided into a data sending part and a decision tree model updating part, wherein,
the data transmission section includes:
step K1: an application program of first equipment transmits data to be transmitted to a transmission layer, and the data is waiting to be transmitted in a transmission cache;
step K2: predicting a result of a preferred path according to the generated offline decision tree model and the current network parameter characteristics, determining a preferred set of each path by a path selector of the first equipment according to the result of the preferred path, and preferentially pushing data to a path with higher priority for sending;
the decision tree updating part comprises:
step M1: acquiring related network parameters of an MAC layer and a transmission layer through a system network interface;
step M2: collecting operation collected data for forming a newly added sample to obtain an updated data set;
step M3: periodically, a training set is obtained from the updated data set to regenerate the online decision tree model.
Step M4: the offline decision tree model is periodically updated with the online decision tree model.
It should be noted that, although the steps are described in a specific order, the steps are not necessarily executed in the specific order, and in fact, some of the steps may be executed concurrently or even in a changed order as long as the required functions are achieved.
The present invention may be a system, method and/or computer program product. The computer program product may include a computer readable storage medium having computer readable program instructions embodied therewith for causing a processor to implement various aspects of the present invention.
The computer readable storage medium may be a tangible device that retains and stores instructions for use by an instruction execution device. The computer readable storage medium may include, for example, but is not limited to, an electronic memory device, a magnetic memory device, an optical memory device, an electromagnetic memory device, a semiconductor memory device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a Static Random Access Memory (SRAM), a portable compact disc read-only memory (CD-ROM), a Digital Versatile Disc (DVD), a memory stick, a floppy disk, a mechanical coding device, such as a punch card or an in-groove protruding structure with instructions stored thereon, and any suitable combination of the foregoing.
Having described embodiments of the present invention, the foregoing description is intended to be exemplary, not exhaustive, and not limited to the embodiments disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The terminology used herein was chosen in order to best explain the principles of the embodiments, the practical application, or technical improvements to the market, or to enable others of ordinary skill in the art to understand the embodiments disclosed herein.
Claims (10)
1. A method for generating a decision tree model for determining a preferred path in an MPTCP connection, comprising:
acquiring a training set consisting of a plurality of training samples, wherein each training sample comprises network parameter characteristics and a label, the network parameter characteristics comprise MAC layer network parameters and transmission layer network parameters corresponding to at least two network paths in MPTCP connection established based on a multi-path transmission control protocol, and the label indicates a preferred path in the at least two network paths under the corresponding network parameter characteristics;
and generating a decision tree model by utilizing a preset decision tree learning rule based on the training set so that the decision tree model makes a decision according to the input network parameter characteristics to output a result of a preferred path in at least two network paths corresponding to the MPTCP connection.
2. The method of claim 1, wherein the MAC layer network parameters include a signal strength and a signal-to-noise ratio for each of the at least two network paths corresponding to the MPTCP connection.
3. The method of claim 2, wherein the at least two network paths comprise a mobile network path and a WiFi path, and wherein the MAC layer network parameters further comprise:
reference signal received power and reference signal received quality for a mobile network path, an
The transmit data rate and the receive data rate of the WiFi path.
4. The method of claim 3, wherein the transport layer network parameters comprise: round trip delay, congestion window value, packet delivery rate, and packet loss rate, or a combination thereof, for each of at least two network paths corresponding to an MPTCP connection.
5. The method according to any one of claims 1-4, characterized in that the method comprises:
when two devices establishing the MPTCP connection communicate, collecting a network path adopted by data transmission and corresponding network parameter characteristics to form initial acquisition data;
when the deviation of each corresponding parameter in the network parameter characteristics of the two pieces of initial acquisition data is smaller than or equal to a preset threshold value and different network paths are adopted, determining a preferred path according to the data transmission quality of the network paths corresponding to the two pieces of initial acquisition data;
and adding labels to the two initial acquisition data according to the determined preferred path to form a sample, and storing the sample into a data set, wherein the training set is a subset of the data set.
6. The method of claim 5, wherein the quality of data transmission comprises goodput and/or application latency of an application sending the data.
7. An MPTCP connection path selection method based on a decision tree model is used for realizing the selection of a path for sending data in an MPTCP connection between a first device and a second device, and is characterized by comprising the following steps:
acquiring network parameter characteristics corresponding to MPTCP connection established between first equipment and second equipment based on a multi-path transmission control protocol as characteristics to be predicted, wherein the MPTCP connection comprises at least two network paths;
inputting the feature to be predicted into a decision tree model obtained by a decision tree model generation method for determining a preferred path in an MPTCP connection according to any one of claims 1 to 7 deployed in a first device, and outputting a result of the preferred path;
and determining the priority of each path in at least two network paths according to the result of the preferred path, and determining the transmitted path according to the priority of each path when the first equipment transmits data to the second equipment.
8. The method of claim 7, further comprising:
when an MPTCP connection is established between first equipment and second equipment for communication, collecting a network path adopted by the first equipment for sending data and corresponding network parameter characteristics to form operation collected data;
when the deviation of each corresponding parameter in the network parameter characteristics of the two operation collected data is smaller than or equal to a preset threshold value and different network paths are adopted, determining an optimal path according to the data transmission quality of the network paths corresponding to the two operation collected data;
adding labels to the two running collected data according to the determined preferred path to form a newly added sample, and storing the newly added sample into a data set to obtain an updated data set;
a training set is obtained from the updated data set to regenerate the decision tree model, and the decision tree model deployed on the first device is updated with the regenerated decision tree model.
9. A computer-readable storage medium, characterized in that a computer program is stored thereon, which computer program is executable by a processor for implementing the steps of the method as claimed in any one of claims 1 to 8.
10. An electronic device, comprising:
one or more processors; and
a memory, wherein the memory is to store executable instructions;
the one or more processors are configured to implement the steps of the method of any one of claims 1-8 via execution of the executable instructions.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210899601.5A CN115278811A (en) | 2022-07-28 | 2022-07-28 | MPTCP connection path selection method based on decision tree model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210899601.5A CN115278811A (en) | 2022-07-28 | 2022-07-28 | MPTCP connection path selection method based on decision tree model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115278811A true CN115278811A (en) | 2022-11-01 |
Family
ID=83771642
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210899601.5A Pending CN115278811A (en) | 2022-07-28 | 2022-07-28 | MPTCP connection path selection method based on decision tree model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115278811A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116887338A (en) * | 2023-09-06 | 2023-10-13 | 南京欣网通信科技股份有限公司 | Big data-based 5G mobile network real-time adjustment method |
| CN117459188A (en) * | 2023-12-25 | 2024-01-26 | 吉林省吉能电力通信有限公司 | Electric Beidou communication system and communication method based on Beidou communication technology |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108809839A (en) * | 2018-07-17 | 2018-11-13 | 湖南理工学院 | Wireless Mesh backbone network network flow control methods and device |
| CN110546958A (en) * | 2017-05-18 | 2019-12-06 | 利弗有限公司 | Apparatus, system and method for wireless multilink vehicle communication |
| CN111131017A (en) * | 2019-11-19 | 2020-05-08 | 中国科学院计算技术研究所 | MPTCP (Multi-protocol Transmission control protocol) cross-layer optimization method and system based on multi-cellular wireless access gateway |
| CN112600759A (en) * | 2020-12-10 | 2021-04-02 | 东北大学 | Multipath traffic scheduling method and system based on deep reinforcement learning under Overlay network |
| CN113033709A (en) * | 2021-04-30 | 2021-06-25 | 中国工商银行股份有限公司 | Link prediction method and device |
| CN113038560A (en) * | 2021-04-13 | 2021-06-25 | 展讯通信（上海）有限公司 | Link switching prediction method and device and electronic equipment |
| US20210329528A1 (en) * | 2018-12-25 | 2021-10-21 | Huawei Technologies Co.,Ltd. | Connection establishment method and related device |
| CN114760644A (en) * | 2022-03-17 | 2022-07-15 | 华南理工大学 | Multilink transmission intelligent message scheduling method based on deep reinforcement learning |
| CN114785731A (en) * | 2022-04-25 | 2022-07-22 | 中国电信股份有限公司 | Model determination system and method, computer-readable storage medium |
| CN114792138A (en) * | 2021-01-26 | 2022-07-26 | 中国电信股份有限公司 | Machine learning model interpretation method, device and storage medium |
-
2022
- 2022-07-28 CN CN202210899601.5A patent/CN115278811A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110546958A (en) * | 2017-05-18 | 2019-12-06 | 利弗有限公司 | Apparatus, system and method for wireless multilink vehicle communication |
| CN108809839A (en) * | 2018-07-17 | 2018-11-13 | 湖南理工学院 | Wireless Mesh backbone network network flow control methods and device |
| US20210329528A1 (en) * | 2018-12-25 | 2021-10-21 | Huawei Technologies Co.,Ltd. | Connection establishment method and related device |
| CN111131017A (en) * | 2019-11-19 | 2020-05-08 | 中国科学院计算技术研究所 | MPTCP (Multi-protocol Transmission control protocol) cross-layer optimization method and system based on multi-cellular wireless access gateway |
| CN112600759A (en) * | 2020-12-10 | 2021-04-02 | 东北大学 | Multipath traffic scheduling method and system based on deep reinforcement learning under Overlay network |
| CN114792138A (en) * | 2021-01-26 | 2022-07-26 | 中国电信股份有限公司 | Machine learning model interpretation method, device and storage medium |
| CN113038560A (en) * | 2021-04-13 | 2021-06-25 | 展讯通信（上海）有限公司 | Link switching prediction method and device and electronic equipment |
| CN113033709A (en) * | 2021-04-30 | 2021-06-25 | 中国工商银行股份有限公司 | Link prediction method and device |
| CN114760644A (en) * | 2022-03-17 | 2022-07-15 | 华南理工大学 | Multilink transmission intelligent message scheduling method based on deep reinforcement learning |
| CN114785731A (en) * | 2022-04-25 | 2022-07-22 | 中国电信股份有限公司 | Model determination system and method, computer-readable storage medium |
Non-Patent Citations (1)
| Title |
|---|
| BINBIN LIAO等: "Cross-layer Path Selection in Multi-path TransportProtocol for Mobile Devices", 《ARXIV.ORG E-PRINT ARCHIVE》, 3 July 2020 (2020-07-03), pages 1 - 6 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116887338A (en) * | 2023-09-06 | 2023-10-13 | 南京欣网通信科技股份有限公司 | Big data-based 5G mobile network real-time adjustment method |
| CN116887338B (en) * | 2023-09-06 | 2023-12-08 | 南京欣网通信科技股份有限公司 | Big data-based 5G mobile network real-time adjustment method |
| CN117459188A (en) * | 2023-12-25 | 2024-01-26 | 吉林省吉能电力通信有限公司 | Electric Beidou communication system and communication method based on Beidou communication technology |
| CN117459188B (en) * | 2023-12-25 | 2024-04-05 | 吉林省吉能电力通信有限公司 | Electric Beidou communication system and communication method based on Beidou communication technology |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3143721B1 (en) | Probing technique for predictive routing in computer networks | |
| CN115278811A (en) | MPTCP connection path selection method based on decision tree model | |
| US9246822B2 (en) | Methods and apparatus for using a layered gear to analyze and manage real-time network quality of service transmission for mobile devices on public networks | |
| US11032219B2 (en) | Network aware application dependent adaptive protocol selection for IoT communications | |
| EP3310009A1 (en) | A framework for traffic engineering in software defined networking | |
| US11178056B2 (en) | Communication method and apparatus for optimizing TCP congestion window | |
| Flushing et al. | A mobility-assisted protocol for supervised learning of link quality estimates in wireless networks | |
| US20140120930A1 (en) | Method, Apparatus, Computer Program Product and System for Communicating Predictions | |
| WO2021103706A1 (en) | Data packet sending control method, model training method, device, and system | |
| Kudelski et al. | A mobility-controlled link quality learning protocol for multi-robot coordination tasks | |
| Yu et al. | Reliable cybertwin-driven concurrent multipath transfer with deep reinforcement learning | |
| Shreyas et al. | CAFOR: congestion avoidance using fuzzy logic to find an optimal routing path in 6LoWPAN networks | |
| Scanzio et al. | Predicting Wi‐Fi link quality through artificial neural networks | |
| CN114616810A (en) | Network path redirection | |
| Houidi et al. | Constrained deep reinforcement learning for smart load balancing | |
| Montoya et al. | Energy-efficient and delay sensitive routing paths using mobility prediction in mobile WSN: mathematical optimization, Markov chains, and deep learning approaches | |
| CN111901237B (en) | Source routing method and system, related device and computer readable storage medium | |
| US11140083B1 (en) | Load balancing over a plurality of packet forwarding components | |
| CN114500561B (en) | Power Internet of things network resource allocation decision-making method, system, equipment and medium | |
| CN115866610A (en) | Radio access network resource management based on Reinforcement Learning (RL) and Graph Neural Network (GNN) | |
| Kadhim et al. | Improving IoT applications using a proposed routing protocol | |
| Millan et al. | Tracking and predicting end-to-end quality in wireless community networks | |
| Liao et al. | Cross-layer path selection in multi-path transport protocol for mobile devices | |
| WO2021219214A1 (en) | Device and method for load balancing | |
| Rege et al. | Generation of realistic cloud access times for mobile application testing using transfer learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |