Cell switching method, device, equipment, medium and program product

Through the deep reinforcement learning model, filtering the target neighborhood and configuring temporary link interfaces, the problem of difficulty in switching drones in low-altitude environments is solved, seamless cell handover is achieved under the lack of neighborhood relationships, and the mobile communication network is improved Coverage performance.

CN118984470APending Publication Date: 2024-11-19CHINA MOBILE GROUP DESIGN INST +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410935088.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-12
Publication Date
2024-11-19

AI Technical Summary

Technical Problem

In cellular wireless systems, in low-altitude environment, the signal disorder caused by the lack of fixed base station coverage, and the user terminal is prone to switch to a remote cell, resulting in disconnection and affecting the switching performance of low-altitude cell.

Method used

Through the deep reinforcement learning model, filter reasonable target neighbors, define temporary neighbor relationships, and configure temporary link interfaces to enable user terminals to switch from the current serving cell to the target neighbor, breaking through the limitations of neighbor relationship handover.

Benefits of technology

In the absence of neighborhood relationship, user terminals can seamlessly switch to cells with far distances, improving the coverage performance and user experience of the mobile communication network.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118984470A_ABST
    Figure CN118984470A_ABST
Patent Text Reader

Abstract

Disclosed are a cell switching method, apparatus, device, medium and program product, based on link data reported by a user terminal, a reasonable target neighbor cell can be screened out through a deep reinforcement learning model, and then a temporary neighbor cell relationship between a current serving cell and the target neighbor cell is defined, and when the base station corresponding to the current service cell is different from the base station corresponding to the target adjacent cell, configuring a temporary link interface between the base station corresponding to the current service cell and the base station corresponding to the target adjacent cell so as to indicate the user terminal to switch from the current service cell to the target adjacent cell. And finally, deleting the temporary neighbor relation and the temporary link interface, thereby breaking through the limitation of switching the neighbor relation, and enabling the user terminal to be switched to a long-distance cell in the absence of the neighbor relation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of mobile communication technology, and in particular to a cell switching method, device, equipment, computer-readable storage medium and computer program product. Background Art

[0002] As the military and civilian applications of drone technology continue to expand, the demand for communication capabilities continues to increase. 4G / 5G networked drones have become an inevitable trend, and the expanding number of drone terminal users has gradually become a user group that operators focus on protecting. In cellular wireless systems, mobility in connected mode is achieved through a switching mechanism. By reallocating an ongoing session handled by one cell to another cell, the user terminal can move seamlessly within the coverage of the network. Therefore, the quality of the cell switching performance of the user terminal can largely determine the user's mobility performance, which in turn affects the user's usage perception.

[0003] However, the development of the low-altitude economy has made the coverage requirements of mobile communication networks more three-dimensional. The original two-dimensional coverage mode has increased the requirements for airspace. In the existing technology, since there is no fixed base station coverage at low altitudes, the signal is abnormally disordered, and the user terminal is prone to switching to a distant cell, resulting in disconnection due to the lack of neighboring cell relations, which affects the low-altitude cell switching performance of the user terminal. Summary of the invention

[0004] The present invention provides a cell switching method, device, equipment, medium and program product. Based on the link data reported by the user terminal, a reasonable target neighboring area can be screened out through a deep reinforcement learning model, and then a temporary neighboring area relationship between the current serving cell and the target neighboring area is defined. When the base station corresponding to the current serving cell is different from the base station corresponding to the target neighboring area, a temporary link interface between the base station corresponding to the current serving cell and the base station corresponding to the target neighboring area is configured to instruct the user terminal to switch from the current serving cell to the target neighboring area. Finally, the temporary neighboring area relationship and the temporary link interface are deleted, thereby breaking through the limitation of the neighboring area relationship switching, so that the user terminal can switch to a cell farther away in the absence of a neighboring area relationship.

[0005] In order to solve the above technical problem, a first aspect of an embodiment of the present invention provides a cell switching method, including:

[0006] Obtaining serving cell link data and neighboring cell link data reported by the user terminal;

[0007] When the link scenario of the user terminal belongs to the preset long-distance coverage scenario, determining a number of candidate target neighboring cells that have no neighboring cell relationship with the current serving cell according to the neighboring cell link data;

[0008] According to the serving cell link data, respectively calculating the distance between each candidate target neighboring cell and the user terminal and the coverage angle difference;

[0009] According to the serving cell link data, the neighboring cell link data, the distance and the coverage angle difference, a preset deep reinforcement learning model is used to select a target neighboring cell from a plurality of candidate target neighboring cells;

[0010] Defining a temporary neighboring cell relationship between the current serving cell and the target neighboring cell, and configuring a temporary link interface between the base station corresponding to the current serving cell and the base station corresponding to the target neighboring cell when the base station corresponding to the current serving cell is different from the base station corresponding to the target neighboring cell, so as to instruct the user terminal to switch from the current serving cell to the target neighboring cell;

[0011] When it is detected that the user terminal meets the preset deletion conditions of the temporary neighbor relationship and the temporary link interface, the temporary neighbor relationship and the temporary link interface are deleted.

[0012] As a preferred solution, the method of selecting a target neighboring area from a plurality of candidate target neighboring areas by using a preset deep reinforcement learning model according to the serving cell link data, the neighboring area link data, the distance and the coverage angle difference specifically includes:

[0013] Determining a current connection state of the user terminal according to the serving cell link data, the neighboring cell link data, the distance and the coverage angle difference;

[0014] Based on the connection state and with the goal of maximizing the Q function value, the deep reinforcement learning model is used to screen out the target neighboring area from a plurality of the candidate target neighboring areas.

[0015] As a preferred embodiment, the method further comprises:

[0016] According to a first preset threshold value corresponding to the serving cell link data, the serving cell link data is compared with the first preset threshold value to obtain a first link data comparison result;

[0017] According to the second preset threshold value corresponding to the neighboring cell link data, the neighboring cell link data is compared with the second threshold value to obtain a second link data comparison result;

[0018] When the comparison result of the first link data and the comparison result of the second link data meet a preset judgment condition, determining that the link scenario of the user terminal belongs to the long-distance coverage scenario;

[0019] The first preset threshold value includes a timing advance threshold value, an uplink signal-to-noise ratio threshold value, and a downlink signal-to-noise ratio threshold value; the second preset threshold value includes a link signal strength threshold value;

[0020] The preset decision condition is specifically: the timing advance of the current serving cell exceeds the timing advance threshold, the uplink signal-to-noise ratio or the downlink signal-to-noise ratio of the current serving cell is lower than the uplink signal-to-noise ratio threshold or the downlink signal-to-noise ratio threshold, and the link signal strength of the neighboring cell exceeds the link signal strength threshold and the link signal strength of the current serving cell.

[0021] As a preferred solution, determining a number of candidate target neighboring cells that have no neighboring cell relationship with the current serving cell according to the neighboring cell link data specifically includes:

[0022] Acquire several neighboring cell center frequency point numbers and physical cell identifiers from the neighboring cell link data;

[0023] A plurality of neighboring cells having the same neighboring cell center frequency number as the physical cell identifier and having no neighboring cell relationship with the current serving cell are screened, and the screened plurality of neighboring cells are used as the candidate target neighboring cells.

[0024] As a preferred solution, the method calculates the distance between each candidate target neighboring area and the user terminal according to the serving cell link data, specifically comprising:

[0025] Acquire the timing advance, antenna arrival angle and location information of the current serving cell from the serving cell link data;

[0026] Calculate the location information of the user terminal according to the time advance, the antenna arrival angle, the location information, a preset unit time advance distance, a unit antenna arrival angle, and a conversion relationship between longitude and latitude and length;

[0027] According to the location information of the user terminal and the preset location information of each candidate target neighboring area, the distance between each candidate target neighboring area and the user terminal is calculated.

[0028] As a preferred solution, the method calculates the coverage angle difference between each candidate target neighboring area and the user terminal according to the serving cell link data, specifically comprising:

[0029] Acquire the coverage azimuth of the current serving cell from the serving cell link data;

[0030] According to the coverage azimuth of the current serving cell and the preset coverage azimuth of each candidate target neighboring cell, the coverage angle difference between each candidate target neighboring cell and the user terminal is calculated.

[0031] As a preferred solution, when it is detected that the user terminal satisfies the preset deletion condition of the temporary neighbor relationship and the temporary link interface, deleting the temporary neighbor relationship and the temporary link interface specifically includes:

[0032] When it is detected that the user terminal switches to the target neighboring area, or switches to any candidate target neighboring area other than the target neighboring area, or the switching fails, or the switching window time exceeds a preset time threshold, the temporary neighboring area relationship and the temporary link interface are deleted; wherein the switching window time starts when the definition of the temporary neighboring area relationship and the temporary link interface is completed.

[0033] As a preferred embodiment, the method further comprises:

[0034] Obtaining a reward value corresponding to each cell switching result of the user terminal; wherein the cell switching result includes: switching from the current serving cell to the target neighboring cell, not initiating a cell switching, and switching from the current serving cell to any candidate target neighboring cell other than the target neighboring cell;

[0035] Generate a training sample according to the connection state of the user terminal at each cell switching, the screening action corresponding to the target neighboring cell, and the reward value;

[0036] When the number of the training samples reaches a preset threshold, a number of the training samples are randomly selected, and the deep reinforcement learning model is trained using a gradient descent algorithm to update the parameters of the deep reinforcement learning model.

[0037] A second aspect of an embodiment of the present invention provides a cell switching device, including:

[0038] A link data acquisition module is used to acquire the serving cell link data and neighboring cell link data reported by the user terminal;

[0039] A candidate target neighboring cell determination module, configured to determine, when the link scenario of the user terminal belongs to a preset long-distance coverage scenario, a number of candidate target neighboring cells that do not have a neighboring cell relationship with the current serving cell according to the neighboring cell link data;

[0040] A data processing module, used to calculate the distance and coverage angle difference between each candidate target neighboring cell and the user terminal according to the serving cell link data;

[0041] A target neighboring area screening module, configured to screen a target neighboring area from a plurality of candidate target neighboring areas using a preset deep reinforcement learning model according to the serving cell link data, the neighboring area link data, the distance and the coverage angle difference;

[0042] A cell switching indication module, used to define a temporary neighboring cell relationship between a current serving cell and the target neighboring cell, and when the base station corresponding to the current serving cell is different from the base station corresponding to the target neighboring cell, configure a temporary link interface between the base station corresponding to the current serving cell and the base station corresponding to the target neighboring cell, so as to instruct the user terminal to switch from the current serving cell to the target neighboring cell;

[0043] The temporary configuration information deletion module is used to delete the temporary neighbor relationship and the temporary link interface when it is detected that the user terminal meets the preset deletion conditions of the temporary neighbor relationship and the temporary link interface.

[0044] A third aspect of an embodiment of the present invention provides a cell switching device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the cell switching method described in any one of the first aspects when executing the computer program.

[0045] A fourth aspect of an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the cell switching method described in any one of the first aspects.

[0046] A fifth aspect of an embodiment of the present invention provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the cell switching method described in any one of the first aspects.

[0047] Compared with the prior art, the beneficial effect of the embodiments of the present invention lies in that, based on the link data reported by the user terminal, a reasonable target neighboring area can be screened out through a deep reinforcement learning model, and then a temporary neighboring area relationship between the current service cell and the target neighboring area is defined, and when the base station corresponding to the current service cell is different from the base station corresponding to the target neighboring area, a temporary link interface is configured between the base station corresponding to the current service cell and the base station corresponding to the target neighboring area to instruct the user terminal to switch from the current service cell to the target neighboring area, and finally the temporary neighboring area relationship and the temporary link interface are deleted, thereby breaking through the limitation of the neighboring area relationship switching, so that the user terminal can switch to a cell farther away in the absence of a neighboring area relationship. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 is a schematic flow chart of a cell switching method in an embodiment of the present invention;

[0049] Figure 2 is a node schematic diagram of deep reinforcement learning in an embodiment of the present invention;

[0050] Figure 3 is a schematic diagram of the overall architecture of a base station in an embodiment of the present invention;

[0051] Figure 4 is a cell switching flow chart in an embodiment of the present invention;

[0052] Figure 5 is a schematic structural diagram of a cell switching device in an embodiment of the present invention;

[0053] Figure 6 It is a structural diagram of a cell switching device in an embodiment of the present invention. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] See also Figure 1 A first aspect of an embodiment of the present invention provides a cell switching method, comprising the following steps S1 to S6:

[0056] Step S1, obtaining serving cell link data and neighboring cell link data reported by the user terminal;

[0057] Step S2, when the link scenario of the user terminal belongs to the preset long-distance coverage scenario, determining a number of candidate target neighboring cells that have no neighboring cell relationship with the current serving cell according to the neighboring cell link data;

[0058] Step S3, calculating the distance and coverage angle difference between each candidate target neighboring cell and the user terminal according to the serving cell link data;

[0059] Step S4, screening out a target neighboring area from a plurality of candidate target neighboring areas using a preset deep reinforcement learning model according to the serving cell link data, the neighboring area link data, the distance and the coverage angle difference;

[0060] Step S5, defining a temporary neighboring cell relationship between the current serving cell and the target neighboring cell, and when the base station corresponding to the current serving cell is different from the base station corresponding to the target neighboring cell, configuring a temporary link interface between the base station corresponding to the current serving cell and the base station corresponding to the target neighboring cell, so as to instruct the user terminal to switch from the current serving cell to the target neighboring cell;

[0061] Step S6: When it is detected that the user terminal meets the preset deletion conditions of the temporary neighbor relationship and the temporary link interface, the temporary neighbor relationship and the temporary link interface are deleted.

[0062] Specifically, in order to break through the limitation of neighboring cell relationship switching, this embodiment needs to screen the target neighboring cell based on the link data reported by the user terminal, and then define the temporary neighboring cell relationship between the current service cell and the target neighboring cell. It can be understood that the link data reported by the user terminal includes not only the link data of the current service cell, but also the link data of the cell that has no neighboring cell relationship with the current service cell, so that the target neighboring cell can be screened in the absence of a neighboring cell relationship. Exemplarily, the service cell link data in this embodiment includes but is not limited to the downlink signal strength, downlink signal-to-noise ratio, uplink signal strength, uplink signal-to-noise ratio, time advance and antenna arrival angle of the current service cell; the neighboring cell link data includes but is not limited to the center frequency point number, physical cell identifier and downlink signal strength of the neighboring cell measured by the user terminal.

[0063] Furthermore, the low-altitude scenario is an airspace below 500 meters, which involves a large area. User terminals in low-altitude scenarios are prone to occupy distant cells. If the cell switching method based on neighboring cell relationships in the prior art is adopted, it is easy to cause disconnection due to the lack of neighboring cell relationships. This embodiment is applicable to cell switching in long-distance coverage scenarios. When it is determined that the link scenario of the user terminal belongs to the preset long-distance coverage scenario, several candidate target neighboring cells that do not have a neighboring cell relationship with the current serving cell are determined based on the neighboring cell link data measured by the user terminal.

[0064] Furthermore, the distance between the current serving cell and the neighboring cell and the coverage angle difference have a significant impact on cell switching. Specifically, when a user terminal moves from one cell to another, a cell switching is required to maintain the continuity of calls or data transmission. If the distance between the two cells is too far, it may cause switching delays, or even call failures in extreme cases; in addition, long-distance switching may increase the risk of switching errors, such as misjudging as cross-area coverage, resulting in the user being unable to smoothly switch to the correct cell. The coverage angle difference will affect the coverage overlap area between cells, and thus affect the behavior of cell switching. If the coverage angle difference between the two cells is large, their coverage areas may not overlap well, which may cause users in the cell edge area to find a suitable target cell for switching, especially in the case of high-speed movement. Therefore, in order to reasonably screen the target neighboring area that needs to be switched, this embodiment calculates the distance between each candidate target neighboring area and the user terminal and the coverage angle difference as the indicator information to be considered when screening the target neighboring area.

[0065] Furthermore, since the user terminal measures only the central frequency number and physical cell identifier of the neighboring cell, which cannot uniquely identify a cell, this embodiment uses a deep reinforcement learning model to learn multi-dimensional link feature data, including service cell link data, neighboring cell link data, the distance between each candidate target neighboring cell and the user terminal, and the coverage angle difference, so that the user link in a long-distance coverage scenario and lacking a neighboring cell relationship can be accurately and reasonably switched to the target neighboring cell.

[0066] Furthermore, this embodiment defines a temporary neighboring cell relationship between the current serving cell and the target neighboring cell, so that the user terminal can be switched to a cell farther away when the corresponding neighboring cell relationship is originally missing. In addition, if the base station corresponding to the current serving cell is different from the base station corresponding to the target neighboring cell, in order to ensure successful cell switching, a temporary link interface between the base stations of the two cells needs to be configured, so that during the cell switching process, the current serving cell can send a switching request to the target neighboring cell, and the target neighboring cell can also feedback a corresponding switching response.

[0067] Furthermore, considering that the added temporary neighbor relationship is a redundant neighbor relationship for the ground user terminal, when it is detected that the user terminal meets the preset deletion conditions of the temporary neighbor relationship and the temporary link interface, it indicates that the current cell switching process of the user terminal has ended. In order to avoid negative impact on the mobile communication network of the ground user terminal, this embodiment deletes the current temporary neighbor relationship and temporary link interface.

[0068] The cell switching method provided in an embodiment of the present invention can screen out a reasonable target neighboring area through a deep reinforcement learning model based on the link data reported by the user terminal, and then define a temporary neighboring area relationship between the current service cell and the target neighboring area, and when the base station corresponding to the current service cell is different from the base station corresponding to the target neighboring area, configure a temporary link interface between the base station corresponding to the current service cell and the base station corresponding to the target neighboring area to instruct the user terminal to switch from the current service cell to the target neighboring area, and finally delete the temporary neighboring area relationship and the temporary link interface, thereby breaking through the limitation of the neighboring area relationship switching, so that the user terminal can switch to a cell farther away in the absence of a neighboring area relationship.

[0069] As a preferred solution, the method of selecting a target neighboring area from a plurality of candidate target neighboring areas by using a preset deep reinforcement learning model according to the serving cell link data, the neighboring area link data, the distance and the coverage angle difference specifically includes:

[0070] Determining a current connection state of the user terminal according to the serving cell link data, the neighboring cell link data, the distance and the coverage angle difference;

[0071] Based on the connection state, with the goal of maximizing the Q function value, the deep reinforcement learning model is used to screen out the target neighboring area from the plurality of candidate target neighboring areas.

[0072] Specifically, Figure 2 As shown in the figure, each node setting of deep reinforcement learning is shown. The deep reinforcement learning model is used to determine the best target neighboring area for the connection of the user terminal. The connection state of the user terminal at the cycle time t is represented by S t , the process of the deep reinforcement learning model determining the target neighborhood is to execute action A t , and then the environment feedbacks the reward or punishment for the action and enters the next state S t+1 .

[0073] This embodiment first defines the action space, state space and reward function of deep reinforcement learning. The action space refers to the set of all candidate target neighboring cells and empty cells. During the cell switching process, a decision needs to be made to select a cell as the target neighboring cell. An empty cell means that there is no suitable target neighboring cell. The state space refers to the set of all possible states of the user terminal connection in the wireless environment. The user terminal connection is based on the current state and selects a reasonable target neighboring cell from the action space to perform a switching operation. After the switching operation, the connection will enter a new state and the effect of the cell switching process will be evaluated according to the reward function. This embodiment defines the connection state of the user terminal as the indicator information of the current serving cell and each candidate target neighboring cell. For example, the connection state of the user terminal at time t is expressed as: s(t)=[s c (t),r c (t),u c (t),v c (t),w c (t),q c (t),h c (t),d c1 (t),...,d cn (t),z c1 (t),...,z cn (t)],s c (t) represents the downlink signal strength of the current serving cell, r c (t) represents the downlink signal-to-noise ratio of the current serving cell, u c (t) represents the uplink signal strength of the current serving cell, v c (t) represents the uplink signal-to-noise ratio of the current serving cell, w c (t) represents the timing advance of the current serving cell, q c (t) represents the antenna arrival angle of the current serving cell, h c (t) represents the downlink signal strength of the candidate target neighboring cell, d ci (t) represents the distance between the ith candidate target neighboring area and the user terminal, z ci (t) represents the coverage angle difference between the ith candidate target neighboring area and the user terminal. The reward function is used to calculate the reward value obtained after selecting an action in a certain connection state and transferring to the next connection state.

[0074] After determining the current connection status of the user terminal, based on the current connection status, the goal is to maximize the Q function value, that is, argmax a′ Q(s,a′), the candidate target neighborhood corresponding to the screening action a′ with the largest Q function value is the target neighborhood finally output by the deep reinforcement learning model.

[0075] As a preferred embodiment, the method further comprises:

[0076] According to a first preset threshold value corresponding to the serving cell link data, the serving cell link data is compared with the first preset threshold value to obtain a first link data comparison result;

[0077] According to the second preset threshold value corresponding to the neighboring cell link data, the neighboring cell link data is compared with the second threshold value to obtain a second link data comparison result;

[0078] When the comparison result of the first link data and the comparison result of the second link data meet a preset judgment condition, determining that the link scenario of the user terminal belongs to the long-distance coverage scenario;

[0079] The first preset threshold value includes a timing advance threshold value, an uplink signal-to-noise ratio threshold value, and a downlink signal-to-noise ratio threshold value; the second preset threshold value includes a link signal strength threshold value;

[0080] The preset decision condition is specifically: the timing advance of the current serving cell exceeds the timing advance threshold, the uplink signal-to-noise ratio or the downlink signal-to-noise ratio of the current serving cell is lower than the uplink signal-to-noise ratio threshold or the downlink signal-to-noise ratio threshold, and the link signal strength of the neighboring cell exceeds the link signal strength threshold and the link signal strength of the current serving cell.

[0081] It is worth noting that this embodiment determines whether the link of the current user is in a low-altitude long-distance coverage scenario according to the link signal characteristics of the current serving cell and the neighboring cell. Specifically, the first preset threshold value includes a timing advance threshold value, an uplink signal-to-noise ratio threshold value, and a downlink signal-to-noise ratio threshold value. This embodiment compares the serving cell link data with each first preset threshold value. If the timing advance value w of the current serving cell is c (t) exceeds the time advance threshold value w m , that is, w c (t)>w m , it indicates that the user terminal exceeds the ideal coverage of the current serving cell; if the uplink signal-to-noise ratio of the current serving cell is r c (t) or downlink signal-to-noise ratio v c (t) is lower than the uplink signal-to-noise ratio threshold r m Or the downlink signal-to-noise ratio threshold value v m , that is, r c (t) <r m or v c (t) <v m, it indicates that the current user terminal may be too far away, resulting in a low communication quality index, thereby causing signal attenuation. In addition, the second preset threshold value includes a link signal strength threshold value. This embodiment also compares the neighboring area link data with the second threshold value. If the neighboring area link signal strength h c (t) exceeds the link signal strength threshold h m And the link signal strength s of the current serving cell c (t), that is, h c (t)>h m And s c (t) <h c (t), it indicates that the user terminal has left the coverage of the current serving cell and entered the coverage of another neighboring cell. When the serving cell link data and the neighboring cell link data meet the above judgment conditions, it indicates that the link of the user terminal is in a low-altitude long-distance coverage scenario.

[0082] As a preferred solution, determining a number of candidate target neighboring cells that have no neighboring cell relationship with the current serving cell according to the neighboring cell link data specifically includes:

[0083] Acquire several neighboring cell center frequency point numbers and physical cell identifiers from the neighboring cell link data;

[0084] A plurality of neighboring cells having the same neighboring cell center frequency number as the physical cell identifier and having no neighboring cell relationship with the current serving cell are screened, and the screened plurality of neighboring cells are used as the candidate target neighboring cells.

[0085] It is worth noting that in the neighboring cell link data reported by the user terminal, the neighboring cell identity information only includes the center frequency point number and the physical cell identifier. There may be multiple cells in the analyzed area with the same neighboring cell identity information. Therefore, it is necessary to further determine the cell name or CGI cell global identification information of the target neighboring cell through data analysis for subsequent temporary neighboring cell relationship addition. Specifically, this embodiment selects neighboring cells with the same center frequency point number and physical cell identifier, thereby determining several candidate target neighboring cells that do not have a neighboring cell relationship with the current serving cell.

[0086] As a preferred solution, the method calculates the distance between each candidate target neighboring area and the user terminal according to the serving cell link data, specifically comprising:

[0087] Acquire the timing advance, antenna arrival angle and location information of the current serving cell from the serving cell link data;

[0088] Calculate the location information of the user terminal according to the time advance, the antenna arrival angle, the location information, a preset unit time advance distance, a unit antenna arrival angle, and a conversion relationship between longitude and latitude and length;

[0089] According to the location information of the user terminal and the preset location information of each candidate target neighboring area, the distance between each candidate target neighboring area and the user terminal is calculated.

[0090] Specifically, this embodiment calculates the location information of the user terminal through the following expression:

[0091] x=-k*sin(q c (t)*N)*(w c (t)*L+L / 2)+u;

[0092] y=h*cos(q c (t)*N)*(w c (t)*L+L / 2)+v;

[0093] Wherein, x and y represent the longitude and latitude of the user terminal respectively; q c (t) represents the antenna arrival angle of the current serving cell; w c (t) represents the time advance of the current serving cell; N represents the unit antenna arrival angle; L represents the unit time advance distance; u and v represent the longitude and latitude of the current serving cell respectively, which are obtained from the location information of the current serving cell; k represents the conversion coefficient between meters and longitude units, and its value is 0.000009; h represents the conversion coefficient between meters and latitude units, and its value changes with the current latitude, and is calculated by the formula: 1 / (111*cosv) / 360.

[0094] Furthermore, after determining the longitude and latitude of the user terminal and obtaining the longitude and latitude of each candidate target neighboring area from the location information of each candidate target neighboring area, the distance between each candidate target neighboring area and the user terminal can be calculated.

[0095] As a preferred solution, the method calculates the coverage angle difference between each candidate target neighboring area and the user terminal according to the serving cell link data, specifically comprising:

[0096] Acquire the coverage azimuth of the current serving cell from the serving cell link data;

[0097] According to the coverage azimuth of the current serving cell and the preset coverage azimuth of each candidate target neighboring cell, the coverage angle difference between each candidate target neighboring cell and the user terminal is calculated.

[0098] Specifically, the coverage angle difference refers to the relative directional difference in the coverage range of two cells in space. Therefore, this embodiment can obtain the coverage angle difference between each candidate target neighboring cell and the user terminal by calculating the difference between the coverage azimuth of the current service cell and the coverage azimuth of each candidate target neighboring cell.

[0099] As a preferred solution, when it is detected that the user terminal satisfies the preset deletion condition of the temporary neighbor relationship and the temporary link interface, deleting the temporary neighbor relationship and the temporary link interface specifically includes:

[0100] When it is detected that the user terminal switches to the target neighboring area, or switches to any candidate target neighboring area other than the target neighboring area, or the switching fails, or the switching window time exceeds a preset time threshold, the temporary neighboring area relationship and the temporary link interface are deleted; wherein the switching window time starts when the definition of the temporary neighboring area relationship and the temporary link interface is completed.

[0101] It is worth noting that when the user terminal performs a cell switching, it may successfully switch to the target neighboring cell, but it may also switch to any candidate target neighboring cell other than the target neighboring cell due to incorrect configuration of the temporary neighboring cell relationship, or fail to switch directly, or its switching window time exceeds the preset time threshold. Therefore, when this embodiment detects that the user terminal meets the above-mentioned temporary neighboring cell relationship and temporary link interface deletion conditions, considering that the added temporary neighboring cell relationship is a redundant neighboring cell relationship for the ground user terminal, in order to avoid negative impact on the mobile communication network of the ground user terminal, this embodiment deletes the currently configured temporary neighboring cell relationship and temporary link interface.

[0102] As an optional embodiment, Figure 3 As shown, this embodiment proposes an overall base station architecture. It is worth noting that the switching decision in the 4G / 5G network is located on the base station side. In order to realize the switching of temporary neighboring cell relationships of users, it is necessary to add corresponding functions on the base station side. The existing switching method is based on the switching of neighboring cell relationships, that is, all switching needs to be based on the existence of neighboring cell relationships. This method is suitable for ground communication networks. However, it has obvious shortcomings for low-altitude networks with cluttered signals. Therefore, a temporary neighboring cell relationship switching function module is supplemented on this basis to be applicable to the cell switching method proposed in the embodiment of the present invention. Figure 3In the present invention, the neighbor cell relationship switching function is the original switching function on the base station side, including the delivery of measurement control, execution of switching algorithm (switching decision), application of various interface resources, processing of switching signaling flow, docking with the temporary neighbor cell relationship switching function, etc. The temporary neighbor cell relationship switching function is used for the temporary definition and deletion of neighbor cell relationships, definition and deletion of temporary link interfaces between base stations, determination of target neighbor cells, docking with the original neighbor cell relationship switching function, etc. The operation and maintenance system is used to provide basic information and configuration data for the base station side, and to update and maintain changes in a timely manner.

[0103] Accordingly, if Figure 4 As shown, it is a cell switching flow chart in this embodiment. The user terminal reports the measured link data to the serving cell base station, which uses the temporary neighbor relationship switching function for further processing. First, the link conditions are analyzed to determine whether it belongs to a long-distance coverage scenario, and then the target neighbor is determined based on the data analysis. The temporary neighbor relationship between the serving cell and the target neighbor is further defined for the link, and the temporary link interface between the base stations is configured. Finally, the switching is determined and executed according to the existing switching decision algorithm. The user terminal switches from the serving cell to the target neighbor. After the switching is completed, the temporary neighbor relationship for the link is deleted, and the link interface configuration between the base stations is deleted.

[0104] As a preferred embodiment, the method further comprises:

[0105] Obtaining a reward value corresponding to each cell switching result of the user terminal; wherein the cell switching result includes: switching from the current serving cell to the target neighboring cell, not initiating a cell switching, and switching from the current serving cell to any candidate target neighboring cell other than the target neighboring cell;

[0106] Generate a training sample according to the connection state of the user terminal at each cell switching, the screening action corresponding to the target neighboring cell, and the reward value;

[0107] When the number of the training samples reaches a preset threshold, a number of the training samples are randomly selected, and the deep reinforcement learning model is trained using a gradient descent algorithm to update the parameters of the deep reinforcement learning model.

[0108] Specifically, the reward value in this embodiment is mainly evaluated based on the cell switching situation of the user terminal after the target neighboring area is confirmed. If the cell switching result is switching from the current serving cell to the target neighboring area, a reward value b will be obtained. If the cell switching result is that the cell switching is not initiated, neither the reward value nor the penalty value is obtained. If the cell switching result is switching from the current serving cell to any candidate target neighboring area other than the target neighboring area, it means that an erroneous target neighboring area has been added, and a penalty value of -b will be obtained. It is worth noting that this embodiment does not specifically limit the value of b, and it can be set according to actual application requirements. The specific situation of obtaining the reward value is shown in the following expression:

[0109]

[0110] This embodiment uses a deep Q network to characterize the mapping relationship between the connection state s(t) and the screening action a(t). The connection state s(t) of the user terminal is input into the deep reinforcement learning model, and then the screening action a(t) of the target neighborhood is output. After executing the screening action a(t), the obtained reward value r(t) is collected and stored as a training sample (s(t), a(t), r(t)). When the training sample reaches a preset number threshold, the gradient descent algorithm is used to train the deep reinforcement learning model to maximize the overall reward.

[0111] Specifically, the deep reinforcement learning model needs to be initialized before training, including: initializing the model parameters θ and learning rate γ, setting the exploration rate coefficient ε to 1, defining the exploration decay coefficient m∈(0,1) and the minimum exploration rate ε min .

[0112] At each cell switching, a screening action a(t) is selected with probability according to the proportion of the inverse of the distance between each candidate target neighboring cell and the user terminal; then, based on the deep reinforcement learning model, the connection state s(t) is input into the model and the screening action a(t) that maximizes the Q function value is obtained; further, the screening action a(t) is executed, that is, a temporary neighboring cell relationship between the current serving cell and the target neighboring cell is configured, and a temporary link interface between the base station corresponding to the current serving cell and the base station corresponding to the target neighboring cell is configured, the cell switching result after the screening action a(t) is executed is observed, the reward value r(t) is obtained, and a training sample (s(t), a(t), r(t)) is generated and stored in a preset buffer.

[0113] When the number of training samples reaches the preset threshold, several training samples are randomly selected, and then the reward value is calculated using the following reward function:

[0114] y(t)=r(t)+γmax a′ Q(s(t+1),a′);

[0115] Thus, training data (s(t), a(t), y(t)) is generated, and the gradient descent algorithm is used to update the parameters of the deep reinforcement learning model, and then the exploration rate is set to: ε = max(ε*m,ε min ). It is worth noting that the deep reinforcement learning model in this embodiment can continuously accumulate training samples and continuously update parameters as the number of cell switching increases, thereby achieving continuous optimization of cell switching.

[0116] See also Figure 5 According to a second aspect of an embodiment of the present invention, a cell switching device 100 is provided, comprising:

[0117] The link data acquisition module 11 is used to acquire the serving cell link data and the neighboring cell link data reported by the user terminal;

[0118] A candidate target neighboring cell determination module 12 is used to determine, when the link scenario of the user terminal belongs to a preset long-distance coverage scenario, a number of candidate target neighboring cells that do not have a neighboring cell relationship with the current serving cell according to the neighboring cell link data;

[0119] The data processing module 13 is used to calculate the distance and coverage angle difference between each candidate target neighboring cell and the user terminal according to the serving cell link data;

[0120] A target neighboring area screening module 14 is used to screen a target neighboring area from a plurality of candidate target neighboring areas using a preset deep reinforcement learning model according to the serving cell link data, the neighboring area link data, the distance and the coverage angle difference;

[0121] The cell switching indication module 15 is used to define a temporary neighboring cell relationship between the current serving cell and the target neighboring cell, and when the base station corresponding to the current serving cell is different from the base station corresponding to the target neighboring cell, configure a temporary link interface between the base station corresponding to the current serving cell and the base station corresponding to the target neighboring cell to instruct the user terminal to switch from the current serving cell to the target neighboring cell;

[0122] The temporary configuration information deleting module 16 is used to delete the temporary neighboring cell relationship and the temporary link interface when it is detected that the user terminal meets the preset deletion conditions of the temporary neighboring cell relationship and the temporary link interface.

[0123] As a preferred solution, the target neighboring area screening module 14 is used to screen out a target neighboring area from a plurality of candidate target neighboring areas using a preset deep reinforcement learning model according to the serving cell link data, the neighboring area link data, the distance and the coverage angle difference, specifically including:

[0124] Determining a current connection state of the user terminal according to the serving cell link data, the neighboring cell link data, the distance and the coverage angle difference;

[0125] Based on the connection state, with the goal of maximizing the Q function value, the deep reinforcement learning model is used to screen out the target neighboring area from the plurality of candidate target neighboring areas.

[0126] As a preferred solution, the device further includes a long-distance coverage scene determination module, which is used to:

[0127] According to a first preset threshold value corresponding to the serving cell link data, the serving cell link data is compared with the first preset threshold value to obtain a first link data comparison result;

[0128] According to the second preset threshold value corresponding to the neighboring cell link data, the neighboring cell link data is compared with the second threshold value to obtain a second link data comparison result;

[0129] When the comparison result of the first link data and the comparison result of the second link data meet a preset judgment condition, determining that the link scenario of the user terminal belongs to the long-distance coverage scenario;

[0130] The first preset threshold value includes a timing advance threshold value, an uplink signal-to-noise ratio threshold value, and a downlink signal-to-noise ratio threshold value; the second preset threshold value includes a link signal strength threshold value;

[0131] The preset decision condition is specifically: the timing advance of the current serving cell exceeds the timing advance threshold, the uplink signal-to-noise ratio or the downlink signal-to-noise ratio of the current serving cell is lower than the uplink signal-to-noise ratio threshold or the downlink signal-to-noise ratio threshold, and the link signal strength of the neighboring cell exceeds the link signal strength threshold and the link signal strength of the current serving cell.

[0132] As a preferred solution, the candidate target neighboring area determination module 12 is used to determine a number of candidate target neighboring areas that have no neighboring area relationship with the current serving cell according to the neighboring area link data, specifically including:

[0133] Acquire several neighboring cell center frequency point numbers and physical cell identifiers from the neighboring cell link data;

[0134] A plurality of neighboring cells having the same neighboring cell center frequency number as the physical cell identifier and having no neighboring cell relationship with the current serving cell are screened, and the screened plurality of neighboring cells are used as the candidate target neighboring cells.

[0135] As a preferred solution, the data processing module 13 is used to calculate the distance between each candidate target neighboring area and the user terminal according to the serving cell link data, specifically including:

[0136] Acquire the timing advance, antenna arrival angle and location information of the current serving cell from the serving cell link data;

[0137] Calculate the location information of the user terminal according to the time advance, the antenna arrival angle, the location information, a preset unit time advance distance, a unit antenna arrival angle, and a conversion relationship between longitude and latitude and length;

[0138] According to the location information of the user terminal and the preset location information of each candidate target neighboring area, the distance between each candidate target neighboring area and the user terminal is calculated.

[0139] As a preferred solution, the data processing module 13 is used to calculate the coverage angle difference between each candidate target neighboring area and the user terminal according to the serving cell link data, specifically including:

[0140] Acquire the coverage azimuth of the current serving cell from the serving cell link data;

[0141] According to the coverage azimuth of the current serving cell and the preset coverage azimuth of each candidate target neighboring cell, the coverage angle difference between each candidate target neighboring cell and the user terminal is calculated.

[0142] As a preferred solution, the temporary configuration information deletion module 16 is used to delete the temporary neighbor relationship and the temporary link interface when it is detected that the user terminal meets the preset deletion condition of the temporary neighbor relationship and the temporary link interface, specifically including:

[0143] When it is detected that the user terminal switches to the target neighboring area, or switches to any candidate target neighboring area other than the target neighboring area, or the switching fails, or the switching window time exceeds a preset time threshold, the temporary neighboring area relationship and the temporary link interface are deleted; wherein the switching window time starts when the definition of the temporary neighboring area relationship and the temporary link interface is completed.

[0144] As a preferred solution, the device further includes a model training module, which is used to:

[0145] Obtaining a reward value corresponding to each cell switching result of the user terminal; wherein the cell switching result includes: switching from the current serving cell to the target neighboring cell, not initiating a cell switching, and switching from the current serving cell to any candidate target neighboring cell other than the target neighboring cell;

[0146] Generate a training sample according to the connection state of the user terminal at each cell switching, the screening action corresponding to the target neighboring cell, and the reward value;

[0147] When the number of the training samples reaches a preset threshold, a number of the training samples are randomly selected, and the deep reinforcement learning model is trained using a gradient descent algorithm to update the parameters of the deep reinforcement learning model.

[0148] The cell switching device 100 provided in an embodiment of the present invention can screen out a reasonable target neighboring area through a deep reinforcement learning model based on the link data reported by the user terminal, and then define a temporary neighboring area relationship between the current service cell and the target neighboring area, and when the base station corresponding to the current service cell is different from the base station corresponding to the target neighboring area, configure a temporary link interface between the base station corresponding to the current service cell and the base station corresponding to the target neighboring area to instruct the user terminal to switch from the current service cell to the target neighboring area, and finally delete the temporary neighboring area relationship and the temporary link interface, thereby breaking through the limitation of the neighboring area relationship switching, so that the user terminal can switch to a cell farther away in the absence of a neighboring area relationship.

[0149] See also Figure 6 A third aspect of an embodiment of the present invention provides a cell switching device 200, comprising a memory 22, a processor 21, and a computer program stored in the memory 22 and executable on the processor 21, wherein the processor 21 implements the cell switching method described in any embodiment of the first aspect when executing the computer program.

[0150] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory 22 and executed by the processor 21 to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the cell switching device 200.

[0151] The cell switching device 200 may include, but is not limited to, a processor 21 and a memory 22. Those skilled in the art may understand that the schematic diagram is merely an example of the cell switching device 200 and does not constitute a limitation on the cell switching device 200. The cell switching device 200 may include more or fewer components than shown in the figure, or may combine certain components, or different components. For example, the cell switching device 200 may also include input and output devices, network access devices, buses, etc.

[0152] The processor 21 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor 21 may also be any conventional processor 21, etc. The processor 21 is the control center of the cell switching device 200, and uses various interfaces and lines to connect various parts of the entire cell switching device 200.

[0153] The memory 22 can be used to store the computer program and / or module. The processor 21 implements various functions of the cell switching device 200 by running or executing the computer program and / or module stored in the memory 22 and calling the data stored in the memory 22. The memory 22 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory 22 can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0154] A fourth aspect of an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the cell switching method described in any embodiment of the first aspect.

[0155] A fifth aspect of an embodiment of the present invention provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the cell switching method described in any embodiment of the first aspect.

[0156] Wherein, if the module / unit integrated in the cell switching device 200 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0157] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A cell switching method, characterized in that: include: Obtaining serving cell link data and neighboring cell link data reported by the user terminal; When the link scenario of the user terminal belongs to the preset long-distance coverage scenario, determining a number of candidate target neighboring cells that have no neighboring cell relationship with the current serving cell according to the neighboring cell link data; According to the serving cell link data, respectively calculating the distance between each candidate target neighboring cell and the user terminal and the coverage angle difference; According to the serving cell link data, the neighboring cell link data, the distance and the coverage angle difference, a preset deep reinforcement learning model is used to select a target neighboring cell from a plurality of candidate target neighboring cells; Defining a temporary neighboring cell relationship between the current serving cell and the target neighboring cell, and configuring a temporary link interface between the base station corresponding to the current serving cell and the base station corresponding to the target neighboring cell when the base station corresponding to the current serving cell is different from the base station corresponding to the target neighboring cell, so as to instruct the user terminal to switch from the current serving cell to the target neighboring cell; When it is detected that the user terminal meets the preset deletion conditions of the temporary neighbor relationship and the temporary link interface, the temporary neighbor relationship and the temporary link interface are deleted.

2. The cell switching method according to claim 1, characterized in that: The method of selecting a target neighboring area from a plurality of candidate target neighboring areas by using a preset deep reinforcement learning model according to the serving cell link data, the neighboring area link data, the distance and the coverage angle difference specifically includes: Determining a current connection state of the user terminal according to the serving cell link data, the neighboring cell link data, the distance and the coverage angle difference; Based on the connection state and with the goal of maximizing the Q function value, the deep reinforcement learning model is used to screen out the target neighboring area from a plurality of the candidate target neighboring areas.

3. The cell switching method according to claim 1, characterized in that: The method further comprises: According to a first preset threshold value corresponding to the serving cell link data, the serving cell link data is compared with the first preset threshold value to obtain a first link data comparison result; According to the second preset threshold value corresponding to the neighboring cell link data, the neighboring cell link data is compared with the second threshold value to obtain a second link data comparison result; When the comparison result of the first link data and the comparison result of the second link data meet a preset judgment condition, determining that the link scenario of the user terminal belongs to the long-distance coverage scenario; The first preset threshold value includes a timing advance threshold value, an uplink signal-to-noise ratio threshold value, and a downlink signal-to-noise ratio threshold value; the second preset threshold value includes a link signal strength threshold value; The preset decision condition is specifically: the timing advance of the current serving cell exceeds the timing advance threshold, the uplink signal-to-noise ratio or the downlink signal-to-noise ratio of the current serving cell is lower than the uplink signal-to-noise ratio threshold or the downlink signal-to-noise ratio threshold, and the link signal strength of the neighboring cell exceeds the link signal strength threshold and the link signal strength of the current serving cell.

4. The cell switching method according to claim 1, characterized in that: The determining, according to the neighboring cell link data, a plurality of candidate target neighboring cells that have no neighboring cell relationship with the current serving cell specifically includes: Acquire several neighboring cell center frequency point numbers and physical cell identifiers from the neighboring cell link data; A plurality of neighboring cells having the same neighboring cell center frequency number as the physical cell identifier and having no neighboring cell relationship with the current serving cell are screened, and the screened plurality of neighboring cells are used as the candidate target neighboring cells.

5. The cell switching method according to claim 1, characterized in that: The method calculates the distance between each candidate target neighboring cell and the user terminal according to the serving cell link data, specifically comprising: Acquire the timing advance, antenna arrival angle and location information of the current serving cell from the serving cell link data; Calculate the location information of the user terminal according to the time advance, the antenna arrival angle, the location information, a preset unit time advance distance, a unit antenna arrival angle, and a conversion relationship between longitude and latitude and length; According to the location information of the user terminal and the preset location information of each candidate target neighboring area, the distance between each candidate target neighboring area and the user terminal is calculated.

6. The cell switching method according to claim 1, characterized in that: The method calculates the coverage angle difference between each candidate target neighboring cell and the user terminal according to the serving cell link data, specifically comprising: Acquire the coverage azimuth of the current serving cell from the serving cell link data; According to the coverage azimuth of the current serving cell and the preset coverage azimuth of each candidate target neighboring cell, the coverage angle difference between each candidate target neighboring cell and the user terminal is calculated.

7. The cell switching method according to claim 1, characterized in that: The deleting the temporary neighbor relationship and the temporary link interface when detecting that the user terminal meets the preset deletion condition of the temporary neighbor relationship and the temporary link interface specifically includes: When it is detected that the user terminal switches to the target neighboring area, or switches to any candidate target neighboring area other than the target neighboring area, or the switching fails, or the switching window time exceeds a preset time threshold, the temporary neighboring area relationship and the temporary link interface are deleted; wherein the switching window time starts when the definition of the temporary neighboring area relationship and the temporary link interface is completed.

8. The cell switching method according to claim 2, characterized in that: The method further comprises: Obtaining a reward value corresponding to each cell switching result of the user terminal; wherein the cell switching result includes: switching from the current serving cell to the target neighboring cell, not initiating a cell switching, and switching from the current serving cell to any candidate target neighboring cell other than the target neighboring cell; Generate a training sample according to the connection state of the user terminal at each cell switching, the screening action corresponding to the target neighboring cell, and the reward value; When the number of the training samples reaches a preset threshold, a number of the training samples are randomly selected, and the deep reinforcement learning model is trained using a gradient descent algorithm to update the parameters of the deep reinforcement learning model.

9. A cell switching device, characterized in that: include: A link data acquisition module is used to acquire the serving cell link data and neighboring cell link data reported by the user terminal; A candidate target neighboring cell determination module, configured to determine, when the link scenario of the user terminal belongs to a preset long-distance coverage scenario, a number of candidate target neighboring cells that do not have a neighboring cell relationship with the current serving cell according to the neighboring cell link data; A data processing module, used to calculate the distance and coverage angle difference between each candidate target neighboring cell and the user terminal according to the serving cell link data; A target neighboring area screening module, configured to screen a target neighboring area from a plurality of candidate target neighboring areas using a preset deep reinforcement learning model according to the serving cell link data, the neighboring area link data, the distance and the coverage angle difference; A cell switching indication module, used to define a temporary neighboring cell relationship between a current serving cell and the target neighboring cell, and when the base station corresponding to the current serving cell is different from the base station corresponding to the target neighboring cell, configure a temporary link interface between the base station corresponding to the current serving cell and the base station corresponding to the target neighboring cell, so as to instruct the user terminal to switch from the current serving cell to the target neighboring cell; The temporary configuration information deletion module is used to delete the temporary neighbor relationship and the temporary link interface when it is detected that the user terminal meets the preset deletion conditions of the temporary neighbor relationship and the temporary link interface.

10. A cell switching device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the cell switching method according to any one of claims 1 to 8 when executing the computer program.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the cell switching method according to any one of claims 1 to 8.

12. A computer program product, characterized in that The method comprises a computer program / instruction, which, when executed by a processor, implements the steps of the cell switching method according to any one of claims 1 to 8.