Internet of Things card data rendering method and device, equipment, storage medium and program product

Abstract

The invention discloses an Internet of Things card data rendering method and device, equipment, a storage medium and a program product, and the method comprises the steps: determining a first rendering data range according to a current view region of a map; performing fine adjustment on the first rendering data range according to a range adjustment factor to obtain a second rendering data range; wherein the range adjustment factor is obtained by performing multi-agent reinforcement learning on a plurality of databases storing GIS data of the Internet of Things card; querying candidate GIS data in the range of the second rendering data; rendering the candidate GIS data in the current visual field area; according to the embodiment of the invention, the multi-agent learning algorithm is used for dynamic scheduling, so that the dynamic response time of the GIS data can be effectively shortened, the efficiency of rendering the GIS data is improved, and the query efficiency of the GIS data can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet of Things, and in particular to an Internet of Things card data rendering method, device, equipment, storage medium and program product. Background Art

[0002] IoT cards are mobile communication access services provided by basic network operators to IoT users based on the public IoT. Using IoT-specific number segments, they support basic communication services such as SMS, wireless data, and voice through dedicated network element devices. They can also provide intelligent connection services such as communication connection management and terminals. With the cross-border integration, integrated innovation, and large-scale development of the IoT industry, the use of IoT cards has increased dramatically. Considering the long IoT card industry chain, large terminal scale, and complex application scenarios, it is necessary to control the massive IoT card data rendering and clarify the data flow of IoT cards.

[0003] The control of massive IoT cards requires querying the IoT card data renderings of the required control area from the massive IoT card data renderings. Currently, data query is mainly performed based on big data scheduling. However, data query based on big data scheduling requires scheduling and querying massive data, and the data volume is too large, resulting in low data query efficiency and long data scheduling response time. Summary of the invention

[0004] In view of the problems existing in the prior art, the embodiments of the present invention provide an IoT card data rendering method, device, equipment, storage medium and program product, which can improve the efficiency of GIS data query and shorten the response time of GIS data scheduling.

[0005] In a first aspect, an embodiment of the present invention provides a method for rendering IoT card data, comprising:

[0006] Determine a first rendering data range according to a current field of view of the map;

[0007] The first rendering data range is fine-tuned according to a range adjustment factor to obtain a second rendering data range; wherein the range adjustment factor is obtained by performing multi-agent reinforcement learning on a plurality of databases storing GIS data of IoT cards;

[0008] Querying candidate GIS data within the range of the second rendering data;

[0009] The candidate GIS data is rendered in the current field of view.

[0010] As an improvement of the above solution, the method further includes:

[0011] Collect GIS data and basic data from multiple IoT cards;

[0012] According to the location of each IoT card, the GIS data and basic data of each IoT card are divided into regions to build multiple databases;

[0013] The GIS data of each IoT card is compressed and encoded to obtain the geographic index of each IoT card in the corresponding database.

[0014] As an improvement of the above solution, the GIS data of each IoT card is compressed and encoded to obtain the geographic index of each IoT card in the corresponding database, including:

[0015] Performing GeoHash encoding on each of the GIS data to obtain a first location code of each of the GIS data;

[0016] Perform base32 encoding mapping on each of the first location codes to obtain a second location code of the corresponding GIS data;

[0017] The coding prefix of each of the second coding positions is obtained to obtain the geographic index of the corresponding GIS data.

[0018] As an improvement of the above solution, determining the first rendering data range according to the current field of view of the map includes:

[0019] Expand the current field of view;

[0020] The first rendering data range is determined according to the coordinate point positions corresponding to all GIS data within the current field of view after expansion.

[0021] As an improvement of the above solution, the calculation process of the range adjustment factor includes:

[0022] Constructing a set of environmental states in which the agent is located and a set of actions taken by the agent;

[0023] According to the state set and the action set, a multi-agent reinforcement learning model is constructed based on a flexible action-evaluation algorithm to obtain the range adjustment factor.

[0024] As an improvement of the above scheme, the state set includes environmental information composed of observations of all intelligent agents and data dynamic scheduling characteristics of intelligent agents; the environmental information includes: the data range of all databases during scheduling; the data dynamic scheduling characteristics include: the longitude and latitude coordinates of the top, bottom, left, right and middle of the current field of view, the current scale size of the map, the current screen display scale size, the spatial scope of the current viewport data retrieval, the map scale feature of moving or reducing the map, and a set of mouse movement rate features in all directions.

[0025] As an improvement of the above solution, the action set is a set including a binary representation of the longitude and latitude of the GeoHash string within the data range of the corresponding database when querying any database, the number of the scheduled database, and a range adjustment factor of the corresponding database query.

[0026] As an improvement of the above scheme, the multi-agent reinforcement learning model includes several Q evaluation networks constructed based on the action value function after the agent performs the actions in the action set under the state in the environment state set, several action networks that perform strategy scheduling to execute actions based on the state in the environment state set, a strategy network for outputting the probability distribution of taking actions according to the current state, and a bidirectional long short-term memory neural network for memorizing the actions and strategies performed by the agent each time and predicting the next action to be performed by the agent; wherein each agent learns a Q evaluation network and an action network accordingly.

[0027] As an improvement of the above solution, the range adjustment factor is obtained by constructing a multi-agent reinforcement learning model based on a flexible action-evaluation algorithm according to the state set and the action set, including:

[0028] The environmental information in the state set is used as the state, and the binary representation of the longitude and latitude of the GeoHash string within the data range of the corresponding database when querying any database in the action set is used as the action to input into the Q evaluation network of each intelligent agent, and the action value function of each intelligent agent is calculated;

[0029] The data dynamic scheduling feature in the state set is input into the action network of each agent as the current state, and the execution action output by each agent is obtained based on the action value function of each agent;

[0030] By using the strategy network, the probability distribution of taking corresponding execution actions in the current state is calculated to obtain the strategy corresponding to the corresponding execution action;

[0031] Based on memorizing the execution actions and strategies of each agent each time through a bidirectional long short-term memory neural network, the predicted actions of each agent are obtained; wherein the predicted actions include the number of the scheduled database and the range adjustment factor of the corresponding database query.

[0032] As an improvement of the above solution, the rendering of the candidate GIS data in the current field of view includes:

[0033] Traversing the current visual field after expansion, and selecting target GIS data within the current visual field after expansion from the candidate GIS data;

[0034] Rendering is performed using the target GIS data as a node to obtain rendering data of the corresponding target GIS data;

[0035] The rendering data in the current field of view is displayed, and the rendering data in the area outside the current field of view is hidden.

[0036] In a second aspect, an embodiment of the present invention provides an IoT card data rendering device, comprising:

[0037] A data range determination module, used to determine a first rendering data range according to a current field of view of the map;

[0038] A data range fine-tuning module, used to fine-tune the first rendering data range according to a range adjustment factor to obtain a second rendering data range; wherein the range adjustment factor is obtained by performing multi-agent reinforcement learning on a plurality of databases storing GIS data of IoT cards;

[0039] A data query module, used for querying candidate GIS data within the range of the second rendering data;

[0040] The data rendering module is used to render the candidate GIS data in the current field of view.

[0041] In a third aspect, an embodiment of the present invention provides an Internet of Things card data rendering device, comprising: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements the Internet of Things card data rendering method as described in any one of the first aspects.

[0042] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the IoT card data rendering method as described in any one of the first aspects.

[0043] In a fifth aspect, an embodiment of the present invention provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the IoT card data rendering method as described in any one of the first aspects.

[0044] Compared with the prior art, the method, device, equipment, storage medium and program product for rendering IoT card data in the embodiment of the present invention determines the first rendering data range according to the current field of view of the map; fine-tunes the first rendering data range according to the range adjustment factor to obtain the second rendering data range; wherein the range adjustment factor is obtained by multi-agent reinforcement learning of multiple databases storing GIS data of IoT cards; queries candidate GIS data within the second rendering data range; renders the candidate GIS data in the current field of view to achieve query and rendering display of IoT card GIS data in the current field of view. The embodiment of the present invention uses the range adjustment factor generated by the multi-agent learning algorithm to dynamically schedule the GIS data query rendering range. In the scenario where a large number of IoT cards need to be controlled, only the dynamically scheduled required control area needs to be GIS data queried and rendered, without the need for scheduling and querying of massive data, which can effectively shorten the dynamic response time of GIS data and improve the efficiency of rendering GIS data, thereby improving the efficiency of GIS data query, thereby improving the control effect of massive IoT cards. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solution of the present invention, the drawings used in the implementation mode will be briefly introduced below. Obviously, the drawings described below are only some implementation modes of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 It is a flow chart of a method for rendering IoT card data provided by an embodiment of the present invention;

[0047] Figure 2 It is a schematic diagram of a massive IoT card GIS data dynamic scheduling and rendering process provided by an embodiment of the present invention;

[0048] Figure 3 is a schematic diagram of the network structure of a multi-agent reinforcement learning model provided by an embodiment of the present invention;

[0049] Figure 4 It is a schematic diagram of the architecture of the IoT card data rendering system provided by an embodiment of the present invention;

[0050] Figure 5 It is a structural block diagram of an IoT card data rendering device provided by an embodiment of the present invention;

[0051] Figure 6 It is a structural block diagram of an IoT card data rendering device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0052] 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.

[0053] It is understood that the various numbers involved in the embodiments of the present invention are only for the convenience of description and are not intended to limit the scope of the present application. The size of the sequence number of each process does not mean the order of execution, and the execution order of each process should be determined by its function and internal logic.

[0054] In embodiments of the present invention, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. The terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or equipment. In the absence of more restrictions, the elements defined by the statement "include..." do not exclude the existence of other identical elements in the process, method, article or equipment including the elements. The term "multiple or several" refers to two or more, and the same applies to "multiple / items or several / items". "And / or" describes the association relationship of associated objects, indicating that there can be three relationships, for example, A and / or B, which can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the associated objects before and after are in an "or" relationship.

[0055] See also Figure 1 , Figure 1 1 is a flow chart of a method for rendering data of an IoT card provided by an embodiment of the present invention. The method for rendering data of an IoT card specifically comprises:

[0056] S11: determining a first rendering data range according to a current field of view of the map;

[0057] S12: fine-tuning the first rendering data range according to a range adjustment factor to obtain a second rendering data range; wherein the range adjustment factor is obtained by performing multi-agent reinforcement learning on a plurality of databases storing GIS data of IoT cards;

[0058] S13: Query candidate GIS data within the second rendering data range;

[0059] S14: Rendering the candidate GIS data in the current field of view.

[0060] It should be noted that the IoT card data rendering method described in the embodiment of the present invention can be executed by an IoT card data rendering system, which includes a user end for front-end data rendering and a server end for back-end data query. The user end and the server end can be independently set in different devices and communicated with each other, or the user end and the server end can be integrated in one device and communicated with each other, which is not specifically limited in the embodiment of the present invention.

[0061] The embodiment of the present invention first determines the first rendering data range of the GIS data of the IoT card based on the current field of view of the map; then uses a multi-agent learning algorithm to operate multiple databases storing the GIS data of the IoT card through multiple agents to determine the range adjustment factor of the first rendering data range, that is, the dynamic scheduling range, and then adjusts the first rendering data range based on the range adjustment factor to obtain the second rendering data range; then traverses the second rendering data range to query the candidate GIS data within the range; finally, renders in the current field of view based on the candidate GIS data to achieve query and rendering display of the IoT card GIS data in the current field of view. The embodiment of the present invention uses a multi-agent learning algorithm to dynamically schedule GIS data. In the scenario where a large number of IoT cards need to be controlled, it is only necessary to query and render GIS data for the required control area after dynamic scheduling, without scheduling and querying massive data, effectively shortening the dynamic response time of GIS data, and improving the efficiency of rendering GIS data, thereby improving the efficiency of GIS data query, and thus improving the control effect of massive IoT cards.

[0062] The range adjustment factor may be dynamically changing, for example, it may be called a dynamic range adjustment factor. When the range adjustment is performed according to a certain ratio, it may also be called a range adjustment ratio, a dynamic ratio, or a dynamic range adjustment ratio, etc.

[0063] Furthermore, the method further comprises:

[0064] Collect GIS data and basic data from multiple IoT cards;

[0065] According to the location of each IoT card, the GIS data and basic data of each IoT card are divided into regions to build multiple databases;

[0066] The GIS data of each IoT card is compressed and encoded to obtain the geographic index of each IoT card in the corresponding database.

[0067] For example, the map can be divided into East China, South China, North China, Central China, and West China, and databases for each region can be created; then for each database, multiple database tables are established in the database with province as the granularity, and partitions are established for each urban area of ​​each province with city as the granularity, so that the database table stores the GIS data and basic data of the IoT card at the city granularity. The embodiment of the present invention transmits the GIS data of the IoT card from time to time through the 5G base station to store the distributed database tables representing different regions, thereby realizing the distributed storage of massive IoT card GIS data.

[0068] Each time data is collected, the GIS data of the IoT card in the database will be updated. However, since the GIS data of the massive IoT card is too large, in order to reduce the impact of data scheduling and transmission, the embodiment of the present invention compresses and encodes the GIS data and creates a two-dimensional geographic index of the GIS data in the corresponding database. Through the geographic index, the corresponding GIS data can be queried and scheduled from the database, thereby improving the retrieval efficiency of the IoT card GIS data.

[0069] In some embodiments, compressing and encoding the GIS data of each IoT card to obtain a geographic index of each IoT card in a corresponding database includes:

[0070] Performing GeoHash encoding on each of the GIS data to obtain a first location code of each of the GIS data;

[0071] Perform base32 encoding mapping on each of the first location codes to obtain a second location code of the corresponding GIS data;

[0072] The coding prefix of each of the second coding positions is obtained to obtain the geographic index of the corresponding GIS data.

[0073] For example, the GeoHash encoding algorithm can be used to encode each GIS data first; for example, the GIS data with binary encoding of longitude and latitude on the IoT card is organized together in the form of "longitude string + latitude string arranged alternately in sequence" to form a one-dimensional string, i.e., the first position code. It should be understood that the more bits of this string, the smaller the corresponding rectangular block, and the small rectangular blocks with the same prefix are likely to belong to the same large rectangular block, which means that the longer the number of matching bits of the two string prefixes, the closer the relative positions of the two rectangular blocks.

[0074] Then the first position code is mapped using the base32 encoding format, for example, the 5-bit binary number of the first position code is taken from left to right each time, and the selected 5-bit binary number is converted into decimal corresponding to 10 numeric characters (0-9) and 26 English letters to obtain the second position code. Through the above GeoHash encoding mapping, the encoding length can be shortened.

[0075] Then, considering that the longer the GeoHash is, the smaller the area it represents and the higher the location accuracy is; conversely, the shorter the GeoHash is, the larger the area it represents and the lower the location accuracy is; the embodiment of the present invention can represent a location of any accuracy by changing the length of the second location code, such as selecting the first N characters of the second location code as the code prefix to create a geographic index of the corresponding GIS data.

[0076] In some embodiments, determining the first rendering data range according to the current field of view of the map includes:

[0077] Expand the current field of view;

[0078] The first rendering data range is determined according to the coordinate point positions corresponding to all GIS data within the current field of view after expansion.

[0079] Specifically, rendering the candidate GIS data in the current field of view includes:

[0080] Traversing the current visual field after expansion, and selecting target GIS data within the current visual field after expansion from the candidate GIS data;

[0081] Rendering is performed using the target GIS data as a node to obtain rendering data of the corresponding target GIS data;

[0082] The rendering data in the current field of view is displayed, and the rendering data in the area outside the current field of view is hidden.

[0083] Furthermore, after expanding the current field of view, the method further includes:

[0084] Check whether the current cache contains GIS data that matches the current field of view after expansion.

[0085] If it exists, the matching GIS data is directly used as candidate GIS data;

[0086] If not, execute steps S11 to S13 to determine candidate GIS data.

[0087] It should be understood that “matched GIS data” refers to the GIS data corresponding to the IoT card belonging to the geographical area indicated by the current field of view after expansion.

[0088] In an embodiment of the present invention, in order to reduce the number of requests, when the user terminal initializes the rendering of the GIS data of the IoT card in the current field of view, or when the current field of view can be changed by moving or shrinking the map, the current field of view is first expanded, for example, by 15%, as a rendering space, and only the data in the current field of view is displayed when the data is displayed.

[0089] It should be understood that the 15% expansion means that each time the user end expands the range of ±15% of the GIS coordinate point within the current field of view, and then queries the server for GIS data with each expanded range. In the subsequent traversal rendering, the range expanded by 15% outside the field of view will be rendered, but only the GIS data within the current field of view will be displayed when the data is displayed.

[0090] like Figure 2 The massive IoT card GIS data dynamic scheduling and rendering process shown in the figure, when the user terminal initializes the rendering of the IoT card GIS data in the current field of view, or when the current field of view can be changed by moving or reducing the map, in the current field of view after expansion, it will first query whether there is a matching IoT card GIS data in the cache. If matching GIS data is found in the cache, the GIS data found in the cache will be returned to the user terminal for rendering. If no matching GIS data is found in the cache, the longitude and latitude coordinates of the expanded current field of view will be sampled to determine the first rendering data range. During the sampling process, the longitude and latitude coordinates of the upper, lower, left, right and middle of the expanded current field of view will be selected, and then the current scale size of the map, the current screen display scale size, the spatial range of the current viewport data, the scale features of moving or reducing the map, the speed of the mouse in each direction (such as the southeast, northwest and northeast directions) and other data features will be obtained, and finally a data request will be sent to the server.

[0091] When the server receives the data request, it uses the GenHash algorithm to encode the latitude and longitude coordinates of the selected current field of view into a one-dimensional target GeoHash string (i.e., the first rendering data range). For the specific encoding method, please refer to the GeoHash encoding and base32 encoding mapping process of GIS data encoding above, which will not be repeated here. Then, the multi-agent reinforcement learning algorithm is used to dynamically schedule the database. Each agent will make a strategic decision whether to schedule its corresponding database. If the data needs to be scheduled, the target GeoHash string encoding prefix and the corresponding database geographic index will be matched by assembling SQL statements, and finally a location with a similar GeoHash value (i.e., the encoding prefix of the target GeoHash string matches the geographic index) is found in the database.

[0092] In the process of matching the target GeoHash string, if the matching target GeoHash string is too long, it will lead to too much GIS data of the IoT card being scheduled, the data scheduling time is too long, and the user end will render the GIS data of the IoT card too slowly, resulting in freezes and long loading times; if the matching target GeoHash string is too short, the user end will render incomplete GIS data, resulting in omission of rendering IoT card information, and the user end will request the server to schedule data again for a long time, so the length of the matching target GeoHash string is crucial. Moreover, due to the edge problem of the GeoHash algorithm, some IoT cards with a closer actual distance and different GeoHash codes within the current field of view may be omitted. Therefore, the embodiment of the present invention fine-tunes the binary representation of the longitude and latitude of the target GeoHash string (i.e., the first rendering data range), such as adding or subtracting the range adjustment factor Y, to obtain the GeoHash string of the surrounding area (i.e., the second rendering data range). For the range adjustment factor Y added or subtracted to the target GeoHash string (i.e., the first rendering data range), the multi-agent reinforcement learning algorithm can be used to allow each agent to dynamically operate the database, and then estimate the range adjustment factor Y that needs to be added or subtracted for each data scheduling. Finally, the GIS data of all IoT cards in the vicinity are queried as candidate GIS data and returned to the user end for dynamic rendering of GIS data.

[0093] After the client obtains the query results from the server, it performs front-end rendering, for example, it can use canvas Canvas to render on the HTML5 page. Specifically, the client uses the GIS data within the current field of view as the cached area to traverse the data, and determines whether the traversal process is completed. If not, it will repeatedly query the GIS data. If completed, the number of nodes after rendering is determined according to the traversal, and the rendering degree of each node is automatically adjusted. Among them, the larger the amount of node data and the more concentrated the range, the heavier and darker the rendering color, that is, the higher the rendering degree.

[0094] The data traversal process is as follows: based on the second rendering data range determined above, the candidate GIS data searched and queried are traversed, and the target GIS data existing in the current field of view after expansion is rendered as a node.

[0095] The embodiment of the present invention will appropriately expand the current field of view after obtaining it, such as expanding the length and width by 15% respectively, to form a data rendering space. When the map is panned and browsed, if the new screen space range does not exceed the actual data request range of the last time, there is no need to request rendering again, thereby reducing the number of requests and improving data response efficiency.

[0096] Specifically, the calculation process of the range adjustment factor includes:

[0097] Constructing a set of environmental states in which the agent is located and a set of actions taken by the agent;

[0098] According to the state set and the action set, a multi-agent reinforcement learning model is constructed based on a flexible action-evaluation algorithm to obtain the range adjustment factor.

[0099] Among them, the state set includes environmental information composed of observations of all intelligent agents and data dynamic scheduling characteristics of intelligent agents; the environmental information includes: the data range of all databases during scheduling; the data dynamic scheduling characteristics include: the longitude and latitude coordinates of the top, bottom, left, right and middle of the current field of view, the current scale size of the map, the current screen display scale size, the spatial scope of the current viewport data retrieval, the map scale feature of moving or reducing the map, and the mouse movement rate characteristics in all directions.

[0100] The action set is a set including a binary representation of the longitude and latitude of a GeoHash string within a data range of a corresponding database when querying any database, a number of a scheduled database, and a range adjustment factor of the corresponding database query.

[0101] The multi-agent reinforcement learning model includes several Q evaluation networks constructed based on the action value function after the agent performs the actions in the action set in the state in the environment state set, several action networks that perform strategy scheduling to execute actions based on the state in the environment state set, a strategy network for outputting the probability distribution of taking actions according to the current state, and a bidirectional long short-term memory neural network for memorizing the actions and strategies performed by the agent each time and predicting the next action to be performed by the agent; wherein each agent learns a Q evaluation network and an action network accordingly.

[0102] Specifically, the obtaining of the range adjustment factor by constructing a multi-agent reinforcement learning model based on a flexible action-evaluation algorithm according to the state set and the action set includes:

[0103] The environmental information in the state set is used as the state, and the binary representation of the longitude and latitude of the GeoHash string within the data range of the corresponding database when querying any database in the action set is used as the action to input into the Q evaluation network of each intelligent agent, and the action value function of each intelligent agent is calculated;

[0104] The data dynamic scheduling feature in the state set is input into the action network of each agent as the current state, and the execution action output by each agent is obtained based on the action value function of each agent;

[0105] By using the strategy network, the probability distribution of taking corresponding execution actions in the current state is calculated to obtain the strategy corresponding to the corresponding execution action;

[0106] Based on memorizing the execution actions and strategies of each agent each time through a bidirectional long short-term memory neural network, the predicted actions of each agent are obtained; wherein the predicted actions include the number of the scheduled database and the range adjustment factor of the corresponding database query.

[0107] During the roaming of massive IoT card GIS data, the GIS data of each IoT card can be rendered. In order to effectively shorten the response time of dynamic data scheduling, the embodiment of the present invention adopts a multi-agent reinforcement learning algorithm to perform dynamic data scheduling. The dynamic data scheduling features include: the latitude and longitude coordinates of the top, bottom, left, right and middle of the current field of view, the current scale size of the map, the current screen display scale size, the spatial range of the current viewport data, the characteristics of moving or reducing the map scale, and the characteristics of the mouse moving speed in all directions (east, south, west and north). The environmental information composed of the observations of all agents includes the data range of all databases when scheduling.

[0108] Each agent is allowed to dynamically operate the corresponding database. The embodiment of the present invention builds a multi-agent reinforcement learning model based on the SAC (Soft Actor-Critic, flexible action-evaluation) reinforcement learning algorithm and the bidirectional long short-term memory neural network BilSTM. The SAC (Soft Actor-Critic, flexible action-evaluation) reinforcement learning algorithm is used to formulate a decision strategy for each agent, and the bidirectional long short-term memory neural network BilSTM is used to memorize the execution actions and strategies of each agent. The multi-agent reinforcement learning model mainly involves the update and interaction of the strategy network, the Q evaluation network (also referred to as the Q network), the action network (also described as the Actor network), and the bidirectional long short-term memory neural network BilSTM.

[0109] Each agent needs to learn a Q-evaluation network and an action network. During the training phase, the Q-evaluation network receives the environment information composed of the observations of all agents (i.e., the data range of all database scheduling) and the actions {a1, a2, ..., a n} (i.e., when querying a database, the binary representation of the longitude and latitude of the GeoHash string of the data range of the database) as input to centrally calculate the action value function of the agent; the action network takes the dynamic scheduling characteristics of the agent's data as input, and outputs the execution action taken in the current state based on the corresponding action value function; and the policy network is used to output the probability distribution of the execution action taken according to the current state; the execution action output by each agent's action network is then input into the bidirectional long short-term memory neural network BilSTM, and the bidirectional long short-term memory neural network BilSTM is used to memorize the agent's execution actions and strategies each time, and finally outputs the next execution action to be performed by each agent (i.e., the number of the database dynamically scheduled by each agent and the range adjustment factor Y that is predicted to be added or subtracted for the second rendering data range). Since each agent learns a Q value separately, each agent can design any reward that belongs to it.

[0110] In the process of dynamic data scheduling, each agent needs to use the observation-action information of other agents, which means that observation sharing is carried out during the training process.

[0111] Q evaluates the objective function of the network y i as follows:

[0112]

[0113] Where t represents the time step, s t , a t They represent the state and action of the agent at time t, γ represents the reward discount coefficient, r represents the reward, α represents the learning rate, θ represents the network parameters, and j represents different goals, where 1 represents the database called, 2 represents the range adjustment factor, represents the minimum value of the network parameters of j targets; π represents the strategy, and φ represents the feature extraction function of the state.

[0114] Among them, r(s t , a t )=(p true -p pred ) 2 , p true 、p pred They represent the agent in state s t Next, perform action a t The true and predicted probabilities.

[0115] The policy network is based on the current state s i Output takes action a i The probability distribution formula is as follows:

[0116] π φ(a t |s t )=tanh(μ φ (s t )+σ φ (s t )⊙∈) (2);

[0117] Among them, ∈ represents the exploration coefficient, tanh represents the activation function, μ and σ represent the two fully connected layer functions of reinforcement learning respectively; among them, μ φ (s t ) means to change the current state s i Input fully connected layer, σ φ (s t ) means to change the current state s i Input another fully connected layer, the output of the fully connected layer is S t =ws t +b, w is the weight of the neuron, and b is the bias value.

[0118] The formula for the action-value function (which can also be described as the state-action-value function) is as follows:

[0119]

[0120] in, Under given condition d, the next state s t+1 , such as the number of databases that the agent chooses to call, or the expected range adjustment factor of the prediction. π (s t+1 ) is the next state v π (s t+1 )’s value function.

[0121]

[0122] Among them, E represents the mathematical expectation, which is used to calculate the average value of the cumulative reward, t represents the time step, S0 represents the initial state, r i (s t ,a 1,t ...a n,t ) represents the reward obtained by agent i (i∈n, n represents the number of agents) at time step t, which depends on the current global state s t The action a taken at this time and space 1，t , ..., a n，t .

[0123] Finally, the probability of each agent performing the predicted action is:

[0124] P(s′|s,a1,...,a n ,π1,...,πn )=

[0125] P(s′|s,a1,...,a n ) = P(s′|s,a1,...,a n ,π′1,...,π′ n ) (5);

[0126] Among them, a n represents the action of the nth agent, and π′ represents the nth strategy executed.

[0127] The network structure of the multi-agent reinforcement learning model is as follows: Figure 3 As shown, the predicted actions and probabilities of each agent can be obtained through the multi-agent reinforcement learning model, and the predicted action output corresponding to the maximum probability is selected for fine-tuning the subsequent first rendering data range.

[0128] The embodiment of the present invention utilizes a multi-agent learning algorithm to dynamically schedule it, shortening the dynamic response time of data and improving the efficiency of rendering GIS data on the user side. At the same time, it can also avoid the phenomenon that too little node data is scheduled, resulting in incomplete GIS data rendering on the user side, leading to omission of rendering IoT card information, and the user side requesting the server to schedule data again for a long time. It can also avoid too much node dynamic scheduling data, which will result in too much IoT card data being returned to the user side for rendering, too long data scheduling time, and the user side rendering of IoT card GIS data is too slow, resulting in freezes, long loading times, and other phenomena.

[0129] For example, the server and the client are set up independently. Figure 4 The schematic diagram of the architecture of the IoT card data rendering system used for dynamic scheduling and rendering of IoT card GIS data is shown. The user end sends a data request related to data rendering to the server end. Based on the data request, the server end dynamically schedules the database through multiple intelligent agents to query the required candidate GIS data and responds to the user end with data. The user end renders the data based on the query results.

[0130] The embodiment of the present invention uses 5G base stations and IoT cards to transmit GIS data in real time, stores massive IoT card GIS data at the provincial and municipal levels through regional features, and uses GeoHash coding to encode the IoT card GIS data into a one-dimensional string and use it as the 2D geographic index of the database IoT card. Then, a multi-agent reinforcement learning algorithm is used to dynamically schedule the database. Each agent will dynamically operate the database. It dynamically learns the GIS data range that needs to be scheduled for the IoT card through the map features collected and rendered by the front end, the screen display ratio, the latitude and longitude coordinates within the current field of view, and the manual operation loading display features, thereby improving the efficiency of loading and rendering data on the user side, reducing the number of requests, and improving the data response rate.

[0131] See also Figure 5 , Figure 5 The embodiment of the present invention provides a structural block diagram of an IoT card data rendering device, wherein the IoT card data rendering device includes:

[0132] A data range determination module 11, used to determine a first rendering data range according to a current field of view of the map;

[0133] A data range fine-tuning module 12, configured to fine-tune the first rendering data range according to a range adjustment factor to obtain a second rendering data range; wherein the range adjustment factor is obtained by performing multi-agent reinforcement learning on a plurality of databases storing GIS data of IoT cards;

[0134] A data query module 13, used for querying candidate GIS data within the second rendering data range;

[0135] The data rendering module 14 is used to render the candidate GIS data in the current field of view.

[0136] In an optional embodiment, the device further includes:

[0137] Data collection module, used to collect GIS data and basic data from multiple IoT cards;

[0138] A database construction module, used to divide the GIS data and basic data of each IoT card into regions according to the location of each IoT card, and construct multiple databases;

[0139] The geographic index generation module is used to compress and encode the GIS data of each IoT card to obtain the geographic index of each IoT card in the corresponding database.

[0140] In an optional embodiment, the geographic index generation module includes:

[0141] An encoding unit, used for performing GeoHash encoding on each of the GIS data to obtain a first position code of each of the GIS data;

[0142] A code mapping unit, used for performing base32 code mapping on each of the first position codes to obtain a second position code of the corresponding GIS data;

[0143] The coding prefix acquisition unit is used to acquire the coding prefix of each of the second coding positions to obtain the geographic index of the corresponding GIS data.

[0144] In an optional embodiment, the data range determination module 11 includes:

[0145] A field of view expansion unit, used to expand the current field of view area;

[0146] The rendering data range determination unit is used to determine the first rendering data range according to the coordinate point positions corresponding to all GIS data in the current field of view after expansion.

[0147] In an optional embodiment, the device further includes:

[0148] A state-action set construction module is used to construct a set of environmental states in which the agent is located and a set of actions taken by the agent;

[0149] The range adjustment factor acquisition module is used to obtain the range adjustment factor by constructing a multi-agent reinforcement learning model based on a flexible action-evaluation algorithm according to the state set and the action set.

[0150] In an optional embodiment, the state set includes environmental information composed of observations of all intelligent agents and data dynamic scheduling characteristics of the intelligent agents; the environmental information includes: the data range of all databases during scheduling; the data dynamic scheduling characteristics include: the longitude and latitude coordinates of the top, bottom, left, right and middle of the current field of view, the current scale size of the map, the current screen display scale size, the spatial scope of the current viewport data retrieval, the map scale feature of moving or reducing, and the mouse movement rate characteristics in all directions.

[0151] In an optional embodiment, the action set is a set including a binary representation of the longitude and latitude of a GeoHash string within a data range of a corresponding database when querying any database, a number of a scheduled database, and a range adjustment factor of the corresponding database query.

[0152] In an optional embodiment, the multi-agent reinforcement learning model includes several Q evaluation networks constructed based on the action value function after the agent performs an action in the action set under the state in the environment state set, several action networks that perform strategy scheduling to execute actions based on the state in the environment state set, a strategy network for outputting the probability distribution of taking actions according to the current state, and a bidirectional long short-term memory neural network for memorizing the actions and strategies performed by the agent each time and predicting the next action to be performed by the agent; wherein each agent learns a Q evaluation network and an action network accordingly.

[0153] In an optional embodiment, the range adjustment factor obtaining module includes:

[0154] An action value function calculation unit, used to input the environmental information in the state set as the state, and the longitude and latitude binary representation of the GeoHash string within the data range of the corresponding database when querying any database in the action set as the action into the Q evaluation network of each intelligent agent, and calculate the action value function of each intelligent agent;

[0155] An execution action acquisition unit, used to input the data dynamic scheduling feature in the state set as the current state into the input of the action network of each agent, and obtain the execution action output by each agent based on the action value function of each agent;

[0156] A strategy calculation unit, used to calculate the probability distribution of taking a corresponding execution action in the current state through the strategy network, and obtain a strategy corresponding to the corresponding execution action;

[0157] The execution action prediction unit is used to obtain the predicted action of each intelligent agent based on memorizing the execution actions and strategies performed by each intelligent agent each time through a bidirectional long short-term memory neural network; wherein the predicted action includes the number of the scheduled database and the range adjustment factor of the corresponding database query.

[0158] In an optional embodiment, the data rendering module 14 includes:

[0159] A GIS data screening unit, used for traversing the current visual field after expansion, and screening out target GIS data within the current visual field after expansion from the candidate GIS data;

[0160] A rendering unit, used for rendering with the target GIS data as a node to obtain rendering data of the corresponding target GIS data;

[0161] The rendering display unit is used to display the rendering data in the current field of view and hide the rendering data in the area outside the current field of view.

[0162] It should be noted that the working process of each module in the IoT card data rendering device described in the embodiment of the present invention can refer to the working process of the IoT card data rendering method described in the above embodiment, and the technical effect achieved is the same as the IoT card data rendering method described in the above embodiment, which will not be repeated here.

[0163] See also Figure 6 , Figure 6: is a structural block diagram of an IoT card data rendering device provided by an embodiment of the present invention. The IoT card data rendering device includes a processor 21, a memory 31, and a computer program stored in the memory 31 and executable on the processor 21. When the processor 21 executes the computer program, the steps in the above-mentioned embodiments of the IoT card data rendering method are implemented, such as steps S11 to S14.

[0164] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory 31 and executed by the processor 21 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program in the IoT card data rendering device.

[0165] The IoT card data rendering device may include, but is not limited to, a processor 21 and a memory 31. Those skilled in the art may understand that the schematic diagram is only an example of the IoT card data rendering device and does not constitute a limitation on the IoT card data rendering device. The IoT card data rendering device may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the IoT card data rendering device may also include input and output devices, network access devices, buses, etc.

[0166] 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 any conventional processor, etc. The processor 21 is the control center of the IoT card data rendering device, and uses various interfaces and lines to connect various parts of the entire IoT card data rendering device.

[0167] The memory 31 can be used to store the computer program and / or module. The processor 21 realizes various functions of the IoT card data rendering device by running or executing the computer program and / or module stored in the memory 31 and calling the data stored in the memory 31. The memory 31 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.); the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.). In addition, the memory 31 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.

[0168] Wherein, if the module / unit integrated in the IoT card data rendering device 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 21. 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), electric carrier signal, telecommunication signal and software distribution medium, etc.

[0169] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0170] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, many improvements and modifications can be made 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 method for rendering data of an Internet of Things card, characterized in that: include: Determine a first rendering data range according to a current field of view of the map; The first rendering data range is fine-tuned according to a range adjustment factor to obtain a second rendering data range; wherein the range adjustment factor is obtained by performing multi-agent reinforcement learning on a plurality of databases storing GIS data of IoT cards; Querying candidate GIS data within the range of the second rendering data; The candidate GIS data is rendered in the current field of view.

2. The method for rendering IoT card data according to claim 1, characterized in that: The method further comprises: Collect GIS data and basic data from multiple IoT cards; According to the location of each IoT card, the GIS data and basic data of each IoT card are divided into regions to build multiple databases; The GIS data of each IoT card is compressed and encoded to obtain the geographic index of each IoT card in the corresponding database.

3. The method for rendering IoT card data as claimed in claim 2, characterized in that: The step of compressing and encoding the GIS data of each IoT card to obtain a geographic index of each IoT card in a corresponding database includes: Performing GeoHash encoding on each of the GIS data to obtain a first location code of each of the GIS data; Perform base32 encoding mapping on each of the first location codes to obtain a second location code of the corresponding GIS data; The coding prefix of each of the second coding positions is obtained to obtain the geographic index of the corresponding GIS data.

4. The method for rendering IoT card data according to claim 1, characterized in that: The determining of the first rendering data range according to the current field of view of the map includes: Expand the current field of view; The first rendering data range is determined according to the coordinate point positions corresponding to all GIS data within the current field of view after expansion.

5. The method for rendering IoT card data according to claim 1, wherein: The calculation process of the range adjustment factor includes: Constructing a set of environmental states in which the agent is located and a set of actions taken by the agent; According to the state set and the action set, a multi-agent reinforcement learning model is constructed based on a flexible action-evaluation algorithm to obtain the range adjustment factor.

6. The method for rendering IoT card data as claimed in claim 5, characterized in that: The state set includes environmental information composed of observations of all agents and data dynamic scheduling characteristics of the agents; The environmental information includes: the data range of all databases during scheduling; the data dynamic scheduling characteristics include: the latitude and longitude coordinates of the top, bottom, left, right and middle of the current field of view, the current scale size of the map, the current screen display scale size, the spatial range of the current viewport data retrieval, the map scale feature of moving or reducing the map, and a collection of mouse movement rate features in all directions.

7. The method for rendering IoT card data as claimed in claim 5, characterized in that: The action set is a set including a binary representation of the longitude and latitude of a GeoHash string within a data range of a corresponding database when querying any database, a number of a scheduled database, and a range adjustment factor of the corresponding database query.

8. The method for rendering IoT card data as claimed in claim 5, characterized in that: The multi-agent reinforcement learning model includes several Q evaluation networks constructed based on the action value function after the agent performs the actions in the action set in the state in the environment state set, several action networks that perform strategy scheduling to execute actions based on the state in the environment state set, a strategy network for outputting the probability distribution of taking actions according to the current state, and a bidirectional long short-term memory neural network for memorizing the actions and strategies performed by the agent each time and predicting the next action to be performed by the agent; wherein each agent learns a Q evaluation network and an action network accordingly.

9. The method for rendering IoT card data as claimed in claim 8, characterized in that: The step of obtaining the range adjustment factor by constructing a multi-agent reinforcement learning model based on a flexible action-evaluation algorithm according to the state set and the action set includes: The environmental information in the state set is used as the state, and the binary representation of the longitude and latitude of the GeoHash string within the data range of the corresponding database when querying any database in the action set is used as the action to input into the Q evaluation network of each intelligent agent, and the action value function of each intelligent agent is calculated; The data dynamic scheduling feature in the state set is input into the action network of each agent as the current state, and the execution action output by each agent is obtained based on the action value function of each agent; By using the strategy network, the probability distribution of taking corresponding execution actions in the current state is calculated to obtain the strategy corresponding to the corresponding execution action; Based on memorizing the execution actions and strategies of each agent each time through a bidirectional long short-term memory neural network, the predicted actions of each agent are obtained; wherein the predicted actions include the number of the scheduled database and the range adjustment factor of the corresponding database query.

10. The method for rendering IoT card data according to claim 4, characterized in that: The rendering of the candidate GIS data in the current field of view includes: Traversing the current visual field after expansion, and selecting target GIS data within the current visual field after expansion from the candidate GIS data; Rendering is performed using the target GIS data as a node to obtain rendering data of the corresponding target GIS data; The rendering data in the current field of view is displayed, and the rendering data in the area outside the current field of view is hidden.

11. A data rendering device for an Internet of Things card, characterized in that: include: A data range determination module, used to determine a first rendering data range according to a current field of view of the map; A data range fine-tuning module, used to fine-tune the first rendering data range according to a range adjustment factor to obtain a second rendering data range; wherein the range adjustment factor is obtained by performing multi-agent reinforcement learning on a plurality of databases storing GIS data of IoT cards; A data query module, used for querying candidate GIS data within the range of the second rendering data; The data rendering module is used to render the candidate GIS data in the current field of view.

12. An IoT card data rendering device, characterized in that: include: A processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for rendering IoT card data as described in any one of claims 1 to 10 is implemented.

13. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the IoT card data rendering method according to any one of claims 1 to 10.

14. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the method for rendering IoT card data as described in any one of claims 1 to 10 is implemented.

Images