RedCap terminal multi-mode communication switching method based on electric red-ong
By initializing performance index weights and reinforcement learning modules on the RedCap terminal, the priority of the operator's network is dynamically adjusted, solving the problem of inefficient network switching in multimodal communication of power networks, and achieving more efficient network switching and lower energy consumption.
Patent Information
- Application Number
- CN202511135965.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-14
- Publication Date
- 2025-11-11
AI Technical Summary
In a multi-carrier network environment, the network handover of power networks is not efficient enough. Traditional handover algorithms based on signal strength cannot meet the needs of diverse scenarios in power networks, resulting in frequent handovers and network congestion.
Using RedCap terminals based on Elec-Tech, the system initializes the performance index weights of service types, uses a reinforcement learning module for iterative learning, dynamically adjusts the priority score and normalized weights of the operator's network, and selects the optimal network for switching based on real-time network performance indicators.
The network handover strategy has been optimized, reducing device power consumption, decreasing user congestion rate and average number of handovers, improving network performance and user experience, and adapting to complex and ever-changing network environments and diverse business needs.
Smart Images
Figure CN120935689A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of communication technology, and in particular relates to a multi-modal communication switching method for RedCap terminals based on Elec-Tech. Background Technology
[0002] With the digitalization and intelligentization of the power industry, power companies' network needs are gradually shifting from traditional single transmission to multi-dimensional, multi-layered, and high-efficiency communication networks. The introduction of 5G technology provides the power industry with more advanced communication solutions. A power virtual private network (VPN) refers to a dedicated communication network built by power companies to meet their specific communication needs. Multimodal communication technology, by integrating multiple network interfaces (such as multi-carrier cellular networks, Wi-Fi, and power line carrier), enables terminal devices to flexibly select the optimal communication path. In power VPNs, multimodal communication capability, i.e., multi-carrier network access, has become a key requirement. In power VPNs, the power HarmonyOS terminal, as a key access device, plays a crucial role. The power HarmonyOS terminal is a smart terminal device developed based on the HarmonyOS operating system, specifically customized for the power industry, possessing powerful communication capabilities, data processing capabilities, and security.
[0003] Traditional 5G terminals, due to their high bandwidth, multiple antennas, and complex modulation designs, suffer from high costs and power consumption, making it difficult to meet the large-scale deployment needs of low- and medium-speed IoT devices. While existing LPWA technologies (such as NB-IoT and LTE-M) are low-cost, they cannot support 5G-level reliability and latency-sensitive services. RedCap, as a "lightweight" 5G technology, has become an important evolution direction for cellular IoT due to its low power consumption, low complexity, high bandwidth, and low latency. However, in multi-operator network environments, E-Tech's RedCap terminals face network handover challenges. On the one hand, the power network environment is complex and variable, with significant differences in network coverage, bandwidth, and latency performance indicators across different regions and scenarios. On the other hand, the network resources of a single operator cannot fully meet the needs of the power network in diverse scenarios, especially in cases of network congestion or insufficient signal coverage. Furthermore, traditional RSS vertical handover algorithms based on signal strength only consider signal strength as a single indicator, failing to meet the diverse service needs of power networks. Moreover, the algorithm only considers short-term gains, switching immediately when a better network is detected. Frequent handovers can lead to significant congestion. Summary of the Invention
[0004] To address the aforementioned shortcomings in existing technologies, this invention provides a RedCap terminal multimodal communication switching method based on Elec-Tech, which solves the problem of inefficient multimodal communication network switching.
[0005] To achieve the above-mentioned objectives, the technical solution adopted by this invention is as follows: This invention provides a RedCap terminal multimodal communication handover method based on Elec-Tech, comprising the following steps: S1. Using the RedCap terminal, initialize the performance index weight values of each service type according to the service type, and dynamically adjust the performance index weight values based on the normalized performance index of each operator network to obtain the priority score and normalized weight of each operator network. S2. Iterative learning is performed using the reinforcement learning module of the RedCap terminal to obtain a trained reinforcement learning module; S3. Use the trained reinforcement learning module to obtain the Q value of each operator network in the real-time Q table, and combine it with the real-time normalized weights of the operator network to select and access the target operator network.
[0006] Further, step S1 includes the following steps: S11. Use the data processing unit of the Dianhong RedCap terminal to initialize the performance index weight values of each service type according to the service type; S12. Based on the normalized performance indicators of each operator's network, dynamically adjust the weight values of the performance indicators of each service type in each operator's network, and calculate the priority score of each operator's network. The calculation expression for the priority score of the operator's network is as follows: , in, Indicates the first Priority rating of each operator's network. The weight value represents the round-trip delay. Indicates the first Real-time round-trip latency of each operator's network. The weight value representing latency jitter. Indicates the first Real-time latency jitter of each operator's network. The weight value represents the bandwidth. Indicates the first Real-time bandwidth of each operator's network. The weight values representing the packet loss rate, Indicates the first Real-time packet loss rate of each operator's network; S14. Based on the priority scores of each operator's network, calculate the normalized weight of each operator's network. The formula for calculating the normalized weights of each operator's network is as follows: , in, Indicates the first Normalized weights for each operator's network Indicates the first One operator network, This represents the total number of operator networks, of which, .
[0007] Further, step S2 includes the following steps: S21. Use the RedCap terminal to obtain the normalized performance indicators of the currently accessed operator network as the network status of each accessed operator network, and input them into the reinforcement learning module. S22. Based on network status and The strategy utilizes a reinforcement learning module to explore probabilities of action selection. Randomly select operator network handover actions, and with The network operator with the highest current Q value is selected based on the corresponding probability. S23. Calculate the performance index return value of the operator network based on the network status corresponding to the operator network handover action. S24. Calculate the Q value based on the performance index return value of the operator's network and update the Q table; S25. Repeat S21~S24 several times until the number of iterations of learning after initialization is reached. Afterwards, a trained reinforcement learning module is obtained.
[0008] Furthermore, the calculation expression for the performance index return value of the operator network in S23 is as follows: , , in, This represents the performance metrics return value of the operator's network. This indicates the current network status of the carrier's network. This indicates the current carrier network handover action. Indicates the business type. This represents the weight value of latency jitter initialized based on the service type. Indicates based on and The reward function for latency jitter in a given carrier network. This represents the weight value of the round-trip latency initialized based on the service type. Indicates based on and The reward function for round-trip latency of a given carrier network. This represents the weighted value of the packet loss rate initialized based on the business type. Indicates based on and The reward function for a given packet loss rate in a carrier network. This represents the weight value of the bandwidth initialized based on the service type. Indicates based on and The payoff function for bandwidth in a given carrier network. Indicates based on and The first determined operator network The reward function for each performance metric Indicates the basis constant. This represents the minimum performance requirements for network services. This represents the maximum performance metric required by network services, where... .
[0009] Furthermore, the expression for calculating the Q value in S24 is as follows: , in, This represents the current network status of the carrier network and the Q value corresponding to the carrier network handover action. This indicates assignment. This represents the current performance metrics return value of the operator's network. Indicates the discount factor. This represents the Q-value corresponding to the operator network handover action that maximizes the Q-value under the new network condition. This indicates the new network state achieved after the current carrier network handover action. This indicates the operator network switching action that will be prioritized under the new network conditions.
[0010] Further, step S3 includes the following steps: S31. Use the trained reinforcement learning module to obtain the Q value of each operator's network in the real-time acquired Q table; S32. Multiply the Q value of each operator network in the real-time obtained Q table with the real-time normalized weight of the operator network to calculate the network selection score of the operator network. S33. Sort the network selection scores of the operators' networks from high to low to obtain the operator network selection ranking table. S34. Select the operator network with the highest network selection score in the operator network selection ranking table as the candidate network. S35. Determine whether the normalized performance indicators of the candidate network can meet the current business requirements. If yes, proceed to S37; otherwise, proceed to S36. S36. Remove the candidate network from the operator network selection sorting table. Repeat S34~S35 until the operator network selection sorting table is empty. Then, keep the current operator network connected and record the event, waiting for the next iteration update. S37. Select the candidate network as the target operator network and switch to the target operator network through the network interface of the Elec-Tech RedCap terminal.
[0011] The beneficial effects of this invention are as follows: This invention provides a multi-modal communication switching method for RedCap terminals based on Elec-Tech's RedCap architecture. It obtains hardware performance indicators by reading hardware parameters from the Elec-Tech RedCap terminal and optimizes the parameter size of the reinforcement learning algorithm model based on a dynamic power management mechanism, reducing the computational complexity of the model and thus reducing device energy consumption. This invention adopts a multi-dimensional state space definition, comprehensively considering key indicators such as network latency, jitter, bandwidth, packet loss rate, and service type, and quantifies them uniformly through standardized processing. This comprehensively reflects the actual network performance, avoiding misjudgments caused by a single indicator. When the network state does not change drastically, it can effectively reduce user blocking rate and average switching frequency. This invention improves network performance and user experience by prioritizing and normalizing the weights of different operator networks. It meets the varying network performance requirements of different types of services in real-world applications. By dynamically adjusting the weights of each network based on real-time collected network performance metrics, it better adapts to complex and ever-changing network environments and diverse service needs, optimizing network switching strategies. Furthermore, this invention enables the RedCap terminal to flexibly switch between multiple networks in multi-operator access scenarios, dynamically selecting the optimal network based on service requirements and network performance. By combining intelligent monitoring with multimodal communication, it significantly enhances the adaptability and flexibility of the power virtual private network in complex environments.
[0012] Other advantages of the present invention will be analyzed in more detail in the following embodiments. Attached Figure Description
[0013] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0014] Figure 1 This is a flowchart illustrating the steps of a RedCap terminal multimodal communication switching method based on Elec-Tech in an embodiment of the present invention. Detailed Implementation
[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.
[0016] RedCap (Reduced Capability) is a lightweight 5G technology standard defined by 3GPP. It achieves low power consumption and low cost by eliminating unnecessary functions, such as reducing the number of antennas and simplifying modulation methods, making it suitable for low-to-medium speed IoT scenarios. When used in power virtual private networks, E-Hong RedCap terminals support simultaneous access to multiple operator networks.
[0017] ICMP, or Internet Control Message Protocol, is used for network diagnostics.
[0018] DRB (Data Radio Bearer) is a logical channel used in 5G to transmit user data, which determines the number of data streams that a terminal can process simultaneously.
[0019] like Figure 1 As shown, in one embodiment of the present invention, the present invention provides a multi-modal communication handover method for RedCap terminals based on Elec-Tech, comprising the following steps: S01. Start the E-Hong RedCap terminal and initialize the reinforcement learning parameters and Q-table through the baseband processing unit; S02. Read the memory resources and remaining battery power of the RedCap terminal and dynamically adjust the reinforcement learning parameters to obtain real-time reinforcement learning parameters; The reinforcement learning parameters in S01 include learning factors. Discount factor Number of learning iterations And the probability of action selection exploration In this embodiment, the number of iterations of learning after initialization... The number of iterations is 1000. The value in the initialized Q-table is 0. The initialized reinforcement learning parameters and Q-table are stored in the storage unit for the reinforcement learning module to perform iterative learning.
[0020] S02 includes the following steps: S021. Read the memory resources and remaining battery power of the Dianhong RedCap terminal; S022. If the memory resource utilization rate of the Dianhong RedCap terminal exceeds the preset utilization rate threshold, then the learning factor will be reduced. And reduce the number of learning iterations. Reduce the frequency of Q table updates; In this embodiment, the occupancy threshold refers to a DRB cache occupancy rate of 70%, which is achieved by reducing the learning factor. This reduces computational and storage pressure by decreasing the number of iterations and the frequency of Q-table updates, thus limiting memory usage. S023. If the memory resource utilization rate of the Dianhong RedCap terminal is lower than the preset utilization rate threshold, then the learning factor is adjusted. Maintain the value above the learning factor threshold, and increase the Q-table update frequency and the number of learning iterations. ; In this embodiment, the learning factor threshold is 0.3, by making the learning factor... To maintain a value greater than the learning factor threshold, increase the Q-table update frequency and the number of learning iterations. It can accelerate convergence and fully explore the state space; S024. If the remaining battery power of the Dianhong RedCap terminal is greater than the remaining battery power threshold, then the discount factor is adjusted. Keep the probability of action selection above the discount factor threshold and allow the action to be explored. The probability threshold for action selection exploration is greater than the threshold. In this embodiment, the discount factor threshold is 0.9, the probability threshold for action selection exploration is 0.9, and the remaining battery threshold is 50%. When the remaining battery is greater than 50%, a higher discount factor is maintained. Number of learning iterations It can encourage the exploration of new strategies and focus on long-term strategy optimization; S025. If the remaining battery power of the Dianhong RedCap terminal is less than the remaining battery power threshold, then the discount factor will be reduced. And the probability of action selection exploration ; S026. Use the reinforcement learning parameters that have been dynamically adjusted based on S021~S025 at the current time as the real-time reinforcement learning parameters.
[0021] In this embodiment, when the remaining battery power is less than 50%, the discount factor is reduced. And the probability of action selection exploration This allows reinforcement learning to be biased towards known policies, thus shortening the computational chain. In this embodiment, the real-time reinforcement learning parameters that change are stored in a storage unit for the reinforcement learning module to perform iterative learning.
[0022] S03. Using the Dianhong RedCap terminal, send a service request to the server through the network interface module, receive service data packets, and parse them to obtain the service type; S03 includes the following steps: S031. Use the RedCap terminal to send a service request to the server through the network interface module. S032. Receive the service data packets responded by the server through the same network interface module using the Dianhong RedCap terminal; S033. Based on the communication protocol and port number, use the data processing unit of the Dianhong RedCap terminal to parse the service data packets and obtain the service type corresponding to the service data packets.
[0023] In this embodiment, based on the various service types in the six stages of power generation, transmission, transformation, distribution, consumption, and integration described by the 5G power virtual private network, the service types can be roughly divided into data collection services, control services, and information services. In this scheme, the service type is determined by parsing the communication protocol and port number. For example, if the collected data packet uses the TCP protocol and the destination port number is 80, then the service type is determined to be a data collection service. If the data packet uses the UDP protocol and the destination port number is 53, then the service type is determined to be an information service.
[0024] S04. Use the RedCap terminal to collect and normalize the performance indicators of each operator's end-to-end network to obtain the normalized performance indicators. S04 includes the following steps: S041. Use the RedCap terminal to collect the performance indicators of each operator's end-to-end network. The performance indicators of the operator's end-to-end network include round-trip time, latency jitter, bandwidth and packet loss rate. S042. Measure round-trip delay, delay jitter, and bandwidth using the baseband processing unit of the Dianhong RedCap terminal; In this embodiment, round-trip time refers to the total time it takes for data to travel from the sending end to the receiving end and back to the sending end. This time is determined by the baseband processing unit according to the ICMP protocol. The command measures the round-trip time, and the terminal sends an ICMP protocol signal to the server. Message, and record the sending timestamp The server received Reply to message using the ICMP protocol The baseband processing unit records the received timestamp. Then the round-trip time delay ; Latency jitter is calculated by measuring the latency difference between adjacent data packets and averaging the results. Bandwidth is calculated by averaging the time intervals between consecutive data packets, and the expression for bandwidth calculation is as follows: , in, Indicates bandwidth. Indicates the number of bits in the data packet. This indicates the average interval between adjacent data packets.
[0025] S043. Calculate the packet loss rate using the data processing unit of the Dianhong RedCap terminal; The expression for calculating the packet loss rate is as follows: , in, Indicates packet loss rate. Indicates the number of data packets received. Indicates the number of data packets sent; S044. Perform nonlinear normalization on round-trip delay, delay jitter, and packet loss rate, and perform linear normalization on bandwidth to obtain normalized round-trip delay, delay jitter, bandwidth, and packet loss rate, which are used as normalized performance indicators.
[0026] In this embodiment, the normalized performance indicators are stored in the storage unit for dynamic adjustment of operator network weight values, operator network priority scoring, calculation of operator network normalized weights, and iterative learning by the reinforcement learning module.
[0027] S1. Using the RedCap terminal, initialize the performance index weight values of each service type according to the service type, and dynamically adjust the performance index weight values based on the normalized performance index of each operator network to obtain the priority score and normalized weight of each operator network. S1 includes the following steps: S11. Use the data processing unit of the Dianhong RedCap terminal to initialize the performance index weight values of each service type according to the service type; In this embodiment, data acquisition services have high requirements for bandwidth and packet loss rate, and data integrity is very important, such as high-definition video surveillance and drone inspection; control services typically have extremely high requirements for latency and reliability, such as distribution network differential protection and distribution automation; information services usually involve a large amount of data transmission and real-time interaction, and have high requirements for bandwidth and latency, such as smart construction sites and smart navigation. The performance indicators required for different service types vary depending on the specific service scenario. In practice, the initialization weights can be adjusted according to the needs of the service scenario and the service type. When considering only the service type, the initialization performance indicator weight values corresponding to each service type are shown in Table 1: Table 1 Initial weight values for performance metrics of different business types S12. Based on the normalized performance indicators of each operator's network, dynamically adjust the weight values of the performance indicators of each service type in each operator's network, and calculate the priority score of each operator's network. In this embodiment, if the bandwidth of a certain network in the operator's network suddenly drops based on the normalized bandwidth, the weight of that network is reduced, which can effectively reduce the dependence on that network.
[0028] The calculation expression for the priority score of the operator's network is as follows: , in, Indicates the first Priority rating of each operator's network. The weight value represents the round-trip delay. Indicates the first Real-time round-trip latency of each operator's network. The weight value representing latency jitter. Indicates the first Real-time latency jitter of each operator's network. The weight value represents the bandwidth. Indicates the first Real-time bandwidth of each operator's network The weight values representing the packet loss rate Indicates the first Real-time packet loss rate of each operator's network; S14. Based on the priority scores of each operator's network, calculate the normalized weight of each operator's network. The formula for calculating the normalized weights of each operator's network is as follows: , in, Indicates the first Normalized weights for each operator's network Indicates the first One operator network, This represents the total number of operator networks, of which, .
[0029] S2. Iterative learning is performed using the reinforcement learning module of the RedCap terminal to obtain a trained reinforcement learning module; S2 includes the following steps: S21. Use the RedCap terminal to obtain the normalized performance indicators of the currently accessed operator network as the network status of each accessed operator network, and input them into the reinforcement learning module. S22. Based on network status and The strategy utilizes a reinforcement learning module to explore probabilities of action selection. Randomly select operator network handover actions, and with The network operator with the highest current Q value is selected based on the corresponding probability. In this scheme, the following is adopted: Strategy, exploring probabilities based on action selection Randomly select operator network handover actions, and with The corresponding probability selection of the operator network with the largest current Q value can effectively balance the discovery of new operator networks with the selection of the known best operator network.
[0030] S23. Calculate the performance index return value of the operator network based on the network status corresponding to the operator network handover action. The calculation formula for the performance index return value of the operator network in S23 is as follows: , , in, This represents the performance metrics return value of the operator's network. This indicates the current network status of the carrier's network. This indicates the current carrier network handover action. Indicates the business type. This represents the weight value of latency jitter initialized based on the service type. Indicates based on and The reward function for latency jitter in a given operator network. This represents the weight value of the round-trip latency initialized based on the service type. Indicates based on and The reward function for round-trip latency of a given carrier network. This represents the weighted value of the packet loss rate initialized based on the business type. Indicates based on and The reward function for a given packet loss rate in a carrier network. This represents the weight value of the bandwidth initialized based on the service type. Indicates based on and The payoff function for bandwidth in a given carrier network. Indicates based on and The first determined operator network The reward function for each performance metric Indicates the basis constant. This represents the minimum performance requirements for network services. This represents the maximum performance metric required by network services, where... .
[0031] In this embodiment, when The corresponding performance metric is latency jitter. The corresponding performance metric is round-trip latency. The corresponding performance metric is packet loss rate. The corresponding performance metric is bandwidth; when At that time, the reward function value is 1, and as... As the return function increases, its value decreases exponentially.
[0032] S24. Calculate the Q value based on the performance index return value of the operator's network and update the Q table; The expression for calculating the Q value in S24 is as follows: , in, This represents the current network status of the carrier network and the Q value corresponding to the carrier network handover action. This indicates assignment. This represents the current performance metrics return value of the operator's network. Indicates the discount factor. This represents the Q-value corresponding to the operator network handover action that maximizes the Q-value under the new network condition. This indicates the new network state achieved after the current carrier network handover action. This indicates the operator network switching action that will be prioritized under the new network conditions.
[0033] In this embodiment, the discount factor is used to weigh the importance of the current performance metric return value of the operator's network against the future performance metric return value of the operator's network, and its value range is [value range missing]. learning factor Used to represent the speed at which new information overwrites old information, with a value range of [value missing]. .
[0034] S25. Repeat S21~S24 several times until the number of iterations of learning after initialization is reached. After that, a trained reinforcement learning module is obtained.
[0035] S3. Use the trained reinforcement learning module to obtain the Q value of each operator network in the real-time Q table, and combine it with the real-time normalized weights of the operator network to select and access the target operator network.
[0036] S3 includes the following steps: S31. Use the trained reinforcement learning module to obtain the Q value of each operator's network in the real-time acquired Q table; S32. Multiply the Q value of each operator network in the real-time obtained Q table with the real-time normalized weight of the operator network to calculate the network selection score of the operator network. S33. Sort the network selection scores of the operators' networks from high to low to obtain the operator network selection ranking table. S34. Select the operator network with the highest network selection score in the operator network selection ranking table as the candidate network. S35. Determine whether the normalized performance indicators of the candidate network can meet the current business requirements. If yes, proceed to S37; otherwise, proceed to S36. S36. Remove the candidate network from the operator network selection sorting table. Repeat S34~S35 until the operator network selection sorting table is empty. Then, keep the current operator network connected and record the event, waiting for the next iteration update. S37. Select the candidate network as the target operator network and switch to the target operator network through the network interface of the Elec-Tech RedCap terminal.
[0037] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.
Claims
1. A multi-mode communication handover method for RedCap terminals based on Elec-Tech, characterized in that, Includes the following steps: S1. Using the RedCap terminal, initialize the performance index weight values of each service type according to the service type, and dynamically adjust the performance index weight values based on the normalized performance index of each operator network to obtain the priority score and normalized weight of each operator network. S2. Iterative learning is performed using the reinforcement learning module of the RedCap terminal to obtain a trained reinforcement learning module; S3. Use the trained reinforcement learning module to obtain the Q value of each operator network in the real-time Q table, and combine it with the real-time normalized weights of the operator network to select and access the target operator network.
2. The RedCap terminal multi-mode communication handover method based on Elec-Tech as described in claim 1, characterized in that, S1 includes the following steps: S11. Use the data processing unit of the Dianhong RedCap terminal to initialize the performance index weight values of each service type according to the service type; S12. Based on the normalized performance indicators of each operator's network, dynamically adjust the weight values of the performance indicators of each service type in each operator's network, and calculate the priority score of each operator's network. The priority score calculation formula for the operator network is as follows: ， in, Indicates the first Priority rating of each operator's network. The weight value represents the round-trip delay. Indicates the first Real-time round-trip latency of each operator's network. The weight value representing latency jitter. Indicates the first Real-time latency jitter of each operator's network. The weight value represents the bandwidth. Indicates the first Real-time bandwidth of each operator's network. The weight values representing the packet loss rate, Indicates the first Real-time packet loss rate of each operator's network; S14. Based on the priority scores of each operator's network, calculate the normalized weight of each operator's network. The formula for calculating the normalized weights of each operator's network is as follows: ， in, Indicates the first Normalized weights for each operator's network Indicates the first One operator network, This represents the total number of operator networks, of which, .
3. The RedCap terminal multimodal communication handover method based on Elec-Tech as described in claim 2, characterized in that, S2 includes the following steps: S21. Use the RedCap terminal to obtain the normalized performance indicators of the currently accessed operator network as the network status of each accessed operator network, and input them into the reinforcement learning module. S22. Based on network status and The strategy utilizes a reinforcement learning module to explore probabilities of action selection. Randomly select operator network handover actions, and with The corresponding probability is to select the operator network with the highest current Q value; S23. Calculate the performance index return value of the operator network based on the network status corresponding to the operator network handover action. S24. Calculate the Q value based on the performance index return value of the operator's network and update the Q table; S25. Repeat S21~S24 several times until the number of iterations of learning after initialization is reached. Afterwards, a trained reinforcement learning module is obtained.
4. The RedCap terminal multimodal communication handover method based on Elec-Tech as described in claim 3, characterized in that, The calculation formula for the performance index return value of the operator network in S23 is as follows: ， ， in, This represents the performance metrics return value of the operator's network. This indicates the current network status of the carrier's network. This indicates the current carrier network handover action. Indicates the business type. This represents the weight value of latency jitter initialized based on the service type. Indicates based on and The reward function for latency jitter in a given carrier network. This represents the weight value of the round-trip latency initialized based on the service type. Indicates based on and The reward function for round-trip latency of a given carrier network. This represents the weighted value of the packet loss rate initialized based on the business type. Indicates based on and The reward function for a given packet loss rate in a carrier network. This represents the weight value of the bandwidth initialized based on the service type. Indicates based on and The payoff function for bandwidth in a given carrier network. Indicates based on and The first determined operator network The reward function for each performance metric Indicates the basis constant. This represents the minimum performance requirements for network services. This represents the maximum performance metric required by network services, where... .
5. The RedCap terminal multimodal communication handover method based on Elec-Tech as described in claim 4, characterized in that, The expression for calculating the Q value in S24 is as follows: ， in, This represents the current network status of the carrier network and the Q value corresponding to the carrier network handover action. This indicates assignment. This represents the current performance metrics return value of the operator's network. Indicates the discount factor. This represents the Q-value corresponding to the operator network handover action that maximizes the Q-value under the new network condition. This indicates the new network state achieved after the current carrier network handover action. This indicates the operator network switching action that will be prioritized under the new network conditions.
6. The RedCap terminal multimodal communication handover method based on Elec-Tech as described in claim 3, characterized in that, S3 includes the following steps: S31. Use the trained reinforcement learning module to obtain the Q value of each operator's network in the real-time acquired Q table; S32. Multiply the Q value of each operator network in the real-time obtained Q table with the real-time normalized weight of the operator network to calculate the network selection score of the operator network. S33. Sort the network selection scores of the operators' networks from high to low to obtain the operator network selection ranking table. S34. Select the operator network with the highest network selection score in the operator network selection ranking table as the candidate network. S35. Determine whether the normalized performance indicators of the candidate network can meet the current business requirements. If yes, proceed to S37; otherwise, proceed to S36. S36. Remove the candidate network from the operator network selection sorting table. Repeat S34~S35 until the operator network selection sorting table is empty. Then, keep the current operator network connected and record the event, waiting for the next iteration update. S37. Select the candidate network as the target operator network and switch to the target operator network through the network interface of the Elec-Tech RedCap terminal.
7. The RedCap terminal multimodal communication handover method based on Elec-Tech as described in claim 1, characterized in that, It also includes the following steps: S01. Start the E-Mobile RedCap terminal and initialize the reinforcement learning parameters and Q-table through the baseband processing unit. The reinforcement learning parameters include the learning factor. Discount factor Number of learning iterations And the probability of action selection exploration ; S02. Read the memory resources and remaining battery power of the RedCap terminal and dynamically adjust the reinforcement learning parameters to obtain real-time reinforcement learning parameters; S03. Using the Dianhong RedCap terminal, send a service request to the server through the network interface module, receive service data packets, and parse them to obtain the service type; S04. Use the RedCap terminal to collect and normalize the performance indicators of each operator's end-to-end network to obtain the normalized performance indicators.
8. The RedCap terminal multimodal communication handover method based on Elec-Tech as described in claim 7, characterized in that, S02 includes the following steps: S021. Read the memory resources and remaining battery power of the Dianhong RedCap terminal; S022. If the memory resource utilization rate of the Dianhong RedCap terminal exceeds the preset utilization rate threshold, then the learning factor will be reduced. And reduce the number of learning iterations. Reduce the update frequency of the Q table until; S023. If the memory resource utilization rate of the Dianhong RedCap terminal is lower than the preset utilization rate threshold, then the learning factor is adjusted. Keep the learning factor above the threshold and increase the Q-table update frequency and the number of learning iterations. ; S024. If the remaining battery power of the Dianhong RedCap terminal is greater than the remaining battery power threshold, then the discount factor is adjusted. Keep the probability of action selection above the discount factor threshold and allow the action to be explored. The probability threshold for action selection exploration is greater than the threshold. S025. If the remaining battery power of the Dianhong RedCap terminal is less than the remaining battery power threshold, then the discount factor will be reduced. And the probability of action selection exploration ; S026. Use the reinforcement learning parameters that have been dynamically adjusted based on S021~S025 at the current time as the real-time reinforcement learning parameters.
9. The RedCap terminal multi-mode communication handover method based on Elec-Tech as described in claim 8, characterized in that, S03 includes the following steps: S031. Use the RedCap terminal to send a service request to the server through the network interface module. S032. Receive the service data packets responded by the server through the same network interface module using the Dianhong RedCap terminal; S033. Based on the communication protocol and port number, use the data processing unit of the Dianhong RedCap terminal to parse the service data packets and obtain the service type corresponding to the service data packets.
10. The RedCap terminal multimodal communication handover method based on Elec-Tech as described in claim 9, characterized in that, S04 includes the following steps: S041. Use the RedCap terminal to collect the performance indicators of each operator's end-to-end network. The performance indicators of the operator's end-to-end network include round-trip time, latency jitter, bandwidth and packet loss rate. S042. Measure round-trip delay, delay jitter, and bandwidth using the baseband processing unit of the Dianhong RedCap terminal; S043. Calculate the packet loss rate using the data processing unit of the Dianhong RedCap terminal; S044. Perform nonlinear normalization on round-trip delay, delay jitter, and packet loss rate, and perform linear normalization on bandwidth to obtain normalized round-trip delay, delay jitter, bandwidth, and packet loss rate, which are used as normalized performance indicators.