Hierarchical sleep mode management in base station

Abstract

A method may include determining, by a system, a first sleep mode action for a first set of cellular base station signal metrics for a cellular base station for a first time period, wherein the first set of cellular base station signal metrics is determined based on a first layer of communications of the cellular base station. The method may also include determining, by the system, a second sleep mode action for a second set of cellular base station signal metrics for the cellular base station for a second time period. The method may also include determining, by the system, an arbitrated sleep mode action based on the first sleep mode action and the second sleep mode action. The method may also include sleeping, by the system, at least a portion of the cellular base station based on the arbitrated sleep mode action.

Description

Related applications

[0001] This application claims priority to U.S. non-provisional patent application serial number 18 / 364,438, filed August 2, 2023, entitled “HIERARCHICAL SLEEP MODEMANAGEMENT IN BASE STATIONS”, the entire contents of which are incorporated herein by reference. Background Technology

[0002] Base stations can communicate with user equipment to facilitate mobile or cellular network communication. In doing so, base stations can consume energy. Summary of the Invention

[0003] The following is a simplified overview of the disclosed subject matter to provide a basic understanding of some of the various embodiments. This overview is not a comprehensive summary of the various embodiments. It is neither intended to identify key or essential elements of the various embodiments nor to define the scope of the various embodiments. Its sole purpose is to introduce some concepts of this disclosure in a concise form as a prelude to the more detailed description that follows.

[0004] Example methods may include the system determining a first sleep mode action for a first set of cellular base station signal indicators for a cellular base station during a first time period, wherein the first set of cellular base station signal indicators is determined based on a first layer of communication of the cellular base station. The method may also include the system determining a second sleep mode action for a second set of cellular base station signal indicators for a cellular base station during a second time period, wherein the first set of cellular base station signal indicators differs from the second set of cellular base station signal indicators, wherein the second set of cellular base station signal indicators is determined based on a second layer of communication of the cellular base station, and wherein the first time period differs from the second time period. The method may also include the system determining an arbitrated sleep mode action based on the first and second sleep mode actions. The method may further include the system putting at least a portion of the cellular base station into sleep mode based on the arbitrated sleep mode action.

[0005] The example system can operate as follows: The system can determine a first sleep mode action relative to a first time period, applicable to a first set of cellular base station signal indicators, wherein the first set of cellular base station signal indicators is determined based on a first layer of cellular base station communication. The system can determine a second sleep mode action relative to a second time period, applicable to a second set of cellular base station signal indicators, wherein the second set of cellular base station signal indicators is determined based on a second layer of cellular base station communication. The system can determine a selected sleep mode action based on the first and second sleep mode actions. The system can put at least a portion of the cellular base station into sleep mode based on the selected sleep mode action.

[0006] Example non-transitory computer-readable media may include instructions that, in response to execution, cause a system including a processor to perform operations. These operations may include determining a corresponding sleep mode action for a base station device based on a corresponding set of base station signal indicators, wherein the corresponding set of base station signal indicators corresponds to a corresponding communication layer of the base station device. These operations may also include determining a sleep mode action based on the corresponding sleep mode action, thereby obtaining the determined sleep mode action. These operations may further include putting at least a portion of the base station device into sleep mode based on the determined sleep mode action. Attached Figure Description

[0007] Numerous embodiments, objects, and advantages of this model will become apparent upon consideration of the following specific implementations taken in conjunction with the accompanying drawings, in which similar reference numerals always refer to similar parts, and in the drawings:

[0008] Figure 1 An example system architecture that can facilitate hierarchical sleep pattern management according to embodiments of the present disclosure is shown;

[0009] Figure 2 An example sleep state of a base station based on different durations, which can facilitate hierarchical sleep pattern management in a base station according to an embodiment of the present disclosure, is shown.

[0010] Figure 3 An example system architecture for a hierarchical sleep mode management framework, which can facilitate hierarchical sleep mode management in a base station, is shown according to embodiments of the present disclosure.

[0011] Figure 4 An example process flow that can facilitate hierarchical sleep mode management in a base station according to embodiments of the present disclosure is shown;

[0012] Figure 5 A system architecture for a two-layer sleep control engine that can facilitate hierarchical sleep pattern management in a base station, according to embodiments of the present disclosure, is shown.

[0013] Figure 6 An example system architecture for a radio-controlled autonomous sleep mode without upper-layer interaction, according to embodiments of the present disclosure, is shown.

[0014] Figure 7 An example process flow that can facilitate hierarchical sleep mode management in a base station according to embodiments of the present disclosure is shown;

[0015] Figure 8 Another example process flow that can facilitate hierarchical sleep mode management in a base station according to embodiments of the present disclosure is shown;

[0016] Figure 9 Another example process flow that can facilitate hierarchical sleep mode management in a base station according to embodiments of the present disclosure is shown;

[0017] Figure 10 An example block diagram of a computer operable for performing embodiments of the present disclosure is shown. Detailed Implementation Overview

[0018] The examples described in this article may refer to optimizing aspects such as base station energy consumption. It can be understood that even if the improvement may not be optimal, this technique can still be applied to improve similar metrics. Similarly, when the examples describe the highest level (such as maximizing a metric), it can be understood that there may be examples where improvements or enhancements to that metric occur.

[0019] Fifth-generation (5G) and 5G+ (B5G) networks can enable features for end users, including augmented reality / virtual reality (AR / VR), autonomous driving, ultra-high speeds for connectivity of hundreds of devices, low-latency modes, and significantly lower latency and reliability in certain modes compared to previous networks. However, achieving such features can also come at the cost of high energy consumption. Therefore, energy efficiency in next-generation networks should be central to the design process, rather than an afterthought feature, to meet various climate goals and reduce the operating expenses (OPEX) of mobile network operators (MNOs). In some base station examples, the radio access network (RAN) can account for up to 70% of this energy consumption, making it a primary target for energy efficiency enhancements. True energy efficiency enhancements can only be achieved when data-driven learning presents energy efficiency opportunities and is properly combined with appropriate actuation within RAN components. The current complexity of RAN design makes deriving a general framework for achieving energy efficiency challenging, and according to this technology, meaningful energy efficiency enhancements may be achieved through cross-layer efforts.

[0020] Leveraging the design flexibility and high data availability offered by 5G New Radio (NR) base station designs, intelligent processing of such data can be systematically simplified to optimize base station performance based on various Key Performance Indicators (KPIs), thereby optimizing network operations. While mechanisms to avoid high-overhead signaling can be incorporated into the downlink / uplink (DL / UL) frame structure of 5G NR, these measures alone may not be sufficient to have a meaningful impact on reducing overall energy consumption, according to some previous approaches. Traditional base station architectures are designed to handle worst-case traffic scenarios with heavy loads. However, in practice, these heavy-load scenarios are not always prevalent, nor are they prevalent for all cell sites in a network deployment. To take advantage of these low-load scenarios, previous approaches could implement sleep states, potentially shutting down significant portions of the RAN to save power. However, in some previous approaches and earlier generations of mobile networks, it was possible to implement methods where cell sites were only operated in sleep mode (sometimes referred to as low-power mode) when traffic was low for extended periods, such as during the night (e.g., between 1 a.m. and 5 a.m.). In other approaches, numerous smaller sleep cycles can be used to handle micro-sleep cycles that are short in duration and may not require shutting down the entire radio frequency (RF) chain. Intelligent control mechanisms may be needed to take advantage of such low-power states in a timely manner. Due to the flexibility offered by 5G protocol design, the sleep duration of various power-consuming components within a base station can vary from sub-milliseconds to hundreds of seconds, depending on the activation / deactivation times of various components and the workload.

[0021] This technology can be implemented to facilitate a multi-level sleep pattern management framework as part of an intelligent control mechanism, where each level is configured to respond with signal statistics data associated with a given time scale at different time granularities. In some examples, each level is configured to collect data related to the optimal sleep control strategy reaching that level and run a decision engine that influences the future sleep pattern behavior of the base station. In some examples, lower-level sleep control engines can influence the sleep pattern behavior of only one engine, while in others, one or more levels provide sleep pattern recommendations that can be incorporated into the decision-making of higher levels. Example Architecture

[0022] Figure 1An example system architecture 100 according to embodiments of the present disclosure is shown that can facilitate hierarchical sleep mode management in a base station. Hierarchical sleep mode management can be a novel approach utilizing the present technology. In this approach, the hierarchy can emphasize the interaction between layers of base station communication, which is lacking in previous methods. Using previous methods, the interaction between layers of the hierarchy may exist or not for other purposes, but for sleep state management, the interaction between layers of the hierarchy does not exist.

[0023] System architecture 100 includes base station 102 and user equipment 104. Base station 102 further includes a hierarchical sleep mode management layer 106 and a hierarchical sleep mode management layer 108 within the base station components.

[0024] Each of base station 102 and / or user equipment 104 can utilize Figure 10 The computing environment 1000 is implemented in multiple parts. Base station 102 may typically include one or more antennas and electronic communication devices to facilitate network communication with user equipment 104. User equipment 104 may typically include computing devices used by end users to communicate with base station 102.

[0025] As part of communication with user equipment 104, base station 102 can consume energy. Base station 102 can reduce or conserve its energy consumption by implementing hierarchical sleep mode management. Hierarchical sleep mode management level 106 can include multiple levels that determine the sleep modes that base station 102 can enter, such as based on different indicators from different communication layers of base station 102, and such as when communicating with user equipment 104 (and other user equipment communicating with base station 102), it can operate in a lower power state (which may sometimes be referred to as a power state) than the current power state. Hierarchical sleep mode management 108 in the base station component can use information from hierarchical sleep mode management level 106 to implement sleep modes for base station 102.

[0026] In some examples, the communication layers of base station 102 may include a radio and physical layer (which may be referred to as layer 1), layer 2, and layer 3. The radio and physical layers can form the actual signals to be transmitted in the electrical domain. Specifically, the physical layer can use digital modules to perform encoding, modulation, and waveform generation functions, and then a digital-to-analog converter (DAC) feeds the modulated analog signal to the radio, which can up-convert the signal to the correct frequency range and apply appropriate gain to the signal to transmit it with the desired signal power using an antenna.

[0027] Layer 2 can be an upper layer relative to Layer 1 and includes a Media Access Control (MAC) layer and a Radio Link Control (RLC) layer, which can schedule data to be sent to various user equipment based on channel conditions and Quality of Service (QoS) protocols between different users.

[0028] Layer 3 can be another higher layer relative to Layer 2, and can control the allocation of resources (Radio Resource Management (RRM)) and aspects such as admission control, mobility management (handover to another base station) and coordination with other base stations to maintain optimal network operation.

[0029] In some examples, hierarchical sleep mode management 108 in the base station component can achieve... Figure 4 and / or Figures 7 to 9 The process flow consists of (multiple) parts to achieve hierarchical sleep mode management in the base station.

[0030] It can be understood that System Architecture 100 is an example system architecture for hierarchical sleep mode management in base stations, and there may be other system architectures that facilitate sleep mode management in base stations.

[0031] Figure 2 An example sleep state 200 of a base station based on different durations, according to embodiments of the present disclosure, is illustrated, which can facilitate hierarchical sleep mode management in a base station. In some examples, portions(s) of sleep state 200 may be used to implement… Figure 1 The system architecture consists of (multiple) parts to facilitate hierarchical sleep mode management in base stations.

[0032] Sleep state 200 includes sleep mode 0 202, sleep mode 1 204, sleep mode 2 206, sleep mode 3 208, and hierarchical sleep mode management 210 in the base station component (which can be similar to...). Figure 1 The base station component includes a hierarchical sleep mode management system (108). Sleep mode 0 (202) can be executed for a time interval t0, sleep mode 1 (204) can be executed for a time interval t1, sleep mode 2 (206) can be executed for a time interval t2, and sleep mode 3 (208) can be executed for a time interval t3. The sequence can be t0 ≤ t1 ≤ t2 ≤ t3.

[0033] Sleep mode 0 202 may include shutting down the power amplifier (PA) and some analog front-end (AFE). Sleep mode 1 204 may include shutting down the PA and the full AFE. Sleep mode 2 206 may include shutting down the PA, digital baseband, and AFE. Sleep mode 3 208 may include shutting down the PA, digital baseband, and AFE for a longer period than sleep mode 2 206.

[0034] To make timely use of idle periods in downlink (DL) / uplink (UL) transmissions, individual parts of the radio unit (RU) can be "shut down" to minimize the overall power consumption of the base station. Depending on the traffic load and activation / deactivation times of individual components, the base station can be placed in various low-power modes, which can be referred to as "sleep modes." However, the timescale of such sleep states can be very long and limited to Time of Day (ToD) and Day of the Week (DoW) modes, which have durations much longer than the frame durations in 5G NR.

[0035] Examples of this technology can be applied to, for example... Figure 1 The sleep modes are categorized as follows: Sleep Mode 0 202 (which may be referred to as SlpMode_0) represents the mode with the shortest duration (t0), and Sleep Mode 3 208 (which may be referred to as SlpMode_3) represents the longest duration (t3), with intermediate durations in between. The specific components that can enter sleep mode can depend on the deactivation and activation times of each of these components. For example, in an RF front-end, several components may be analog and therefore have capacitive behavior with finite non-zero times for charging / discharging them. For example, for a short duration lasting less than a subframe, the power amplifier (PA) can be turned off if there is no DL transmission (SlpMode_0).

[0036] One issue with previous methods may involve the lack of an optimal automatic dynamic sleep management component. Previous methods may lack a mechanism for optimally activating specific sleep states within the base station that could save the maximum average amount of energy during operational periods that explicitly consider base station activity factors. Furthermore, sleep state activation may only be considered at higher layers of the protocol stack, which may not provide a comprehensive understanding of the radio unit's power consumption. Therefore, the base station may be unable to take advantage of micro-sleep opportunities because the feedback loop for actuating sleep states based on service demands received from the user equipment (UE) on the uplink and transmitting relevant control information to the radio unit may be too long and require multiple cycles to formulate a strategy.

[0037] Another issue with previous methods may involve the relatively long network response times under low-traffic conditions, which allows for the use of low-power states when responding in an agile manner. Control mechanisms for binary sleep states in some previous methods may be too slow to respond to network events occurring at the subframe or frame level, and only to day-of-day, weekday-specific traffic patterns that exhibit relevant behavior spanning tens of minutes or hours. This, in turn, can mean that energy-saving features built into the base station cannot take advantage of opportunities lasting relatively short periods, such as tens of minutes, and in some examples, only follow rule-based shutdown operations lasting several hours. The base station may continue to operate in a high-power state during these short periods of low or no traffic and thus exhibit poor energy efficiency.

[0038] Previous approaches in this field addressed the macroscopic issue of base station on / off, such as in 4G-LTE / LTE-Advanced protocols, where base stations need to remain on for several subframes to enable UEs to synchronize and broadcast information. Furthermore, these previous approaches may not have considered energy efficiency as a critical design factor, given the much lower density of scenarios envisioned for 5G and beyond.

[0039] Figure 3 An example system architecture 300 for a hierarchical sleep mode management framework, according to embodiments of the present disclosure, is illustrated, which can facilitate hierarchical sleep mode management in a base station. In some examples, multiple portions of system architecture 300 may be used to implement... Figure 1 The system architecture consists of (multiple) parts to facilitate hierarchical sleep mode management in base stations.

[0040] The system architecture 300 includes: layer 3 302, layer 2 304, layer 1 306, arbitrator 308, RF data collection 310, service scheduler (MAC) 312, MCS selection 314 (open loop and closed loop), lightweight sleep control engine 316 (micro sleep), physical (PHY) and radio control 318, multi-DU data collection and ML prediction / RIC 320 (joint user association and handover control), long-term service prediction engine 322, scheduler buffer queue state 324, short-term service prediction engine 326, and instantaneous signal statistics at the PHY / RF module 328.

[0041] According to this technology, sleep prediction can occur at various time scales to flexibly take advantage of varying traffic loads. In some examples, depending on the statistical patterns utilized, sleep states can vary from sub-milliseconds to seconds or even minutes. Therefore, for micro-sleep strategies, the feedback loop used to actuate low-power states may need to be small compared to relatively long time scales, such as those for hourly patterns. Given these considerations, in some examples, a policy engine can be implemented that is capable of making sleep pattern decisions at various latency levels, and the data-driven decision-making is accurate. This can be used to facilitate hierarchical sleep pattern management, as described according to this technology. System architecture 300 depicts an example of hierarchical sleep patterns for a base station, with blocks that can assist in preprocessing and actuating energy efficiency enhancements based on hierarchical sleep patterns.

[0042] System architecture 300 depicts a multi-tiered sleep control architecture that facilitates hierarchical sleep mode management, where micro-sleep states can be controlled by instantaneous signal statistics collected from telemetry points in RUs (e.g., tier 1 306). Tier 2 mechanisms (e.g., tier 2 304) can then utilize long-term traffic forecasting, and tier 3 mechanisms (e.g., tier 3 302) can consider longer-term and wider-area statistics across multiple RUs and distributed units (DUs), treating user association and handover as feasible power strategies for the entire network. Thus, system architecture 300 can represent three distinct tiers that ingest data at different levels within the base station protocol stack. It is understandable that examples using more or fewer tiers as depicted in system architecture 300 may exist. The level directly interacting with the radio channel could be the RF component, and therefore the lowest level of data collection may occur within the RF component, typically including raw data ingested by the base station from the radio channel, which, after some basic preprocessing using analog and digital components, ensures proper signal integrity and results in signal levels within a useful dynamic range. Business forecasting can be used to initialize sleep states at the correct time. The forecasting engine can have various levels of complexity; in some examples, and at the radio level (labeled Level 1), it can be configured to discern signal energy from the aggregated power of subcarriers, which over time can help infer the percentage of Physical Resource Block (PRB) occupancy and allow the RF control engine to adjust the PA output power level accordingly. On the other hand, a Level 2 sleep mode engine can utilize scheduler policy information and, in some examples, influence scheduling to prioritize energy efficiency. Finally, Level 3 can be implemented in elements such as Radio Intelligent Controllers (RICs), which can accommodate multiple network-specific control layers based on Artificial Intelligence / Machine Learning (AI / ML).

[0043] In some examples, the policies recommended by each of these tiers may not be entirely consistent, and may even conflict with the scheduler's own policies (e.g., if the scheduler is using a variant of the proportional fair scheduler). An arbitrator (as shown in system architecture 300) can then be placed to prioritize policy recommendations based on the network operator's KPIs.

[0044] This technology can be implemented to facilitate a hierarchical sleep pattern framework. It can also facilitate a hierarchical AI / ML engine for sleep pattern management, capable of responding to network events and patterns across all timescales, from milliseconds to hourly patterns, including day of the day, day of the week, and geographically isolated events. While some previous methods have attempted to use binary sleep states, allowing radio units to either transmit at full power or remain dormant for long periods of relatively inactive activity, the success rate of these methods has likely been limited. Several reasons for this may exist, such as the lack of a data-driven policy engine that can (a) determine the optimal power state of a radio unit and (b) predict the optimal duration for which radio units and other RAN elements might be placed in sleep mode to ensure no KPI service level agreement (SLA) violations occur. The hierarchical sleep pattern framework based on this technology, and the methods initiated as part of these technologies through a combination of offline training and dynamic online learning, can ensure that the network can select appropriate sleep states, thereby optimizing energy consumption across various timescales. Example process flow

[0045] Figure 4 An example process flow 400 according to embodiments of the present disclosure is shown that can facilitate hierarchical sleep mode management in a base station. In some examples, one or more embodiments of process flow 400 can be implemented through... Figure 1 Layered sleep mode management in base station components 108 or Figure 10 The computing environment is 1000 to achieve this.

[0046] It is understood that the operational flow of process flow 400 is an example operational flow, and there may be embodiments that implement more or fewer operational flows than depicted, or embodiments that implement the depicted operational flows in a different order than depicted. In some examples, process flow 400 may be... Figure 7 Process flow 700 Figure 8 Process flow 800 and / or Figure 10 The process flow 900 is implemented by combining one or more embodiments of one or more process flows.

[0047] The process flow 400 includes service monitoring 402 (which includes short-term forecasting 404 and long-term forecasting 406), scheduling and MCS allocation 408, CSI reporting 410 (link adaptation), system parameters 412 (bandwidth, antenna port, MIMO mode based on active users, etc.), load-related energy consumption calculation 414, base station power consumption model 416, sleep mode management 418, and sleep mode decision 420.

[0048] This technique can be implemented to facilitate dynamic sequential processing within a hierarchical sleep mode. Process flow 400 describes, at a high level, the procedural steps for determining a specific sleep state for a given epoch or a series of epochs. In some examples, certain operational parameters may exist that are fixed when the base station is configured within the network and can be considered relatively static except where network attributes are significantly reconfigured. These parameters are referred to as system parameters in process flow 400 and, together with channel state information (CSI) reports received by the base station from the UE and the buffer queue states of the individual connected UEs, can form the input to the scheduler. In some examples, for routine operation, this can form a sufficient dataset for scheduling DL transmissions (408) for various UEs.

[0049] For energy-efficiency-focused scheduling, AI / ML-driven service prediction (402, which may include short-term prediction 404 and long-term prediction 406) can be utilized. The service prediction component can provide an estimate of subframe-based service demand in advance, so the scheduler may already have a nominal modulation and coding scheme (MCS) level pre-selected for that subframe. The difference between the predicted scheduling output and the actual output can be determined based on the difference between the service conditions considered by the AI / ML component and the service conditions reported to the base station in real time. Based on these predictions, the scheduler can pre-determine the energy efficiency status of the base station, as described below. In some examples, the scheduler can select the MCS level that provides the highest spectral efficiency while ensuring that the block error rate (BLER) target of the scheduled UE is met. This can be determined by the scheduling and MCS allocation blocks using various inputs. For example, it can be reported from the Channel Quality Indicator (CQI) received from the UE, and in the case of Time-Domain Duplex (TDD) bands, the correct MCS can be established using an uplink reference signal (pilot). However, while previous schedulers may have tended to be conservative in MCS allocation to allow for some additional margin in the link budget, energy-efficient scheduling could recommend a slightly more aggressive MCS level, thereby using fewer resources in the time domain (or a larger set of resource elements in the frequency domain). Such an approach could create more opportunities for the flexible shutdown of power-intensive RF circuitry.

[0050] Once PRB scheduling is complete, the base station's low-power model (which in some examples can be selected from several possible power models) can be used to determine the frame's power consumption and feed it to the sleep mode management component (which can be implemented at layer 2). Scheduling and MCS allocation may have incorporated any suggestions from the RU regarding the selection of prediction model-based modulation and RB allocation.

[0051] In some examples, if there is a momentary deviation in real-time service, sleep mode management can attempt to determine if a more energy-efficient transmission is possible. If sleep mode management does not find a more energy-efficient transmission, it can accept the transmission recommended by the scheduler and MCS allocation 408, and can continue the transmission, and can communicate the sleep mode decision to the corresponding RU. Example Architecture

[0052] Figure 5 A system architecture 500 for a two-layer sleep control engine that can facilitate hierarchical sleep mode management in a base station, according to embodiments of the present disclosure, is illustrated. In some examples, multiple portions of the system architecture 500 may be used to implement... Figure 1 The system architecture consists of (multiple) parts to facilitate hierarchical sleep mode management in base stations.

[0053] The system architecture 500 includes a spatial correlation component 502 for service requirements, a temporal correlation component 504 for service requirements, implementations of Graph Convolutional Networks (GCN) 1 506A, GCN 2 506B and GCN N 506N, implementations of Long Short-Term Memory Networks (LSTM) 1508A, LSTM B 508B and LSTM N 508N, a historical base station state database 510, an AI / ML-driven sleep control engine 512, a high-performance sleep control engine 514 (for determining long-term sleep states, which can be implemented, for example, in O-DU, RIC and / or edge), a lightweight sleep control engine 516 (for determining micro-sleep states, which can be implemented in O-RU), a macro sleep pattern decision 518, a micro sleep pattern decision 520, a RAN actuator 522 (RU, baseband, etc.) for energy efficiency enhancement, a radio control 524, and an RF measurement service predictor 526.

[0054] This technique can be implemented to facilitate deep reinforcement learning for base station sleep mode control. Using a neural network (NN) architecture for reinforcement learning-based sleep mode control can be advantageous because it has the ability to respond to changing RF environments and service demand matrices. Therefore, in some examples of hierarchical sleep mode policy engines, deep reinforcement learning (RL) can be used as the underlying concept with explicit state and action spaces related to network energy efficiency optimization. It can be noted that in some examples, by definition, any two consecutive base station state switching operations may be correlated with each other, and thus the current base station switching operation can further affect overall energy consumption in the long run. However, analytical methods and greedy algorithms almost completely ignore these sequential dependencies between consecutive base station sleep control decisions. Here again, RL according to this technique can be an attractive tool that incorporates such sequential dependencies, as well as a historical base station state database (with limited memory) maintained by each base station to feed into the optimal policy selection, as depicted in System Architecture 500. For large networks, Q-table-based RL techniques may not be suitable because they may suffer from the curse of dimensionality. Specifically, the Markov formula for the problem can be defined as follows: 1. State Space: In this context, the state can include service requirements, current energy consumption, and committed throughput (or other KPIs prioritized by the network operator). In some examples, to accommodate faster training, the service load aspect of the state space can be discretized to consider 10%, 20%, and so on up to 100%, in 10% increments (note that in some examples, other discretization steps beyond 10% are possible). This helps reduce the number of combinations considered, preventing the total number of combinations from growing exponentially as more cells and carriers are considered. In some examples, additional components such as autoencoders can also be used to reduce the dimensionality of the state space. 2. Action Space: This can represent a number of valid actions constrained by a specific sleep state that can be actuated within the base station, and the base station's ability to respond (activate / deactivate) within the time allocated to each of those sleep states. For example, if the PA can operate in a low-power state, it can be understood that a power-down PA will not be able to process a 100% loaded PRB map in the next subframe due to the upslope time required to return to full power. Additionally, it is also possible that the action space should have a reasonable cardinality (the number of actions to be taken); otherwise, the number of actions responding to different service conditions may become too small. On the other hand, a large cardinality may negatively impact the convergence time of RL training. 3. Strategy: Offline or online policy approaches can be used, where, respectively, the agent progressively learns the optimal value function / policy to directly find the action value of the optimal policy, or it can follow a set of greedy policies or the currently effective policy. For example, in some examples of implementing offline policy approaches, Q-learning can be used to obtain the optimal policy, and it can differ from the policy used to generate the samples. In some examples, this approach can help the network decision engine absorb new and unforeseen information and additionally address situations where training data is insufficient. 4. Reward Function: The reward for an action can be negative if it results in one of two situations: (a) failure to meet any weighted combination of specific network KPIs or expected KPIs (such as network throughput, latency, and congestion probability), and (b) an overall increase in energy consumption compared to the base station's long-term average energy consumption trend value. As a minor requirement, the latency threshold is exceeded by a negative value of a lower magnitude that may result in Guaranteed Bit Rate (GBR) services, but has no effect when considering non-GBR service categories. Positive reward values ​​can be set to introduce sleep states, and the longer the sleep state is identified while satisfying KPI constraints, the larger the reward value. In some examples, the reward value can be normalized to between +1 and -1.

[0055] Regarding the high-level architecture described above for tiered sleep modes, several derivative examples can be envisioned that can utilize data available at various levels of the protocol stack.

[0056] This technology can be implemented to facilitate hierarchical sleep control engines for Open RAN (O-RAN) base stations. As an example of a hierarchical sleep mode framework, System Architecture 500 depicts a high-level architecture employing a two-tier deployment of the hierarchical sleep mode concept. In this example, the lower tier resides within the O-RU, and the higher tier resides within the O-DU or edge server.

[0057] This two-layer structure allows for the flow of different granular timescales of each response in the Sleep Control Engine (SCE) and the corresponding inputs required for them to make decisions.

[0058] Figure 6 An example system architecture 600 for a radio-controlled autonomous sleep mode without upper-layer interaction is illustrated according to embodiments of the present disclosure. In some examples, portions(s) of system architecture 600 may be used to implement... Figure 1 The system architecture consists of (multiple) parts to facilitate hierarchical sleep mode management in base stations.

[0059] The system architecture 600 includes: a layer 1 602, a short-term service forecasting engine 604, instantaneous signal statistics at the PHY / RF module 606, RF data collection 608, a lightweight sleep control engine 610 (microsleep), PHY and radio control 612, and a RAN actuator 614 for energy efficiency enhancement.

[0060] This technology can be implemented to facilitate autonomous sleep mode management for radios. The benefits of an O-RAN-based implementation include the ability to use standardized messaging across open interfaces between O-DUs and O-RUs using Open Fronthaul (Open FH) control user synchronization (CUS) and management (M) planes, allowing for the deployment of O-DUs from various sources and O-RUs from different sources.

[0061] System Architecture 600 describes an autonomous operating mode of hierarchical sleep mode hosted on the RU, where the RU only executes... Figure 5 The system architecture 500 captures sleep mode control engine functionality within a lightweight sleep control engine component. This example simplifies the hierarchical sleep mode framework to a possible single layer, allowing the radio unit to implement the framework independently of higher layers to autonomously initiate and wake from sleep modes. More specifically, in this mode, the O-RU may not be able to independently influence scheduling and MCS allocation, so the intention could be to utilize radio control independently while taking into account RF front-end control OA bias, pre-driver gain, and the ability of components in the RF front-end to be put into sleep for a given period. Therefore, a shallow NN model can be used for traffic prediction based on power measurements on a per-subcarrier basis. In some examples, this prediction can be made more accurate if modulation information is made transparently available to the RU on a per-subcarrier basis. Previous methods might involve the RU receiving in-phase quadrature (I / Q) information only from the DU on a per-subframe basis.

[0062] In this example, a distinction can be made: the duration of sleep can be limited by the amount of advance notice the RU has for DL ​​data transmission. For example, the MAC scheduler can schedule DL data three subframes in advance of its physical transmission time. If a notification is also sent to the RU simultaneously, the RU component that can be woken up within approximately two 5G NR subframes can be safely powered down without incurring any degradation to KPIs. Example process flow

[0063] Figure 7 An example process flow 700 according to one embodiment of the present disclosure is shown, which can facilitate hierarchical sleep mode management in a base station. In some examples, one or more embodiments of process flow 700 can be implemented through... Figure 1 Layered sleep mode management in base station components 108 or Figure 10 The computing environment is 1000 to achieve this.

[0064] It is understood that the operation flow of process flow 700 is an example operation flow, and there may be embodiments that implement more or fewer operation flows than depicted, or embodiments that implement the depicted operation flows in a different order than depicted. In some examples, process flow 700 may be... Figure 4 Process flow 400 Figure 8 Process flow 800 and / or Figure 9 One or more embodiments of one or more process flows in process flow 900 are used in combination.

[0065] Process flow 700 starts at 702 and moves to operation 704.

[0066] Operation 704 describes determining a first sleep mode action for a first set of cellular base station signal indicators for a first time period, wherein the first set of cellular base station signal indicators is determined based on a first layer of communication of the cellular base station. This may include using a set of statistical data for the first time period to predict the sleep mode of the base station.

[0067] In some examples, the first set of cellular base station signal metrics includes the instantaneous radio frequency signal metrics of the cellular base station. This can be similar to... Figure 3 Instantaneous signal statistics at PHY / RF module 328.

[0068] In some examples, the first sleep mode action involves modifying the modulation and coding scheme of the cellular base station. This can be similar to... Figure 3 The MCS selection is 314.

[0069] In some examples, the first sleep mode actions involve modifying the physical layer and radio control of the cellular base station independently of modifying its modulation and coding scheme. This can be similar to... Figure 3 The PHY and radio control 318 are located in the lightweight sleep control engine 316, which bypasses the MCS selector 314.

[0070] After operation 704, process flow 700 moves to operation 706.

[0071] Operation 706 describes determining a second sleep mode action for a second set of cellular base station signal indicators for a second time period, wherein the first set of cellular base station signal indicators differs from the second set, wherein the second set of cellular base station signal indicators is determined based on a second communication layer of the cellular base station, and wherein the first time period differs from the second time period. This may include using another set of statistics for the first time period to make another sleep mode prediction for the base station relative to operation 704. In some examples, different layers of sleep mode management may use statistics from different communication layers.

[0072] For example, one layer could be the physical layer. Another layer could include MAC, Radio Link Control (RLC), and Packet Data Convergence Protocol (PDCP). And a third layer could include Radio Resource Control (RRC).

[0073] In some examples, the second set of cellular base station signal metrics includes the scheduler buffer queue status of the cellular base station, as well as the first set of cellular base station signal metrics. Using Figure 3 For example, the scheduler buffer queue state can be similar to scheduler buffer queue state 324, and level 2 304 can use this information in conjunction with the results from level 1 306.

[0074] In some examples, the second set of cellular base station signal metrics includes statistics from multiple distributed units of a split cellular base station (some of which could have been utilized by a single base station in previous methods). In some examples, determining the second sleep mode action involves performing machine learning or AI-driven learning on the statistics from multiple distributed units of the split cellular base station. In some examples, the second set of cellular base station signal metrics includes user association and handover control metrics. Figure 3 For example, this could be similar to multi-DU data collection and ML prediction / RIC 320.

[0075] After operation 706, process flow 700 moves to operation 708.

[0076] Operation 708 describes determining the arbitrated sleep mode action based on a first sleep mode action and a second sleep mode action. This may include selecting one of two sleep modes, and / or a combination of sleep modes. In some examples, this may be achieved by... Figure 3 Arbitrator 308 performs the arbitration. In other examples, this may include service scheduler 312 synthesizing, combining, or selecting from sleep modes from tier 1306 and another sleep mode from tier 3302 and / or tier 2304.

[0077] After operation 708, process flow 700 moves to operation 710.

[0078] Operation 710 describes an arbitration-based sleep mode action to put at least a portion of the cellular base station into sleep mode. This may include enabling... Figure 3 The PHY and radio control 318, for example, enable the actuator to shut down at least a portion of the PA.

[0079] After operation 710, process flow 700 moves to 712, where process flow 700 ends.

[0080] Figure 8 An example process flow 800 according to one embodiment of the present disclosure is shown, which can facilitate hierarchical sleep mode management in a base station. In some examples, one or more embodiments of process flow 800 can be implemented through... Figure 1 Layered sleep mode management in base station components 108 or Figure 10 The computing environment is 1000 to achieve this.

[0081] It is understood that the operation flow of process flow 800 is an example operation flow, and there may be embodiments that implement more or fewer operation flows than depicted, or embodiments that implement the depicted operation flows in a different order than depicted. In some examples, process flow 800 may be... Figure 4 Process flow 400 Figure 7 Process flow 700 and / or Figure 9 One or more embodiments of one or more process flows in process flow 900 are used in combination.

[0082] Process flow 800 starts at 802 and moves to operation 804.

[0083] Operation 804 describes determining a first sleep mode action for a first set of cellular base station signal indicators relative to a first time period, applicable to the cellular base station, wherein the first set of cellular base station signal indicators is determined based on a first layer of communication of the cellular base station. In some examples, operation 804 may be in conjunction with... Figure 7 The operation is implemented in a similar way to 704.

[0084] After operation 804, process flow 800 moves to operation 806.

[0085] Operation 806 describes determining a second sleep mode action for a second set of cellular base station signal indicators relative to a second time period, applicable to the cellular base station, wherein the second set of cellular base station signal indicators is determined based on the second layer of communication of the cellular base station. In some examples, operation 806 may be in conjunction with... Figure 7 It is implemented in a similar way to operation 706.

[0086] In some examples, the second set of cellular base station signal metrics includes spatial correlations in service demand metrics, and determining the second sleep mode action involves using a graph convolutional network to handle the spatial correlations of service demand metrics. This is similar to... Figure 5 The spatial relevance of business requirements 502 can be fed into GCN 1 506A (etc.).

[0087] In some examples, the second set of cellular base station signal metrics includes the temporal correlation of service demand metrics, and determining the second sleep mode action involves utilizing a long short-term memory artificial neural network to process the temporal correlation of service demand metrics. This is similar to... Figure 5 The time-dependent nature of the business requirements 504 can be fed into LSTM 1 508A (etc.).

[0088] In some examples, the determination of the second sleep mode action is based on the historical operational status of the cellular base station. This is similar to... Figure 5 The database of historical base station status 510.

[0089] In some examples, inducing at least a portion of the cellular base station to sleep based on selected sleep mode actions includes enabling actuators for the cellular base station. This can be similar to... Figure 5 RAN actuator 522 for energy efficiency enhancement.

[0090] In some examples, the first time period is shorter than the second time period, and the determination of the second sleep mode action is based on the first sleep mode action. That is, in some examples, the hierarchy can provide sleep mode recommendations incorporated by higher levels in the decision-making process at higher levels.

[0091] After operation 806, process flow 800 moves to operation 808.

[0092] Operation 808 describes determining the selected sleep mode action based on the first sleep mode action and the second sleep mode action. In some examples, operation 808 can be in conjunction with... Figure 7 It is implemented in a similar way to operation 708.

[0093] After operation 808, process flow 800 moves to operation 810.

[0094] Operation 810 describes putting at least a portion of the cellular base station into sleep mode based on a selected sleep mode. In some examples, operation 810 can be in conjunction with... Figure 7 The operation is implemented in a similar way to 710.

[0095] After operation 810, process flow 800 moves to 812, where process flow 800 ends.

[0096] Figure 9 An example process flow 900 according to one embodiment of the present disclosure is shown, which can facilitate hierarchical sleep mode management in a base station. In some examples, one or more embodiments of process flow 900 can be implemented through... Figure 1 Layered sleep mode management in base station components 108 or Figure 10 The computing environment is 1000 to achieve this.

[0097] It is understood that the operational flow of process flow 900 is an example operational flow, and there may be embodiments that implement more or fewer operational flows than depicted, or embodiments that implement the depicted operational flows in a different order than depicted. In some examples, process flow 900 may be... Figure 4 Process flow 400 Figure 8 Process flow 800 and / or Figure 9 and / or Figure 9 One or more embodiments of one or more process flows in process flow 900 are used in combination.

[0098] Process flow 900 starts from 902 and moves to operation 904.

[0099] Operation 904 describes determining the corresponding sleep mode action of the base station equipment based on a corresponding set of base station signal indicators, where the corresponding set of base station signal indicators corresponds to the corresponding communication layer of the base station equipment. In some examples, operation 904 can be compared with... Figure 7 The operation is implemented in a similar way to 704-706.

[0100] After operation 904, process flow 900 moves to operation 906.

[0101] Operation 906 describes determining a sleep pattern action based on a corresponding sleep pattern action, thereby obtaining the determined sleep pattern action. In some examples, operation 906 can be compared with... Figure 7 It is implemented in a similar way to operation 708.

[0102] In some examples, the first time period within a given time period is shorter than the second time period within a given time period, wherein the first sleep mode action associated with the first time period in the corresponding sleep mode action includes shutting down the power amplifier and at least a portion of the analog portion of the base station equipment, and wherein the second sleep mode action associated with the second time period in the corresponding sleep mode action includes shutting down the power amplifier and the entire analog portion of the RF front end. This can be similar to Figure 2 Sleep modes 0 202 and 1 204.

[0103] In some examples, the first time period within a given time period is shorter than the second time period within the same time period, wherein the first sleep mode action associated with the first time period in the corresponding sleep mode action is determined by the radio hardware unit of the base station equipment, and wherein the second sleep mode action associated with the second time period in the corresponding sleep mode action is determined by at least one of the following: the distributed unit of the base station equipment, the radio access network intelligent controller, or the edge device. This can be similar to Figure 5 The system includes a lightweight sleep control engine 516 for the first sleep mode operation and a high-performance sleep control engine 514 for the second sleep mode operation.

[0104] In some examples, the first sleep mode action in the corresponding sleep mode action is determined based on at least one of the following: the bandwidth of the base station device, the number of antenna ports of the base station device, or the per-active-user multiple-input multiple-output mode of the base station device. This can be similar to Figure 4 The system parameter is 412.

[0105] In some examples, the first time period within a given time period is shorter than the second time period within the same time period; the first sleep mode action associated with the first time period in the corresponding sleep mode action is based on a first prediction of network services by the base station equipment; the second sleep mode action associated with the second time period in the corresponding sleep mode action is based on a second prediction of network services by the base station equipment; and the third time amount associated with the first prediction of network services by the base station equipment is shorter than the fourth time amount associated with the first prediction of network services by the base station equipment. This is similar to service monitoring 402, where short-term prediction 404 corresponds to a prediction for a time period shorter than long-term prediction 406.

[0106] In some examples, the radio unit autonomous mode of the base station is configured to determine a first sleep mode action in sleep mode actions, and to implement the first sleep mode action independently of the component configured to determine a second sleep mode action in sleep mode actions.

[0107] After operation 906, process flow 900 moves to operation 908.

[0108] Operation 908 describes actions based on a determined sleep pattern to induce at least a portion of the base station equipment to sleep. In some examples, operation 908 can be performed in conjunction with... Figure 7 It is implemented in a similar way to operation 708.

[0109] After operation 908, process flow 900 moves to 910, where process flow 900 ends. Example operating environment

[0110] To better understand the additional context of the various embodiments described herein, Figure 10 The following discussion is intended to provide a brief overview of a suitable computing environment 1000 in which the various embodiments described herein can be implemented.

[0111] For example, a portion of computing environment 1000 can be used to implement... Figure 1 One or more embodiments of base station 102 and / or user equipment 104.

[0112] In some examples, computing environment 1000 can achieve this. Figure 4 and / or Figures 7 to 9 One or more embodiments of the process flow are used to facilitate hierarchical sleep mode management in base stations.

[0113] Although the embodiments have been described in the general context of computer-executable instructions that can run on a single or multiple computers, those skilled in the art will recognize that the embodiments may also be implemented in combination with other program modules and / or implemented as a combination of hardware and software.

[0114] Typically, a program module includes routines, programs, components, data structures, etc., that perform a specific task or implement a specific abstract data type. Furthermore, those skilled in the art will appreciate that various methods can utilize other computer system configuration practices, including single-processor or multi-processor computer systems, minicomputers, mainframe computers, Internet of Things (IoT) devices, distributed computing systems, and personal computers, handheld computing devices, microprocessor-based or programmable consumer electronics, each of which can be operatively coupled to one or more associated devices.

[0115] The embodiments illustrated herein can also be practiced in a distributed computing environment, where certain tasks are performed by remote processing devices linked via a communication network. In a distributed computing environment, program modules can reside on both local and remote storage devices.

[0116] Figure 10 An example block diagram 1000 of a computer operable to perform embodiments of the present disclosure is shown. UE 1004 may typically include a device used by an end user to access a communication network. UE 1004 may be configured to receive messages from a communication network 1006, which may be, for example, a global communication network such as the Internet.

[0117] Messages from UE 1004 can be received and processed by core network 1008, which may include components of third-generation (3G), fourth-generation (4G), Long Term Evolution (LTE), 5G, or other wireless communication networks. Core network 1008 can be configured to establish connectivity between UE 1004 and communication network 1006, such as by facilitating services such as connectivity and mobility management, authentication and authorization, subscriber data management, and policy management. Messages transmitted between UE 1004 and communication network 1006 can be propagated through centralized unit (CU) 1010, distributed unit (DU) 1012, RU 1014, and antenna 1016.

[0118] CU 1010 can be configured to handle non-real-time RRC and PDCP communications. DU 1012 can be configured to handle communications transmitted according to the RLC, MAC, and PHY layers. RU 1014 can be configured to convert radio signals transmitted to antenna 1016 from digital packets to radio signals, and to convert radio signals received from antenna 1016 from radio signals to digital packets. Antenna 1016 (which may include a transceiver) can be configured to transmit and receive radio waves used for transmitting information. in conclusion

[0119] Computing devices typically include various media, which may include computer-readable storage media, machine-readable storage media, and / or communication media, these two terms being used differently from each other herein. A computer-readable storage medium or a machine-readable storage medium can be any available storage medium accessible to a computer and includes both volatile and non-volatile media, as well as removable and non-removable media. By way of example, and not limitation, a computer-readable storage medium or a machine-readable storage medium can be implemented in conjunction with any method or technology for storing information such as computer-readable or machine-readable instructions, program modules, structured data, or unstructured data.

[0120] Computer-readable storage media may include, but are not limited to, random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, optical disc storage, solid-state drives or other solid-state storage devices, or other tangible and / or non-transitory media that can be used to store desired information. In this regard, the terms “tangible” or “non-transitory” as used herein for storage devices, memories, or computer-readable media shall be understood as modifiers excluding the propagation of only transient signals themselves, and shall not waive the rights to all standard storage devices, memories, or computer-readable media that do not merely propagate transient signals themselves.

[0121] Computer-readable storage media can be accessed by one or more local or remote computing devices, for example, via access requests, queries or other data retrieval protocols, to perform various operations on the information stored on the media.

[0122] Communication media typically embody computer-readable instructions, data structures, program modules, or other structured or unstructured data in data signals (such as modulated data signals, for example, carrier waves or other transmission mechanisms), and include any medium for delivering or transmitting information. The term "modulated data signal" or signal refers to a signal whose one or more characteristics are set or altered to encode information into one or more signals. By way of example, and not limitation, communication media include wired media (such as wired networks or direct wired connections) and wireless media (such as acoustic, RF, infrared, and other wireless media).

[0123] In this specification, terms such as “data repository,” “data storage device,” “database,” “cache,” and any other information storage component related to the operation and function of the component refer to a “memory component,” or an entity embodied in “memory” or a component constituting memory. It will be understood that the memory component or computer-readable storage medium described herein can be volatile memory or non-volatile memory device, or may include both. By way of illustration and not limitation, a non-volatile memory device may include ROM, programmable ROM (PROM), EPROM, EEPROM, or flash memory. Volatile memory may include RAM, which is used as an external cache. By way of illustration and not limitation, RAM may be available in many forms, such as synchronous RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), and Rambus RAM (DRRAM). Additionally, the memory components of the systems or methods disclosed herein are intended to include, but are not limited to, these and any other suitable types of memory.

[0124] The embodiments shown in this disclosure can be practiced in a distributed computing environment, where certain tasks are performed by a remote processing device linked via a communication network. In a distributed computing environment, program modules can reside on both local and remote memory storage devices.

[0125] The aforementioned systems and processes can be embodied in hardware, such as a single integrated circuit (IC) chip, multiple ICs, ASICs, etc. Furthermore, the order in which some or all of the process blocks appear in each process should not be considered restrictive. Rather, it should be understood that some of the process blocks can be executed in various orders, and this document does not explicitly specify all orders.

[0126] As used herein, the terms “component,” “module,” “system,” “interface,” “cluster,” “server,” “node,” etc., are generally intended to refer to a computer-related entity, which is hardware, a combination of hardware and software, software or software in execution, or an entity associated with an operating machine having one or more specific functions. For example, a component can be, but is not limited to, a process running on a processor, a processor, an object, an executable file, a thread of execution, (multiple) computer-executable instructions, a program, and / or a computer. For example, both an application running on a controller and the controller itself can be components. One or more components may reside in a process and / or a thread of execution, and a component may reside on one computer and / or be distributed across two or more computers. As another example, an interface may include input / output (I / O) components and associated processors, applications, and / or application programming interface (API) components.

[0127] Furthermore, various embodiments may be implemented as methods, apparatus, or articles of art that use standard programming and / or engineering techniques to produce software, firmware, hardware, or any combination thereof to control a computer to implement one or more embodiments of the disclosed subject matter. Articles of art may encompass computer programs accessible from any computer-readable device or computer-readable storage / communication medium. For example, computer-readable storage media may include, but are not limited to, magnetic storage devices (e.g., hard disks, floppy disks, magnetic stripes, etc.), optical disks (e.g., CDs, DVDs, etc.), smart cards, and flash memory devices (e.g., cards, sticks, key drives, etc.). Of course, those skilled in the art will recognize that many modifications can be made to this configuration without departing from the scope or spirit of the various embodiments.

[0128] Furthermore, the terms “example” or “exemplary” are used herein to mean as an example, instance, or illustration. Any embodiment or design described herein as “exemplary” should not necessarily be construed as superior to or advantageous to other embodiments or designs. Rather, the use of the term “exemplary” is intended to present concepts in a concrete manner. As used in this application, the term “or” is intended to mean inclusive “or” rather than exclusive “or.” That is, unless otherwise specified or clear from the context, “X adopts A or B” is intended to mean any of the natural inclusive arrangements. That is, if X adopts A; X adopts B; if X adopts both A and B, then “X adopts A or B” is satisfied in any of the above examples. Furthermore, unless otherwise specified or clear from the context to the singular form, the terms “a” and “an” as used in this application and the appended claims should generally be interpreted as meaning “one or more”.

[0129] The foregoing description includes examples from this specification. It is certainly impossible to describe all conceivable combinations of components or methods for the purposes of describing this specification, but those skilled in the art will recognize that many further combinations and arrangements are possible. Therefore, this specification is intended to encompass all such changes, modifications, and variations that fall within the spirit and scope of the appended claims. Furthermore, if the term "comprising" is used in the detailed description or claims, such a term is intended to be inclusive in a manner similar to the term "comprising," as interpreted when "comprising" is used as a transitional word in the claims.

Claims

1. A method comprising: The system determines a first sleep mode action for a first group of cellular base station signal indicators for a first time period, wherein the first group of cellular base station signal indicators is determined based on the first layer of communication of the cellular base station. The system determines a second sleep mode action for the cellular base station and a second set of cellular base station signal indicators for the second time period, wherein the first set of cellular base station signal indicators is different from the second set of cellular base station signal indicators, wherein the second set of cellular base station signal indicators is determined based on the second layer of communication of the cellular base station, and wherein the first time period is different from the second time period. The system determines the arbitrated sleep mode action based on the first sleep mode action and the second sleep mode action; as well as The system induces at least a portion of the cellular base station to sleep based on the sleep pattern action of the arbitration.

2. The method according to claim 1, wherein the first set of cellular base station signal indicators includes the instantaneous radio frequency signal indicators of the cellular base station.

3. The method according to claim 2, wherein the first sleep mode action includes: modifying the modulation and coding scheme of the cellular base station.

4. The method according to claim 2, wherein the first sleep mode action includes: The physical layer and radio control of the cellular base station can be modified independently of the modulation and coding scheme of the cellular base station.

5. The method according to claim 1, wherein the second set of cellular base station signal indicators includes the scheduler buffer queue status of the cellular base station and the first set of cellular base station signal indicators.

6. The method according to claim 1, wherein the second set of cellular base station signal indicators includes statistical data of multiple distributed units of the cellular base station.

7. The method of claim 6, wherein determining the second sleep mode action comprises: Machine learning or artificial intelligence-driven learning is performed on the statistical data of multiple distributed units of the cellular base station.

8. The method according to claim 1, wherein the second set of cellular base station signal indicators includes user association and handover control indicators.

9. A system comprising: processor; as well as A memory coupled to the processor, the memory including instructions that, in response to execution by the processor, cause the system to perform operations, the operations including: A first sleep mode action is determined relative to a first time period, applicable to a first set of cellular base station signal indicators, wherein the first set of cellular base station signal indicators is determined based on a first layer of communication of the cellular base station; A second sleep mode action is determined relative to a second time period for the cellular base station, for a second set of cellular base station signal indicators, wherein the second set of cellular base station signal indicators is determined based on the second layer of communication of the cellular base station; The selected sleep mode action is determined based on the first sleep mode action and the second sleep mode action; and The selected sleep mode action is used to put at least a portion of the cellular base station into sleep.

10. The system of claim 9, wherein the second set of cellular base station signal indicators includes spatial correlation among service demand indicators, and wherein determining the second sleep mode action includes: Graph convolutional networks are used to handle the spatial correlations in the business requirement metrics.

11. The system of claim 10, wherein the second set of cellular base station signal indicators includes the time correlation of service demand indicators, and wherein determining the second sleep mode action includes: Long Short-Term Memory (LSTM) artificial neural networks are used to handle the time-relatedness of business demand metrics.

12. The system of claim 11, wherein the determination of the second sleep mode action is based on the historical operating state of the cellular base station.

13. The system of claim 9, wherein inducing at least a portion of the cellular base station to sleep based on the selected sleep mode action comprises: Enable the actuator of the cellular base station.

14. The system of claim 9, wherein the first time period is shorter than the second time period, and wherein the determination of the second sleep mode action is based on the first sleep mode action.

15. A non-transitory computer-readable medium comprising instructions that, in response to execution, cause a system including a processor to perform an operation, the operation comprising: The corresponding sleep mode action for the base station device is determined based on the corresponding set of base station signal indicators for the base station device, wherein the corresponding set of base station signal indicators corresponds to the corresponding communication layer of the base station device. The sleep pattern action is determined based on the corresponding sleep pattern action, thereby obtaining the determined sleep pattern action; as well as The base station device is put to sleep by actions based on the determined sleep pattern.

16. The non-transitory computer-readable medium of claim 15, wherein the first time period in the corresponding time period is shorter than the second time period in the corresponding time period, wherein the first sleep mode action associated with the first time period in the corresponding sleep mode action includes: The power amplifier of the base station device and at least a portion of the analog section of the base station device are turned off, and the second sleep mode action associated with the second time period in the corresponding sleep mode action includes: turning off all of the analog section of the power amplifier and the radio frequency front end.

17. The non-transitory computer-readable medium of claim 15, wherein the first time period in the corresponding time period is shorter than the second time period in the corresponding time period, wherein the first sleep mode action associated with the first time period in the corresponding sleep mode action is determined by the radio hardware unit of the base station device, and wherein the second sleep mode action associated with the second time period in the corresponding sleep mode action is determined by at least one of: the distributed unit of the base station device, the radio access network intelligent controller of the base station device, or the edge device of the base station device.

18. The non-transitory computer-readable medium of claim 15, wherein the first sleep mode action in the corresponding sleep mode action is determined based on at least one of the following: the bandwidth of the base station device, the number of antenna ports of the base station device, or the per-active-user-equipment multiple-input multiple-output mode of the base station device.

19. The non-transitory computer-readable medium of claim 15, wherein the first time period in the corresponding time period is shorter than the second time period in the corresponding time period, wherein the first sleep mode action associated with the first time period in the corresponding sleep mode action is based on a first prediction by the base station device regarding network services, wherein the second sleep mode action associated with the second time period in the corresponding sleep mode action is based on a second prediction by the base station device regarding network services, and wherein a third time amount associated with the first prediction by the base station device regarding network services is less than a fourth time amount associated with the first prediction by the base station device regarding network services.

20. The non-transitory computer-readable medium of claim 15, wherein the radio unit autonomous mode of the base station is configured to determine a first sleep mode action among the sleep mode actions, and to implement the first sleep mode action independently of a component configured to determine a second sleep mode action among the sleep mode actions.

Images