Resource-constrained group agent communication energy consumption assessment method
Through offline energy consumption modeling and online queue matching optimization, the problem of insufficient energy consumption evaluation in wireless low-power networks is solved, and efficient energy consumption evaluation and energy consumption optimization of resource-constrained swarm agents are achieved.
Patent Information
- Application Number
- CN202510918426.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-03
- Publication Date
- 2025-09-12
AI Technical Summary
Existing wireless low-power networks lack a complete energy consumption evaluation system and are unable to dynamically adjust communication strategies, resulting in resource waste and idle computing resources. In addition, heterogeneous communication protocols are difficult to quantify and optimize energy consumption.
A resource-constrained swarm agent communication energy consumption evaluation method is adopted to separate the offline and online stages, and energy consumption modeling is performed based on the type of deep neural network layer. The computing task allocation is optimized through the queue matching optimization algorithm to achieve more efficient energy consumption evaluation.
The communication energy consumption evaluation efficiency of resource-constrained swarm agents is improved, the overall energy consumption is reduced, and the computing resource utilization and response speed are optimized.
Smart Images
Figure CN120640322A_ABST
Abstract
Description
Technical Field
[0001] The embodiments of the present application relate to the technical field of group intelligent agent communication, and in particular to a method for evaluating energy consumption of resource-constrained group intelligent agent communication. Background Art
[0002] A swarm agent is a group of multiple agents that can exhibit complex intelligent behaviors through simple interactions and collaborative work. Current mobile swarm agents face challenges such as the high complexity of computing tasks, demanding deployment environments, and long computing times. As the application areas of swarm agents continue to expand, the computing tasks that need to be performed are becoming more and more complex. For example, in intelligent drone scenarios, real-time image processing and path planning are required; in intelligent robot scenarios, complex motion control and environmental perception are required. The complexity of these tasks leads to higher demands on computing resources, and traditional mobile swarm agents often can only carry limited computing resources, making it difficult to meet the needs of complex tasks. Among these, energy consumption is the most prominent issue, which limits the computing performance of resource-constrained swarm agents from the source.
[0003] Low-power wireless networks (LPWNs), with their low power consumption, low cost, short-range communication, low data rates, and optimized protocols, enable swarm agents in wireless networks to operate with minimal power consumption, thus overcoming the drawbacks of resource constraints (battery power, computing and storage capacity). LPWNs utilize specially designed communication protocols, such as IEEE 802.15.4, Zigbee, LoRa, and NB-IoT. These protocols are optimized for low-power communication and employ a range of technologies to reduce device power consumption. For example, these protocols often utilize device sleep modes to reduce power consumption when devices are inactive, employ data compression techniques to reduce energy consumption during data transmission, and implement collision avoidance and power control to improve network efficiency and stability.
[0004] However, the inventors of the present application have discovered that there are still deficiencies in wireless low-power networks.
[0005] First, in wireless low-power networks, there is a lack of a complete evaluation system for the energy consumption of swarm intelligent systems.
[0006] The energy consumption assessment of a swarm agent system primarily includes the system's basic operating energy consumption, system loss energy consumption, communication energy consumption, and computing energy consumption. Current energy consumption assessment methods mostly use one or several energy consumption statistics methods, or customize them based on the specific application scenario. These methods fail to consider the relationship between the overall system energy consumption and computing resources. Swarm agents often need to perform complex computational tasks in environments with limited computing resources, and data transmission is closely related to computing resource utilization. For example, data collected by smart sensors may require preliminary local processing and analysis to reduce data transmission and network load. Given the nature of system resources, current assessment methods are limited by the static nature of communication protocols (network communication protocols are often static), making it impossible to dynamically adjust communication strategies based on the swarm agent's computing resources and environmental conditions. This can result in the communication protocol failing to fully utilize available computing resources, resulting in wasted energy or idle computing resources during communication. Furthermore, they are limited by data processing and communication latency, which are crucial to system performance and responsiveness in swarm agent computing systems. However, current communication protocols often fail to fully consider the balance between data processing and communication latency, resulting in an inability to effectively optimize the system's response speed and resource utilization.
[0007] Second, each intelligent agent uses a different communication protocol for network communication, which makes it impossible to quantify the energy consumption generated by the communication network.
[0008] Different types of swarm agents often use different communication protocols, which can vary significantly in terms of communication range, data transmission rate, and power consumption. This heterogeneous communication protocol makes it difficult to uniformly quantify and manage the energy consumption of the entire communication network. Due to the existence of multiple different communication protocols within a communication network, and the different communication modes and energy consumption characteristics of different types of agents, it is difficult to accurately quantify the energy consumption of the entire communication network. Traditional energy consumption assessment methods are not well suited to such heterogeneous communication networks, making it difficult to accurately assess and optimize the energy consumption of the communication network. The inability to quantify the energy consumption of the communication network makes it difficult to effectively optimize the energy consumption of the communication network. Without the ability to accurately calculate the energy consumption of different types of agents, it is difficult to implement targeted optimization strategies to reduce the energy consumption of the entire communication network, which may lead to unnecessary energy waste in the communication network, thereby affecting the overall energy efficiency of the system. Summary of the Invention
[0009] In order to solve the above technical problems, the embodiments of the present application propose a method for evaluating the communication energy consumption of resource-constrained group intelligent agents, which divides the energy consumption evaluation into two parts: online stage and offline stage, and flexibly schedules computing tasks in real time in combination with the current operating environment of the resource-constrained group intelligent agent, thereby achieving more efficient and effective communication energy consumption evaluation of the group intelligent agent.
[0010] In order to achieve the above-mentioned purpose, an embodiment of the present application proposes a method for evaluating the communication energy consumption of a resource-constrained group agent, comprising the following steps: determining the communication protocol currently used by the resource-constrained group agent; wherein the communication protocol currently used by the resource-constrained group agent is a communication protocol with a 2.4GHz communication frequency, specifically a Wi-Fi protocol or a LoRa protocol; in the offline stage, the deep neural network is divided according to the type of layer, and the energy consumption of each layer is calculated separately to realize energy consumption modeling of the computational reasoning task; in the online stage, based on the communication protocol currently used by the resource-constrained group agent and the energy consumption modeling of the computational reasoning task established in the offline stage, combined with the current resource situation of the resource-constrained group agent, corresponding queue matching optimization is performed, and the computational reasoning task is matched to the most suitable resource-constrained group agent for computational reasoning, thereby completing the optimization evaluation of the communication energy consumption of the resource-constrained group agent.
[0011] In order to achieve the above-mentioned purpose, an embodiment of the present application also proposes a resource-constrained group intelligent agent communication energy consumption evaluation system, the system including: a basic setting module, used to determine the resource-constrained group intelligent agent, and determine the communication protocol currently used by the resource-constrained group intelligent agent, the communication protocol currently used by the resource-constrained group intelligent agent is a communication protocol with a 2.4GHz communication frequency, specifically a Wi-Fi protocol or a LoRa protocol; an offline modeling module, used to divide according to the type of layer of the deep neural network, and calculate the energy consumption of each layer respectively to realize energy consumption modeling of computational reasoning tasks; an online matching module, used to perform corresponding queue matching optimization based on the communication protocol currently used by the resource-constrained group intelligent agent and the energy consumption modeling of the computational reasoning tasks established in the offline stage, combined with the current resource situation of the resource-constrained group intelligent agent, and match the computational reasoning tasks to the most suitable resource-constrained group intelligent agent for computational reasoning, thereby completing the optimization evaluation of the communication energy consumption of the resource-constrained group intelligent agent.
[0012] In order to achieve the above-mentioned purpose, an embodiment of the present application also proposes an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute a resource-constrained group intelligent agent communication energy consumption assessment method as described above.
[0013] In order to achieve the above objectives, an embodiment of the present application further proposes a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it can implement the resource-constrained group intelligent agent communication energy consumption evaluation method as described above.
[0014] The present application proposes a method for evaluating the communication energy consumption of a resource-constrained swarm agent. First, the communication protocol currently used by the resource-constrained swarm agent is determined. The communication protocol currently used by the resource-constrained swarm agent is a communication protocol with a communication frequency of 2.4GHz, specifically the Wi-Fi protocol or the LoRa protocol. Next, the communication energy consumption of the resource-constrained swarm agent is evaluated. This process is divided into two stages, namely the offline stage and the online stage. In the offline stage, the energy consumption of the system operation is evaluated by modeling the energy consumption required for the offline stage tasks. In the online stage, an energy consumption resource matching model is designed. Based on the computing tasks and resource energy consumption in the offline stage, an optimization algorithm based on queue matching is designed to optimize the computing process and thus enhance the reasoning computing performance of the low-power swarm agent (resource-constrained swarm agent). The present application combines the current operating environment of the resource-constrained swarm agent to flexibly schedule computing tasks in real time, which can achieve more efficient and effective communication energy consumption evaluation of the swarm agent.
[0015] Optionally, in the offline phase, the deep neural network is divided according to its layer type and the energy consumption of each layer is calculated separately to implement energy consumption modeling of the computational inference task, including: The overall energy consumption of resource-constrained swarm agents is expressed as: ; in, Indicates the size of the L1 memory of the resource-constrained swarm agent, represents the total memory size of the resource-constrained swarm agent, represents the computation frequency of resource-constrained swarm agents, represents the current execution frequency of the resource-constrained swarm agent, represents the rated operating frequency of resource-constrained swarm agents, represents the core frequency of the resource-constrained swarm agent, and They represent the memory energy consumption of L1 level and L2 level of resource-constrained swarm agents, Indicates overall energy consumption; The energy consumption of all convolutional layers of a deep neural network is expressed as: ; in, 、 、 Represent the height, width and number of channels of the input features respectively, represents the total number of convolutional layers in the deep neural network, Indicates the The energy consumption of the convolutional layer, and Respectively represent The height and width of the convolution kernel of the convolution layer, represents the number of cumulative multiplication operations in a deep neural network, 、 、 Represent the height, width and number of channels of the output features respectively, represents the energy consumption of the feature during transmission, Represents the energy consumption of all convolutional layers of a deep neural network; The energy consumption of the pooling and fully connected layers is expressed as: ; in, represents the pooling operation, represents the energy consumption of convolution in the fully connected layer, and Represent the height and width of the convolution kernel in the fully connected layer, represents the energy consumption of pooling and fully connected layers; In the normalization layer, the step size and the size of the input feature are used as regression variables to predict the energy consumption of the normalization layer. The prediction of the energy consumption of the normalization layer is expressed as: ; in, The ability to represent input features, represents the normalization operation, Represents the predicted value of the energy consumption of the normalization layer.
[0016] Alternatively, assume that during the optimization process, the period of each computation task is divided into time slots, and the duration of each time slot is set to , assuming that there is a low-power network The group agents exchange data in each time slot On the swarm agent The amount of communication data exchanged is expressed as , the data transmission rate of the low-power network is ; It is expressed by the formula: ; in, is the preset weight coefficient, Indicates the transmit power, represents the channel gain, represents the noise power; Assuming that the amount of communication data of the group agent is proportional to the amount of computation data, let For the communication data volume The energy consumption required is calculated and the energy consumption coefficient is defined according to different networks and different CPU processing frequencies. , the computing resources of the swarm agent in each time slot are , the total energy consumption of network interaction and computational processing of each swarm agent is ; Based on this, the data transmission and computing energy consumption of the low-power network in each time slot is expressed as: ; in, Indicates time slot Data transmission and computing energy consumption on low-power networks; The average energy consumption over all time slots is expressed as: ; in, represents the average energy consumption over all time slots, .
[0017] Optionally, for the Wi-Fi protocol, the energy consumption of transmitting 1 bit of data is taken into account. , calculate the energy consumption of the current low-power network, and select the low-power network with the best communication quality to obtain the lowest energy consumption by comparing the energy consumption values; In the data transmission process of low-power networks, a multiple-input multiple-output system is usually used, and its antenna is represented by a frequency domain model as follows: ; in, represents the received signal vector, represents the channel matrix, which contains the channel state information, represents the signal vector sent, represents the noise vector, which is usually assumed to be additive Gaussian white noise; In the IEEE 802.11n standard, the signal is divided into 56 carrier channels with a width of 20 MHz, of which 30 carrier channels are represented as follows: ; in, is the total number of carriers; No. The carrier channels are expressed as: ; in, and Represent the amplitude and phase of the wireless signal respectively; During signal transmission, the fluctuation of the signal generates carrier frequency deviation, and the estimated The frequency of the carrier wave is: ; in, For the estimated The frequency of the carrier wave, For the actual The frequency of the carrier wave, Indicates the The index of the carrier, is the frequency offset, is the constant term of frequency offset, is the noise term; The frequency error is: ; ; in, The slope of the frequency error, is the intercept of the frequency error; The channel state information of the carrier can be expressed as a matrix , which means the The receiving antenna and The channel response between the transmit antennas is: ; Finally, the relationship between communication energy consumption and carrier amplitude and phase is established: ; in, Indicates the energy consumption under the Wi-Fi protocol, represents the carrier amplitude, represents the carrier phase, represents the Hadamard product, Represents element-wise addition.
[0018] Optionally, for the Lora protocol, the energy consumption required to transmit one byte in the Lora protocol is Bring in , to calculate ; ; in, represents the total energy consumption, represents the total transmitted energy, Indicates the length of the data being transmitted. Indicates the transmission time of the entire transmission packet; Specifically, it is the sum of the data prefix and the transmission payload length: ; ; in, Indicates the data prefix, Indicates the transmission payload length, represents the prefix code, Indicates the synchronization code period; The total energy consumption of the swarm agent using the Lora protocol is expressed as: ; in, Indicates the total number of bytes, Represents the total energy consumption of the swarm agents using the Lora protocol; Under the Lora protocol, the time slot process of the swarm agent is divided into two states: swarm agent dormancy and activation. The total energy is expressed as: ; ; in, Represents the total energy of the time slot process of the group agent under the Lora protocol, Represents dormant energy, represents activation energy, 、 、 、 、 、 They represent the energy required for system wake-up, data transmission, data processing, system sleep to wake-up, system transmission mode, and reception mode. 、 、 、 、 Respectively expressed as: ; ; ; ; ; in, Indicates the energy consumption of a single operation of the system. Indicates the number of times the system has run. 、 、 and They represent the energy consumption of data transmission, the energy consumed by the system from sleep to wake-up, the energy consumed by data sending, and the energy consumed by data receiving. Indicates the time taken for data transmission. Indicates the time when the system sends data. Indicates the time when the system receives data; ; in, and They represent the energy per bit of transmitted data and the number of bits transmitted, respectively. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the related technologies, the following is a brief introduction to the drawings required for use in the embodiments of the present application or the description of the related technologies. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.
[0020] Figure 1 is a flow chart of a method for evaluating energy consumption of resource-constrained group agent communication provided in one embodiment of the present application; Figure 2 This is a result diagram of a simulation test of a resource-constrained group agent under the Wi-Fi protocol and the LoRa protocol, provided in one embodiment of the present application; Figure 3 is a structural diagram of a resource-constrained group agent communication energy consumption evaluation system provided in another embodiment of the present application; Figure 4 It is a structural diagram of an electronic device provided in another embodiment of the present application. DETAILED DESCRIPTION
[0021] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, each embodiment of the present application will be described in detail below with reference to the accompanying drawings. However, it will be understood by those skilled in the art that in each embodiment of the present application, many technical details are proposed to enable the reader to better understand the present application. However, even without these technical details and various changes and modifications based on the following embodiments, the technical solutions claimed in the present application can be implemented. The division of the following embodiments is for convenience of description and should not constitute any limitation on the specific implementation of the present application. The various embodiments can be combined and referenced with each other under the premise of no contradiction.
[0022] One embodiment of the present application proposes a method for evaluating energy consumption in resource-constrained group agent communication, which is applied to an electronic device, where the electronic device can be a terminal or a server. In this embodiment and the following embodiments, the electronic device is described using a server as an example. The following describes the implementation details of the method for evaluating energy consumption in resource-constrained group agent communication proposed in this embodiment. The following content is merely provided for ease of understanding and is not required for implementing this solution.
[0023] The specific process of the resource-constrained group agent communication energy consumption evaluation method proposed in this embodiment can be as follows: Figure 1 Shown, including: Step 11, determining the communication protocol currently used by the resource-constrained group agent, wherein the communication protocol currently used by the resource-constrained group agent is a communication protocol with a communication frequency of 2.4 GHz, specifically a Wi-Fi protocol or a LoRa protocol.
[0024] In a specific implementation, the server's first step in evaluating communication energy consumption is to determine the communication protocol currently used by the resource-constrained swarm agent. This protocol typically operates at 2.4 GHz, specifically Wi-Fi or LoRa. This embodiment customizes energy consumption evaluation methods for different communication protocols to better meet practical needs.
[0025] Step 12: In the offline stage, the deep neural network is divided according to the layer type, and the energy consumption of each layer is calculated separately to realize the energy consumption modeling of the computational reasoning task.
[0026] In the specific implementation, in the offline stage, the server needs to be divided according to the type of layers of the deep neural network and calculate the energy consumption of each layer separately to realize the energy consumption modeling of the computational reasoning task.
[0027] It is understood that the inference network of a deep neural network contains numerous feature layers, and each neuron iteration involves performing element-wise multiplication between the filter and the input feature map, and accumulating all element-wise products to achieve activation in the output. Computational data is collected from the environment, and an energy consumption model can be established based on the model structure of the deep neural network. Compared with other indicators such as the energy consumption of sensors in data storage and transmission, it is very difficult to estimate the energy consumption of the model based on the layer structure due to its tightly coupled structure. The computational data flow will lead to uncertain computational results during the inference calculation process, and different computational factors will consume different amounts of energy. Therefore, this embodiment further divides the layers according to their types. Taking the convolutional layer as an example, the convolutional layer consists of a series of feature inputs and outputs. According to statistics, the energy consumption of the convolutional layer accounts for a large part of the energy consumption of the entire model. Calculating the energy consumption of the convolutional neural network layer can establish energy consumption based on the granularity of the deep neural network layer.
[0028] In one example, when implementing energy consumption modeling for a computational reasoning task, the server first needs to determine the overall energy consumption of the resource-constrained swarm agent. The server expresses the overall energy consumption of the resource-constrained swarm agent as: ; in, Indicates the size of the L1 memory of the resource-constrained swarm agent, represents the total memory size of the resource-constrained swarm agent, represents the computation frequency of resource-constrained swarm agents, represents the current execution frequency of the resource-constrained swarm agent, represents the rated operating frequency of resource-constrained swarm agents, represents the core frequency of the resource-constrained swarm agent, and They represent the memory energy consumption of L1 level and L2 level of resource-constrained swarm agents, Indicates overall energy consumption.
[0029] Next, the server will disassemble the deep neural network and determine the energy consumption layer by layer.
[0030] The server expresses the energy consumption of all convolutional layers of the deep neural network as: ; in, 、 、 Represent the height, width and number of channels of the input features respectively, represents the total number of convolutional layers in the deep neural network, Indicates the The energy consumption of the convolutional layer, and Respectively represent The height and width of the convolution kernel of the convolution layer, represents the number of cumulative multiplication operations in a deep neural network, 、 、 Represent the height, width and number of channels of the output features respectively, represents the energy consumption of the feature during transmission, Represents the energy consumption of all convolutional layers of a deep neural network.
[0031] The server expresses the energy consumption of the pooling and fully connected layers as: ; in, represents the pooling operation, represents the energy consumption of convolution in the fully connected layer, and Represent the height and width of the convolution kernel in the fully connected layer, represents the energy consumption of pooling and fully connected layers.
[0032] In the normalization layer, the step size and the size of the input feature are used as regression variables to predict the energy consumption of the normalization layer. The server expresses the prediction of the energy consumption of the normalization layer as: ; in, The ability to represent input features, represents the normalization operation, Represents the predicted value of the energy consumption of the normalization layer.
[0033] At this point, the energy consumption modeling of the offline phase has been completed.
[0034] Step 13: In the online stage, based on the communication protocol currently used by the resource-constrained group agent and the energy consumption modeling of the computational reasoning task established in the offline stage, combined with the current resource situation of the resource-constrained group agent, the corresponding queue matching optimization is performed to match the computational reasoning task to the most suitable resource-constrained group agent for computational reasoning, thereby completing the optimization evaluation of the communication energy consumption of the resource-constrained group agent.
[0035] In the specific implementation, in the online stage, the server designs a queue matching optimization method based on the communication protocol currently used by the resource-constrained group intelligent agent and the energy consumption modeling of the computational reasoning tasks established in the offline stage, combined with the current resource situation of the resource-constrained group intelligent agent, and performs corresponding queue matching optimization on the resource-constrained group intelligent agent. In this way, the computational reasoning tasks can be matched (dispatched) to the most suitable resource-constrained group intelligent agent for computational reasoning, thereby completing the optimization evaluation of the communication energy consumption of the resource-constrained group intelligent agent.
[0036] Whether using the Wi-Fi protocol or the LoRa protocol, the data transmission and computing energy consumption of the low-power network in each time slot, as well as the average energy consumption over all time slots, are calculated in the same way.
[0037] Assume that during the optimization process, the cycle of each computing task is divided into time slots, and the duration of each time slot is set to , assuming that there is a low-power network The group agents exchange data in each time slot On the swarm agent The amount of communication data exchanged is expressed as , the data transmission rate of the low-power network is .
[0038] It is expressed by the formula: ; in, is the preset weight coefficient, Indicates the transmit power, represents the channel gain, Represents the noise power.
[0039] Assuming that the amount of communication data of the group agent is proportional to the amount of computation data, let For the communication data volume The energy consumption required is calculated and the energy consumption coefficient is defined according to different networks and different CPU processing frequencies. , the computing resources of the swarm agent in each time slot are , the total energy consumption of network interaction and computational processing of each swarm agent is .
[0040] Based on this, the data transmission and computing energy consumption of the low-power network in each time slot is expressed as: ; in, Indicates time slot Data transmission and computing energy consumption on low-power networks.
[0041] The average energy consumption over all time slots is expressed as: ; in, represents the average energy consumption over all time slots, .
[0042] The following describes queue matching optimization for the Wi-Fi protocol and the LoRa protocol respectively.
[0043] For Wi-Fi protocols, the energy consumption of transmitting 1 bit of data is taken into account. , calculate the energy consumption of the current low-power network, and select the low-power network with the best communication quality to obtain the lowest energy consumption by comparing the energy consumption values.
[0044] In the data transmission process of low-power networks, a multiple-input multiple-output system is usually used, and its antenna is represented by a frequency domain model as follows: ; in, represents the received signal vector, represents the channel matrix, which contains the channel state information, represents the signal vector sent, represents the noise vector, which is usually assumed to be additive white Gaussian noise.
[0045] In the IEEE 802.11n standard, the signal is divided into 56 carrier channels with a width of 20 MHz, of which 30 carrier channels are represented as follows: ; in, is the total number of carriers.
[0046] No. The carrier channels are expressed as: ; in, and Represent the amplitude and phase of the wireless signal respectively.
[0047] During signal transmission, the fluctuation of the signal generates carrier frequency deviation, and the estimated The frequency of the carrier wave is: ; in, For the estimated The frequency of the carrier wave, For the actual The frequency of the carrier wave, Indicates the The index of the carrier, is the frequency offset, is the constant term of frequency offset, is the noise term.
[0048] The frequency error is: ; ; in, The slope of the frequency error, is the intercept of the frequency error.
[0049] The channel state information of the carrier can be expressed as a matrix , which means the The receiving antenna and The channel response between the transmit antennas is: .
[0050] Finally, the relationship between communication energy consumption and carrier amplitude and phase is established: ; in, Indicates the energy consumption under the Wi-Fi protocol, represents the carrier amplitude, represents the carrier phase, represents the Hadamard product, Represents element-wise addition.
[0051] For the Lora protocol, the energy consumption required to transmit one byte in the Lora protocol is Bring in , to calculate ; ; in, represents the total energy consumption, represents the total transmitted energy, Indicates the length of the data being transmitted. Indicates the transmission time of the entire transmission packet.
[0052] Specifically, it is the sum of the data prefix and the transmission payload length: ; ; in, Indicates the data prefix, Indicates the transmission payload length, represents the prefix code, Indicates the synchronization code period.
[0053] The total energy consumption of the swarm agent using the Lora protocol is expressed as: ; in, Indicates the total number of bytes, Represents the total energy consumption of swarm agents using the Lora protocol.
[0054] Under the Lora protocol, the time slot process of the swarm agent is divided into two states: swarm agent dormancy and activation. The total energy is expressed as: ; ; in, Represents the total energy of the time slot process of the group agent under the Lora protocol, Represents dormant energy, represents activation energy, 、 、 、 、 、 They represent the energy required for system wake-up, data transmission, data processing, system sleep to wake-up, system transmission mode, and receiving mode.
[0055] 、 、 、 、 Respectively expressed as: ; ; ; ; ; in, Indicates the energy consumption of a single operation of the system. Indicates the number of times the system has run. 、 、 and They represent the energy consumption of data transmission, the energy consumed by the system from sleep to wake-up, the energy consumed by data sending, and the energy consumed by data receiving. Indicates the time taken for data transmission. Indicates the time when the system sends data. Indicates the time when the system receives data.
[0056] ; in, and They represent the energy per bit of transmitted data and the number of bits transmitted, respectively.
[0057] In one example, we conducted actual tests and deployments on two protocols commonly used by resource-constrained swarm agents, Wi-Fi and LoRa low-power protocols. The test results are as follows: Figure 2 As shown in the figure, after testing swarm agent systems with different numbers of participants, it was determined that the overall energy consumption of the system was reduced by a maximum of 19.09%.
[0058] This embodiment proposes a method for evaluating the communication energy consumption of a resource-constrained group agent. First, the communication protocol currently used by the resource-constrained group agent is determined. The communication protocol currently used by the resource-constrained group agent is a communication protocol with a communication frequency of 2.4GHz, specifically the Wi-Fi protocol or the LoRa protocol. Next, the communication energy consumption of the resource-constrained group agent is evaluated. This process is divided into two stages, namely the offline stage and the online stage. In the offline stage, the energy consumption of the system operation is evaluated by modeling the energy consumption required for the offline stage tasks. In the online stage, an energy consumption resource matching model is designed. Based on the computing tasks and resource energy consumption in the offline stage, an optimization algorithm based on queue matching is designed to optimize the computing process and thereby enhance the reasoning computing performance of the low-power group agent (resource-constrained group agent). This application combines the current operating environment of the resource-constrained group agent to flexibly schedule computing tasks in real time, which can achieve more efficient and effective communication energy consumption evaluation of the group agent.
[0059] The steps of the various methods described above are divided for clarity of description only. During implementation, they can be combined into a single step, or some steps can be split into multiple steps. As long as they contain the same logical relationships, they are all within the scope of protection of this application. Adding minor modifications to the algorithm or process, or introducing minor design changes that do not change the core design of the algorithm or process, are also within the scope of protection of this application.
[0060] Another embodiment of the present application proposes a resource-constrained group intelligent agent communication energy consumption assessment system. The implementation details of the resource-constrained group intelligent agent communication energy consumption assessment system proposed in this embodiment are specifically described below. The following content is only the implementation details provided for easy understanding and is not necessary for the implementation of this solution.
[0061] The specific structure of the resource-constrained group agent communication energy consumption evaluation system proposed in this embodiment can be as follows: Figure 3 As shown, it includes: a basic setting module 21, an offline modeling module 22 and an online matching module 23.
[0062] The basic setting module 21 is used to determine the resource-constrained group agent and the communication protocol currently used by the resource-constrained group agent. The communication protocol currently used by the resource-constrained group agent is a communication protocol with a communication frequency of 2.4 GHz, specifically the Wi-Fi protocol or the LoRa protocol.
[0063] The offline modeling module 22 is used to divide the deep neural network according to the type of layer and calculate the energy consumption of each layer separately to realize energy consumption modeling of computational reasoning tasks.
[0064] The online matching module 23 is used to model the energy consumption of the computational reasoning tasks established in the offline stage based on the communication protocol currently used by the resource-constrained group intelligent agent and the current resource situation of the resource-constrained group intelligent agent, perform corresponding queue matching optimization, match the computational reasoning tasks to the most suitable resource-constrained group intelligent agent for computational reasoning, and thus complete the optimization evaluation of the communication energy consumption of the resource-constrained group intelligent agent.
[0065] It is worth mentioning that all modules and modules involved in this embodiment are logical modules. In actual applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. In addition, to highlight the innovation of this application, this embodiment does not introduce units that are not closely related to solving the technical problems proposed by this application. However, this does not mean that other units do not exist in this embodiment.
[0066] It is not difficult to find that this embodiment is a system embodiment corresponding to the above-mentioned method embodiment, and this embodiment can be implemented in conjunction with the above-mentioned method embodiment. The relevant technical details and technical effects mentioned in the above-mentioned method embodiment are still valid in this embodiment, and to reduce repetition, they are not repeated here. Accordingly, the relevant technical details mentioned in this embodiment can also be applied to the above-mentioned method embodiment.
[0067] Another embodiment of the present application provides an electronic device, such as Figure 4 As shown, it includes: at least one processor 31; and a memory 32 communicatively connected to the at least one processor 31; wherein the memory 32 stores instructions that can be executed by the at least one processor 31, and the instructions are executed by the at least one processor 31 to enable the at least one processor 31 to execute a resource-constrained group intelligent agent communication energy consumption evaluation method as described in the above method embodiment.
[0068] The memory and processor are connected using a bus, which can include any number of interconnected buses and bridges. The bus connects various circuits of one or more processors and memories. The bus can also connect various other circuits such as peripheral devices, voltage regulators, and power management circuits. These are all well known in the art and therefore will not be described further in this article. The bus interface provides an interface between the bus and the transceiver. The transceiver can be a single component or multiple components, such as multiple receivers and transmitters, providing a unit for communicating with various other devices on a transmission medium. Data processed by the processor is transmitted on a wireless medium via an antenna. Furthermore, the antenna also receives data and transmits it to the processor.
[0069] The processor is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory can be used to store data used by the processor when performing operations.
[0070] Another embodiment of the present application provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it can implement a resource-constrained group intelligent agent communication energy consumption evaluation method as described in the above method embodiment.
[0071] That is, those skilled in the art will understand that all or part of the steps in the above-described method embodiments can be implemented by instructing the relevant hardware through a program. The program is stored in a storage medium and includes a number of instructions for causing a device (which may be a single-chip microcomputer or chip) or a processor to execute all or part of the steps in the method embodiments described herein. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory, a random access memory, a magnetic disk, or an optical disk.
[0072] It will be understood by those skilled in the art that the above embodiments are specific embodiments for implementing the present application, and in actual applications, various changes may be made thereto in form and detail without departing from the spirit and scope of the present application.
Claims
1. A method for evaluating energy consumption of communication between resource-constrained group agents, characterized in that: include: Determine the communication protocol currently used by the resource-constrained group agent; wherein the communication protocol currently used by the resource-constrained group agent is a communication protocol with a 2.4 GHz communication frequency, specifically a Wi-Fi protocol or a LoRa protocol; In the offline phase, the deep neural network is divided according to its layer type and the energy consumption of each layer is calculated separately to achieve energy consumption modeling of computational inference tasks. In the online stage, based on the communication protocol currently used by the resource-constrained swarm agent and the energy consumption modeling of the computational reasoning tasks established in the offline stage, combined with the current resource situation of the resource-constrained swarm agent, corresponding queue matching optimization is performed to match the computational reasoning tasks to the most suitable resource-constrained swarm agent for computational reasoning, thereby completing the optimization evaluation of the communication energy consumption of the resource-constrained swarm agent.
2. A resource-constrained group agent communication energy consumption evaluation method according to claim 1, characterized in that: In the offline phase, the deep neural network is divided according to its layer type and the energy consumption of each layer is calculated separately to achieve energy consumption modeling of computational inference tasks, including: The overall energy consumption of resource-constrained swarm agents is expressed as: ； in, Indicates the size of the L1 memory of the resource-constrained swarm agent, represents the total memory size of the resource-constrained swarm agent, represents the computation frequency of resource-constrained swarm agents, represents the current execution frequency of the resource-constrained swarm agent, represents the rated operating frequency of resource-constrained swarm agents, represents the core frequency of the resource-constrained swarm agent, and They represent the memory energy consumption of L1 level and L2 level of resource-constrained swarm agents, Indicates overall energy consumption; The energy consumption of all convolutional layers of a deep neural network is expressed as: ； in, 、 、 Represent the height, width and number of channels of the input features respectively, represents the total number of convolutional layers in the deep neural network, Indicates the The energy consumption of the convolutional layer, and Respectively represent The height and width of the convolution kernel of the convolution layer, represents the number of cumulative multiplication operations in a deep neural network, 、 、 Represent the height, width and number of channels of the output features respectively, represents the energy consumption of the feature during transmission, Represents the energy consumption of all convolutional layers of a deep neural network; The energy consumption of the pooling and fully connected layers is expressed as: ； in, represents the pooling operation, represents the energy consumption of convolution in the fully connected layer, and Represent the height and width of the convolution kernel in the fully connected layer, represents the energy consumption of pooling and fully connected layers; In the normalization layer, the step size and the size of the input feature are used as regression variables to predict the energy consumption of the normalization layer. The prediction of the energy consumption of the normalization layer is expressed as: ； in, The ability to represent input features, represents the normalization operation, Represents the predicted value of the energy consumption of the normalization layer.
3. The method for evaluating communication energy consumption of resource-constrained group agents according to claim 2, wherein: Assume that during the optimization process, the cycle of each computing task is divided into time slots, and the duration of each time slot is set to , assuming that there is a low-power network The group agents exchange data in each time slot On the swarm agent The amount of communication data exchanged is expressed as , the data transmission rate of the low-power network is ; It is expressed by the formula: ； in, is the preset weight coefficient, Indicates the transmit power, represents the channel gain, represents the noise power; Assuming that the amount of communication data of the group agent is proportional to the amount of computation data, let For the communication data volume The energy consumption required is calculated and the energy consumption coefficient is defined according to different networks and different CPU processing frequencies. , the computing resources of the swarm agent in each time slot are , the total energy consumption of network interaction and computational processing of each swarm agent is ; Based on this, the data transmission and computing energy consumption of the low-power network in each time slot is expressed as: ； in, Indicates time slot Data transmission and computing energy consumption on low-power networks; The average energy consumption over all time slots is expressed as: ； in, represents the average energy consumption over all time slots, .
4. The method for evaluating energy consumption of resource-constrained group agent communication according to claim 3, wherein: For Wi-Fi protocols, the energy consumption of transmitting 1 bit of data is taken into account. , calculate the energy consumption of the current low-power network, and select the low-power network with the best communication quality to obtain the lowest energy consumption by comparing the energy consumption values; In the data transmission process of low-power networks, a multiple-input multiple-output system is usually used, and its antenna is represented by a frequency domain model as follows: ； in, represents the received signal vector, represents the channel matrix, which contains the channel state information, represents the signal vector sent, represents the noise vector, which is usually assumed to be additive Gaussian white noise; In the IEEE 802.11n standard, the signal is divided into 56 carrier channels with a width of 20 MHz, of which 30 carrier channels are represented as follows: ； in, is the total number of carriers; No. The carrier channels are expressed as: ; in, and Represent the amplitude and phase of the wireless signal respectively; During signal transmission, the fluctuation of the signal generates carrier frequency deviation, and the estimated The frequency of the carrier wave is: ； in, For the estimated The frequency of the carrier wave, For the actual The frequency of the carrier wave, Indicates the The index of the carrier, is the frequency offset, is the constant term of frequency offset, is the noise term; The frequency error is: ； ； in, The slope of the frequency error, is the intercept of the frequency error; The channel state information of the carrier can be expressed as a matrix , which means the The receiving antenna and The channel response between the transmit antennas is: ； Finally, the relationship between communication energy consumption and carrier amplitude and phase is established: ； in, Indicates the energy consumption under the Wi-Fi protocol, represents the carrier amplitude, represents the carrier phase, represents the Hadamard product, Represents element-wise addition.
5. The method for evaluating energy consumption of resource-constrained group agent communication according to claim 3, characterized in that: For the Lora protocol, the energy consumption required to transmit one byte in the Lora protocol is Bring in , to calculate ; ； in, represents the total energy consumption, represents the total transmitted energy, Indicates the length of the data being transmitted. Indicates the transmission time of the entire transmission packet; Specifically, it is the sum of the data prefix and the transmission payload length: ； ； in, Indicates the data prefix, Indicates the transmission payload length, represents the prefix code, Indicates the synchronization code period; The total energy consumption of the swarm agent using the Lora protocol is expressed as: ； in, Indicates the total number of bytes, Represents the total energy consumption of the swarm agents using the Lora protocol; Under the Lora protocol, the time slot process of the swarm agent is divided into two states: swarm agent dormancy and activation. The total energy is expressed as: ； ； in, Represents the total energy of the time slot process of the group agent under the Lora protocol, Represents dormant energy, represents activation energy, 、 、 、 、 、 They represent the energy required for system wake-up, data transmission, data processing, system sleep to wake-up, system transmission mode, and reception mode. 、 、 、 、 Respectively expressed as: ； ； ； ； ； in, Indicates the energy consumption of a single operation of the system. Indicates the number of times the system has run. 、 、 and They represent the energy consumption of data transmission, the energy consumed by the system from sleep to wake-up, the energy consumed by data sending, and the energy consumed by data receiving. Indicates the time taken for data transmission. Indicates the time when the system sends data. Indicates the time when the system receives data; ； in, and They represent the energy per bit of transmitted data and the number of bits transmitted, respectively.
6. A resource-constrained group agent communication energy consumption evaluation system, characterized in that: include: A basic setting module is used to determine the resource-constrained group agent and the communication protocol currently used by the resource-constrained group agent. The communication protocol currently used by the resource-constrained group agent is a communication protocol with a communication frequency of 2.4 GHz, specifically the Wi-Fi protocol or the LoRa protocol; The offline modeling module is used to divide the deep neural network according to its layer type and calculate the energy consumption of each layer to achieve energy consumption modeling of computational inference tasks; The online matching module is used to model the energy consumption of the computational reasoning tasks established in the offline stage based on the communication protocol currently used by the resource-constrained swarm agent and the current resource situation of the resource-constrained swarm agent, and perform corresponding queue matching optimization to match the computational reasoning tasks to the most suitable resource-constrained swarm agent for computational reasoning, thereby completing the optimization evaluation of the communication energy consumption of the resource-constrained swarm agent.
7. An electronic device, characterized in that: include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the resource-constrained group agent communication energy consumption evaluation method according to any one of claims 1 to 5.
8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, it can implement a resource-constrained group agent communication energy consumption evaluation method according to any one of claims 1 to 5.