Semantic-driven unmanned device communication and computing resource dynamic allocation method and device

Abstract

The application provides a semantic-driven unmanned equipment communication and computing resource dynamic allocation method and device, the method comprising: constructing a joint system model of a multi-semantic coding node transmission task cooperation of a communication and computing resource isomerism; determining the wireless communication condition and semantic compression rate of each semantic coding node according to the system model to realize dynamic control of the transmission data scale; taking the wireless communication condition and semantic compression rate of the multi-semantic coding node as the variable of the transmission data, and constructing an optimization objective function with the minimum overall task completion time of the multi-semantic coding node cooperative task; iteratively solving the objective function of the joint optimization problem through a preset solving strategy to obtain an optimal computing and communication resource allocation scheme that satisfies the semantic available condition; and the application can realize unified modeling and joint optimization of computing resources and communication resources under a dynamic channel environment and a heterogeneous computing condition, and improve the real-time performance and reliability of multi-unmanned equipment cooperative operation.

Description

Technical Field

[0001] This invention relates to the field of communication resource scheduling technology, and in particular to a semantically driven method, apparatus, device, and medium for dynamic allocation of communication and computing resources for unmanned devices. Background Technology

[0002] To achieve efficient centralized control and resource management of multiple unmanned devices, a star-shaped centralized network of "single base station + multiple unmanned devices" has become the mainstream architecture. In this architecture, a ground base station (or edge server) acts as the central node, responsible for communicating with multiple unmanned device terminals. All unmanned devices transmit data back to the base station through a shared uplink wireless channel, and the base station uses its global information (such as channel status and task requirements) to perform unified resource scheduling and coordination.

[0003] In existing technological systems, multi-unmanned equipment systems generally adopt the traditional communication mode of "image / video compression + wireless backhaul + cloud inference," which means that pixel-level data is the core transmission object, and the original data is uploaded completely after compression and encoding. However, the uplink data volume of a single unmanned device is still large. In scenarios where multiple unmanned devices access concurrently, the centralized backhaul of large-scale image or video data will lead to wireless link congestion, significantly increasing the overall system transmission latency, making it difficult to meet the business needs of emergency monitoring, dynamic inspection, and other applications with high real-time requirements.

[0004] In recent years, with the development of semantic communication and generative artificial intelligence technologies, the academic community has proposed replacing traditional pixel-level data transmission with the transmission of semantic feature vectors, semantic compression results, or high-level semantic descriptions, thereby maintaining the effectiveness of the information required for the task at extremely low bit rates. Semantic communication emphasizes task performance (such as detection accuracy, recognition accuracy, or semantic consistency) as the core indicator, providing a new path for unmanned equipment communication systems to break through the traditional Shannon communication paradigm.

[0005] However, existing semantic communication research mostly focuses on single-node or point-to-point scenarios, lacking system-level resource optimization methods for collaborative systems with multiple communication objects or multiple semantic coding nodes. Especially in practical engineering scenarios, how can we provide a technical solution that, while ensuring the quality of semantic reconstruction, can balance real-time performance, energy efficiency, and task reliability under dynamic channel environments and heterogeneous computing conditions, achieving unified modeling and joint optimization of computing and communication resources? Summary of the Invention

[0006] To overcome the problems existing in related technologies, this disclosure provides a semantically driven method, apparatus, device and medium for dynamic allocation of communication and computing resources for unmanned devices, in order to solve the technical problems that are common in multi-unmanned device communication technologies, such as large data scale, high semantic redundancy, fragmented optimization of computing and communication, lack of adaptability in resource allocation and lack of guarantee of semantic quality.

[0007] This specification provides one or more embodiments of a semantically driven method for dynamically allocating communication and computing resources for unmanned devices, including the following steps: A joint system model for collaborative transmission tasks among multiple semantic coding nodes with heterogeneous communication and computing resources is constructed. Based on the system model, the wireless communication conditions and semantic compression rate of each semantic coding node are determined to achieve dynamic control over the scale of transmitted data. The wireless communication conditions and semantic compression rate of the multi-semantic coding nodes are used as variables to control the transmitted data. The optimization objective function and its constraints are constructed by minimizing the overall maximum task completion time of the collaborative task of the multi-semantic coding nodes in the system. By iteratively solving the objective function of the joint optimization problem using a pre-defined solution strategy, the optimal allocation scheme of computing and communication resources that satisfies the semantic availability condition is obtained.

[0008] This specification provides one or more embodiments of a semantically driven dynamic allocation device for communication and computing resources of unmanned devices, including: The system model building module is used to construct a joint system model for the collaborative transmission tasks of multiple semantic coding nodes with heterogeneous communication and computing resources. Based on the system model, the wireless communication conditions and semantic compression rate of each semantic coding node are determined to achieve dynamic control of the scale of transmitted data. The optimization function construction module is used to construct the optimization objective function and its constraints by using the wireless communication conditions and semantic compression rate of the multi-semantic coding nodes as variables to control the transmitted data, and minimizing the overall maximum task completion time of the collaborative task of the multi-semantic coding nodes in the system. The solution module is used to iteratively solve the objective function of the joint optimization problem using a preset solution strategy to obtain the optimal computing and communication resource allocation scheme that satisfies the semantic availability conditions.

[0009] This specification provides one or more embodiments of a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the semantically driven method for dynamic allocation of communication and computing resources for unmanned devices as described above.

[0010] This specification provides one or more embodiments of a computer-readable storage medium storing a computer program that, when executed by a processor, implements the semantically driven method for dynamic allocation of communication and computing resources for unmanned devices as described above.

[0011] This disclosure provides a semantically driven method, apparatus, device, and medium for dynamic allocation of communication and computing resources for unmanned equipment. The advantages of this method lie in its consideration of constructing a multi-semantic coding node collaborative scenario system model for different nodes. This model achieves dynamic control of the transmitted data scale through adaptive adjustment of semantic compression rate and wireless communication conditions. Furthermore, with minimizing the maximum system task completion time as the core optimization objective, it coordinates and schedules computing and communication resources to optimize overall system performance. This method achieves adaptive adjustment of the transmitted data scale using semantic compression rate, transmitting semantic feature information directly related to the task, and realizing joint adaptive allocation of semantic compression rate and wireless communication conditions. Compared to traditional methods targeting average latency, this significantly shortens the overall completion time of collaborative tasks. It enables unified modeling and joint optimization of computing and communication resources in dynamic channel environments and heterogeneous computing conditions, improving the real-time performance, reliability, and energy efficiency of multi-unmanned equipment collaborative operations at the system level. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in one or more embodiments of this specification or in the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A flowchart illustrating a semantically driven dynamic allocation method for communication and computing resources of unmanned devices, provided for one or more embodiments of this specification; Figure 2 This is a schematic diagram of the solution process provided in one or more embodiments of this specification; Figure 3 A schematic diagram of the optimized resource allocation result obtained after the algorithm converges in a simulation experiment provided for one or more embodiments of this specification; Figure 4 A block diagram of a semantically driven dynamic allocation device for communication and computing resources of unmanned equipment, provided for one or more embodiments of this specification; Figure 5 This is a schematic diagram of the structure of a computer device provided for one or more embodiments of this specification. Detailed Implementation

[0014] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this invention.

[0015] This invention addresses the challenges of collaborative semantic communication scenarios involving multiple semantic coding nodes. It tackles the bottlenecks of limited communication resources, computational capabilities, and latency in collaborative tasks faced by heterogeneous nodes in complex wireless environments. The invention proposes a multi-node joint resource optimization method based on adaptive adjustment of semantic compression rate. This method is independent of specific semantic coding networks or wireless protocols and is applicable to various collaborative image or video transmission scenarios involving multiple UAVs, vehicles, and autonomous vehicles. It can be flexibly deployed on edge computing platforms, roadside units, or central servers, and can also be deployed in a distributed or semi-distributed manner among collaborative nodes. The generation and distribution of scheduling results do not affect the core technical effectiveness of this invention. It possesses good versatility and promotional value. The invention will be described in detail below with reference to specific embodiments and accompanying drawings.

[0016] Method Implementation Examples According to embodiments of the present invention, a semantically driven method for dynamic allocation of communication and computing resources for unmanned devices is provided, such as... Figure 1 The diagram shown is a flowchart of a semantically driven dynamic allocation method for communication and computing resources of unmanned devices provided in this embodiment. The semantically driven dynamic allocation method for communication and computing resources of unmanned devices according to this embodiment includes the following steps: Step S1: Construct a joint system model for collaborative transmission tasks of multiple semantic coding nodes with heterogeneous communication and computing resources. Determine the wireless communication conditions and semantic compression rate of each semantic coding node based on the system model to achieve dynamic control of the scale of transmitted data. Step S2: The wireless communication conditions and semantic compression rate of the multi-semantic coding nodes are used as variables to control the transmitted data. The optimization objective function and the constraints of the optimization objective function are constructed by minimizing the overall maximum task completion time of the multi-semantic coding node collaborative task, including the transmission task constraints. Step S3: Iteratively solve the objective function of the joint optimization problem using a preset solution strategy to obtain the optimal computing and communication resource allocation scheme that satisfies the semantic availability condition.

[0017] This embodiment provides a semantically driven dynamic allocation method for communication and computing resources in unmanned devices. It considers constructing a multi-semantic coding node collaborative scenario system model for different nodes with varying characteristics. This model achieves dynamic control of the transmitted data scale through adaptive adjustment of semantic compression ratio and wireless communication conditions. Furthermore, with minimizing the system's maximum task completion time as the core optimization objective, it coordinates and schedules computing and communication resources to optimize overall system performance. This method achieves adaptive adjustment of the transmitted data scale using semantic compression ratio, transmitting semantic feature information directly related to the task, thus realizing joint adaptive adjustment of semantic compression ratio and wireless communication conditions. Compared to transmitting raw images or point cloud data, the data size can be reduced by more than an order of magnitude, effectively alleviating link congestion in scenarios with concurrent access from multiple nodes, improving system scalability and stability. Furthermore, the objective function constrains the performance of the slowest node, avoiding the "barrel effect" where individual nodes slow down the overall task progress. Compared to traditional methods that target average latency, it can significantly shorten the overall completion time of collaborative tasks. It can also achieve unified modeling and joint optimization of computing and communication resources in dynamic channel environments and heterogeneous computing conditions, improving the real-time performance, reliability, and energy efficiency of collaborative operations of multiple unmanned devices at the system level.

[0018] In this embodiment, the joint system model performs full-dimensional modeling of the computation process of semantic coding nodes and the communication process of image transmission, including core indicators in two dimensions: time and energy consumption. It also constructs constraint models such as semantic quality, hardware, and total resources in combination with actual application requirements, thus laying the foundation for optimization problems.

[0019] In this embodiment, the joint system model includes an encoding time model and an encoding energy consumption model that consider the semantic encoding process of multiple semantic coding nodes. These models quantify the time and energy consumption of the encoding process, with energy consumption strongly correlated with semantic compression ratio and image complexity. The model also considers the wireless transmission process of the encoded image, using a transmission time model based on Shannon's theorem and a transmission energy consumption model combined with the energy consumption characteristics of the radio frequency front-end to quantify the time and energy consumption of the transmission process. Furthermore, energy consumption is strongly correlated with semantic compression ratio, bandwidth, transmit power, and channel conditions. The joint system model is detailed below: (1) Semantic coding time model: Based on the number of images of multiple semantic coding nodes, the number of CPU cycles required per image and the working frequency of the coding calculation unit of the semantic coding node, a semantic coding time model of multiple semantic coding nodes is constructed.

[0020] Let the first The total number of CPU cycles required to encode a single image using semantic coding nodes is: (Unit: cycles). This value depends on the complexity of the semantic encoding algorithm used and the complexity of the image content. For example, higher compression ratio requirements or more complex scenes typically require more computation cycles. The more, the more we assume that the encoding computation units of all semantic encoding nodes operate at a fixed frequency. f (Unit: Hz) then, the computation time required for this semantic coding node to complete the encoding. as follows: ; in, For the first The number of images that a semantic coding node needs to encode.

[0021] (2) Semantic coding energy consumption model: characterizes the energy consumption of the semantic coding node during the coding process. Specifically, based on the classic dynamic power consumption model of digital CMOS integrated circuits, the energy consumption is calculated to be proportional to the square of the operating frequency and the amount of switching activity. set up Let be the effective switching capacitor coefficient of the processor chip, which is a constant related to the chip manufacturing process. semantic encoding node The processor coding energy efficiency coefficient (dimensionless) is used to comprehensively reflect the processor's actual energy consumption efficiency under a specific operating voltage. Then, the semantic coding node... Computational energy consumption during encoding As shown in the following formula:

[0022] (3) Transmission time model: Based on the unit image data volume, semantic compression rate of semantic coding nodes, allocated communication bandwidth and transmission power, it is constructed using Shannon's theorem; After semantic encoding, the data is compressed and transmitted to the base station via a wireless channel. The transmission time and energy consumed in this process are as follows, assuming the data size of a single original image is... (Unit: bits). Semantic coding node The semantic compression rate used is It represents the ratio of the compressed data size to the original data size, i.e. Therefore, the amount of data that needs to be transmitted is Bits. Assume the system is a semantic encoding node. The allocated communication bandwidth is (Unit: Hz). Semantic coding node With transmission power (Unit: W) is used for transmission. Assume its channel power gain to the base station is... (Dimensionless, incorporating path loss and fading), the power spectral density of the wireless channel noise is: (Unit: W / Hz). According to Shannon's theorem, semantic coding nodes... achievable data transmission rate (Unit: bps) is:

[0023] in, Let S be the signal-to-noise ratio at the receiving end; then, the transmission... Transmission time required for bit data (Unit: s) is the data volume divided by the rate: .

[0024] (4) Transmission Energy Consumption Model: A transmission energy consumption model is constructed based on the transmit power and transmission time of the semantic coding node; during transmission, the radio frequency front-end of the semantic coding node consumes energy. (Unit: J) Approximately equal to the product of transmit power and transmission time: .

[0025] In summary, semantic encoding nodes Total processing time for completing the entire image acquisition, encoding, and transmission task for:

[0026] Semantic encoding nodes Total energy consumed to complete the entire task for: .

[0027] In this embodiment, the wireless communication conditions and semantic compression rate of the multiple semantic coding nodes are used as control variables in the objective function. The wireless communication conditions may include communication bandwidth, transmission power, and other parameters characterizing the communication or computation of the semantic coding nodes. The semantic compression rate can also be replaced by other equivalent control parameters as needed, such as semantic feature dimension, number of channels, bit rate, or number of coding layers. The pruning ratio, quantization accuracy, or feature selection ratio of the coding network can be selected from the above-mentioned joint optimization variables as needed. The optimization objective function constructed in this embodiment is explained below using communication bandwidth, transmission power, and semantic compression rate as joint optimization variables.

[0028] In this preferred embodiment, the adjustment method for semantic compression ratio is broadly protected, including but not limited to: semantic compression ratio setting methods with continuous or discrete values, and methods for determining semantic compression ratio based on analytical models, lookup tables, or learning models. The core adjustable decision variable in this embodiment is the triplet, which is adaptively allocated for each semantic coding node. X={ , , )|i=1,2,...,N}; in, semantic encoding node semantic compression rate, semantic encoding node Communication bandwidth, semantic encoding node The transmission power.

[0029] (II) Optimization Objectives Based on the aforementioned computation-communication joint model and all-dimensional constraints, an optimization problem is constructed with minimizing the maximum task completion time of the system as its core. The optimization variables, objective function, and constraint set are clearly defined. The goal is to minimize the maximum task completion time of the system, which is the task time of the semantic encoding node that completes the entire task latest. Constraints include total processing time constraints, energy budget constraints, maximum transmit power constraints, bandwidth resource constraints, and semantic compression rate range constraints. The specific objective function and constraint set are as follows:

[0030]

[0031]

[0032]

[0033] ; Among them, the maximum transmit power constraint condition constrains the semantic coding node. Transmission power It cannot exceed the maximum value supported by its RF front-end hardware. The semantic compression ratio range constraint is that the semantic compression ratio of the semantic coding node cannot be lower than the minimum compression ratio. And uncompressed or lossless compression ,Right now The bandwidth resource constraint table stipulates that the sum of the bandwidth allocated to all semantic coding nodes cannot exceed the total bandwidth. , This represents the total energy consumption for the entire mission.

[0034] In this embodiment, the constraints of the transmission task can be indicators that evaluate the semantic quality of the transmission, such as semantic reconstruction quality, task success rate, semantic similarity, or downstream control error. These evaluation indicators can also be based on the accuracy indicators obtained from the analysis of target detection or semantic segmentation tasks. The following uses semantic reconstruction quality as a constraint of the transmission task to illustrate this condition.

[0035] In this implementation example, in an edge monitoring scenario, the ultimate goal of transmitting data is to reconstruct images at the base station with high quality for subsequent tasks such as target detection, recognition, or scene analysis. The quality of the reconstructed image is measured by the peak signal-to-noise ratio (PSNR). The higher the PSNR value, the smaller the difference between the reconstructed image and the original image, and the better the quality.

[0036] However, image quality is affected by two key factors: one is the semantic compression rate of the semantic coding nodes; excessive compression, i.e. If the signal-to-noise ratio is too low, it will introduce irreversible information loss; secondly, it will cause transmission distortion, which is affected by the quality of the wireless channel and is measured by the signal-to-noise ratio at the receiving end. Characteristically, the lower the signal-to-noise ratio, the higher the bit error rate and the worse the reconstruction quality.

[0037] Based on experimental testing of a specific semantic encoder under various compression rates and channel conditions, an empirical relationship can be determined between the PSNR value of the reconstructed image and the two factors mentioned above, which is fitted by the following logarithmic model:

[0038] in, The coefficients (dimensionless) were obtained by fitting offline experimental data. This model shows that PSNR is related to compression ratio. A linear combination of signal-to-noise ratio The value is proportional to the logarithm of the signal-to-noise ratio. This means that increasing the compression ratio or improving the signal-to-noise ratio can improve image quality, and the two can compensate for each other to some extent.

[0039] To ensure the reliability of subsequent intelligent analysis tasks, the system requires that the PSNR of the reconstructed image must not be lower than a preset minimum quality threshold. (Unit: dB). Therefore, the semantic reconstruction quality constraint is as follows: ; For ease of subsequent optimization, the logarithmic inequality is transformed into a linear form. Define a constant: ; because >0 is equivalent to the following formula: .

[0040] In this embodiment, the objective function (P) is a complex non-convex optimization problem. The non-convexity primarily stems from the transmission energy consumption function in the constraints, which is related to the variables... and It is non-convex. Furthermore, the variable T and... , The problem is highly coupled, making it computationally difficult to directly find the global optimum, especially when the number of semantic coding nodes is large. Therefore, an efficient and scalable algorithm is needed to solve it.

[0041] In this embodiment, the solution method for the non-convex problem can use hierarchical solution algorithms, heuristic search algorithms, genetic algorithms, particle swarm optimization algorithms, Lagrange dual decomposition or alternating iterative optimization methods, as well as approximate decision models based on deep reinforcement learning or neural network regression.

[0042] In one embodiment, the solution of the optimization function is illustrated using a solution framework that employs binary search and feasibility determination. The core of this method is that the outer binary search performs a binary search on the system's maximum task time T. max The process involves probing and an inner-layer feasibility check, which breaks down the complex problem into sub-problems with single semantically encoded nodes. These sub-problems are then solved using a one-dimensional numerical search, ultimately approximating the global near-optimal solution to problem (P). The specific solution approach is as follows: Outer layer: for T max A binary search is performed within the range of values ​​T, exploring a target value T at a time to determine if there is a feasible resource allocation scheme X that reduces the task time of all semantic encoding nodes. That is, feasibility assessment; Inner layer: The feasibility judgment problem is decomposed into N independent single semantic coding node sub-problems. Through the substitution and transformation of each variable, the sub-problems are transformed into one-dimensional numerical search problems, which can be solved efficiently and the feasibility can be verified. Iteration: By continuously narrowing T max By dividing the interval into two parts, the smallest feasible T is obtained, which is the optimal solution to the optimization problem (P), and the corresponding resource allocation scheme X is the optimal resource allocation scheme.

[0043] refer to Figure 2 The diagram below illustrates the solution process provided in this embodiment. The specific solution process is as follows: Step 31, Building the outer binary search framework: including determining T max The initial search interval, interval binary search, interval update, and termination condition determination are iteratively performed; the specific steps are as follows: Step A1: Enter all system parameters Parameters of each semantic encoding node and convergence accuracy ; Step A2: Calculation If there exists a certain Make If the problem is not feasible, the algorithm terminates. Step A3: Calculate the lower bound Set an upper bound T. high For a sufficiently large value, ensure |T high T low |<ε; Call feasibility assessment algorithm A:T high If "Feasible == False", then T will be in this state. high (or T) low The optimal solution T to problem (P) is... max If no feasible solution is found within the limit of attempts, the algorithm terminates. Step A4: Let T mid = (T low +T high ) / 2, call feasibility judgment algorithm A: .

[0044] Step A5: If "Feasible == True", let If "Feasible == False", let If Proceed to step A4; Step A6: Output the optimal value And the corresponding (approximate) optimal resource allocation scheme: bandwidth allocation Compression ratio Semantic coding node transmit power .

[0045] Step 32, Inner layer feasibility check: For the target value T of the outer layer probe, the core issue of feasibility checking is: Does a resource allocation scheme X={ , , ), satisfying all constraints of problem (P), and The main core ideas are as follows.

[0046] Determine whether, for a given T, there exists a decision variable { , , Satisfying the constraints is equivalent to determining whether the optimal value of problem (P1) is less than 1 / 2. Furthermore, problem (P1) can be broken down into N independent subproblems:

[0047]

[0048]

[0049]

[0050] .

[0051] The problem becomes a solution Each sub-problem yields its optimal value. ,calculate and Make a comparison. If satisfied... If T is feasible, it means that there exists a bandwidth allocation scheme that makes all semantic coding nodes feasible within time T, thus T is globally feasible; otherwise, T is not feasible.

[0052] Therefore, the core of the inner-layer feasibility check is transformed into solving N parallel subproblems of single semantic coding nodes with identical structures. For semantic coding nodes... To simplify notation, the subscript i will be temporarily omitted when describing the subproblem below. A new intermediate variable—power spectral density—is defined:

[0053] Furthermore, define a function L(y) of y, which represents the spectral efficiency per unit bandwidth: ; Note that L(y) is a concave function of y and is monotonically increasing. Using the new variable y and the function L(y), the problem can be transformed into the following:

[0054]

[0055]

[0056]

[0057] .

[0058] Observation reveals that when (P2) is feasible, the constraints It must be a tight constraint, i.e., the optimal solution. Definitely satisfy .

[0059] Therefore, problem (P2) can be transformed into an optimization problem involving only two variables:

[0060]

[0061]

[0062]

[0063]

[0064] To simplify the analysis, the following assumptions are made. Constraints It is redundant. Observation reveals that for any given... y ,variable x The feasible region is represented as follows:

[0065] Note the objective function with respect to the variable x It is a linear function with coefficients greater than 0. Therefore, when problem (P2) is feasible, the optimal value is minimized. x Obtained from [location]. Therefore, the question becomes about [location]. y One-dimensional search problem: remember And agreed to the constraint When not satisfied, Therefore, the problem is equivalent to finding the interval y > 0 that minimizes F(y). The solution can be found through an efficient numerical search in one-dimensional space y > 0. Although the function F(y) appears complex, its calculation only involves basic elementary functions, namely logarithms, linear functions, maximum and minimum values, thus the calculation speed is very fast.

[0066] Based on the above analysis, step S32 specifically includes the following steps: Step B1: Obtain all system parameters for input Semantic encoding node parameters Time T and convergence accuracy ; Step B2: For each semantic encoding node ,calculate and ,make ; Step B3: Determine the search range : Take a small positive number, such as W / Hz. Upper bound .

[0067] Step B4: In the interval Above, we use efficient, derivative-free numerical optimization methods to find the minimum point. For example: the Golden-Section Search, Brent's Method, to find... .

[0068] Step B5: Determine if If the algorithm returns "Feasible=False", it will terminate.

[0069] Step B6: Calculation ,like If the algorithm returns "Feasible=True", the bandwidth allocation, compression ratio, and semantic coding node transmit power are set, and the algorithm terminates. Otherwise, if the algorithm returns "Feasible=False", it terminates.

[0070] The advantages of the semantically driven dynamic allocation method for communication and computing resources of unmanned equipment provided in this embodiment are illustrated below through specific simulation experiments.

[0071] This experiment involved a monitoring system for 100 drones, randomly distributed within a range of 100m-10km from a base station. The base station's transmission frequency was 4.9GHz, and the total system bandwidth was 10MHz. The key parameter settings for each drone were as follows: [Number of calculation cycles is missing]. Uniformly distributed across cycles, the channel gain follows Rayleigh fading with an average value of 0.01; maximum transmit power is 2 W; energy budget is 5 J. Single raw images are uniformly distributed within the range of 2-6 MB, with a fixed computation frequency of 1 GHz. Image quality parameters are set to... , , , Minimum quality requirement: 35 dB. Compression ratio range: 0.1-1. Noise power spectral density. W / Hz.

[0072] Figure 3 The diagram illustrates the optimized resource allocation results obtained after algorithm convergence. Three figures sequentially present the specific results of bandwidth allocation, semantic compression rate allocation, and transmit power allocation. The results show that the algorithm differentiates bandwidth and semantic compression rate allocation based on the channel conditions and computational requirements of each UAV. UAVs with poor channel conditions are typically allocated more bandwidth and a higher compression rate to compensate for their low signal-to-noise ratio and reduce the amount of data to be transmitted, ensuring transmission can be completed within the specified time. UAVs with better channel conditions are allocated less bandwidth and a lower compression rate because they can achieve higher transmission rates even with lower bandwidth. The transmit power allocation shows that the transmit power of all UAVs is close to 2W, indicating that under the given parameters, the energy consumption required for UAV computation and transmission is low, therefore the power is close to the maximum transmit power.

[0073] The semantic-driven dynamic allocation method for communication and computing resources of unmanned devices provided in this embodiment has the following advantages: (1) The local semantic encoding calculation process of the semantic encoding node and the wireless transmission process are modeled in a unified manner, and the semantic compression rate is introduced as one of the system resource scheduling decision variables, which together with the communication bandwidth and transmission power constitute a set of joint optimization variables. The semantic compression rate is used to control the scale of the transmitted data after semantic encoding and directly participates in the construction of the system objective function and constraints, realizing the unified scheduling of the calculation process and the wireless transmission process; it breaks through the limitation of the semantic compression rate as a fixed parameter or offline adjustment parameter in the existing technology, and makes it have network-level scheduling attributes. The scale of transmitted data is adaptively adjusted by the semantic compression rate, and only semantic feature information directly related to the task is transmitted. Compared with transmitting the original image or point cloud data, the data scale can be reduced by more than an order of magnitude, thereby effectively alleviating the link congestion problem in the scenario of multiple nodes concurrent access and improving the system scalability and stability.

[0074] (2) By establishing a joint model between semantic compression rate, computation delay and transmission delay, computing resources and communication resources are coordinated and scheduled within a unified optimization framework to achieve optimal overall system performance.

[0075] (3) By using the maximum completion time of the system as the optimization target, the present invention can focus on constraining the performance of the slowest node and avoid the "barrel effect" of slowing down the overall task progress due to individual nodes. Compared with the traditional method that targets average latency, it can significantly shorten the overall completion time of collaborative tasks.

[0076] (4) The present invention introduces a semantic reconstruction quality constraint mechanism, which maps the semantic compression rate and wireless communication conditions together into a semantic quality index. During the optimization process, the minimum semantic quality requirement is explicitly constrained, thereby avoiding the problem of result distortion caused by excessive compression or link deterioration in existing methods.

[0077] (5) In view of the differences in computing power, energy status and channel conditions of different nodes, the present invention adopts an adaptive resource allocation strategy to allocate different semantic compression rate, bandwidth and power parameters to different nodes, which can dynamically adapt to complex and changing environments and significantly improve the robustness and engineering feasibility of the system in real scenarios.

[0078] Device Examples According to embodiments of the present invention, a semantically driven dynamic allocation device for communication and computing resources of unmanned equipment is provided, such as... Figure 4 The diagram shown is a block diagram of a semantically driven dynamic allocation device for communication and computing resources of unmanned devices provided in this embodiment. According to an embodiment of the present invention, the semantically driven dynamic allocation device for communication and computing resources of unmanned devices includes: The system model construction module 10 is used to construct a joint system model for the collaborative transmission tasks of multiple semantic coding nodes with heterogeneous communication and computing resources. Based on the system model, the wireless communication conditions and semantic compression rate of each semantic coding node are determined to achieve dynamic control of the scale of transmitted data. The optimization function construction module 20 is used to construct the optimization objective function and the constraints of the optimization objective function, including the transmission task constraints, by using the wireless communication conditions and semantic compression rate of the multi-semantic coding nodes as variables to control the transmitted data, and by minimizing the overall maximum task completion time of the collaborative task of the multi-semantic coding nodes in the system. The solver module 30 is used to iteratively solve the objective function of the joint optimization problem through a preset solver strategy to obtain the optimal computing and communication resource allocation scheme that satisfies the semantic availability condition.

[0079] The semantically driven unmanned device communication and computing resource dynamic allocation device provided in this embodiment has a system model construction module 10 that constructs a multi-semantic coding node collaborative scenario system model for different nodes. This model achieves dynamic control of the transmitted data scale through adaptive adjustment of semantic compression rate and wireless communication conditions. The optimization function construction module 20 takes minimizing the maximum task completion time of the system as the core optimization objective and coordinates the scheduling of computing and communication resources to achieve optimal overall system performance. This method achieves adaptive adjustment of the transmitted data scale with semantic compression rate as the transmitted semantic feature information directly related to the task, realizing the joint adaptive allocation of semantic compression rate and wireless communication conditions. Compared with transmitting original image or point cloud data, the data scale can be reduced by more than an order of magnitude, thereby effectively alleviating the link congestion problem in multi-node concurrent access scenarios, improving system scalability and stability. Moreover, the objective function constrains the performance of the slowest node, avoiding the "barrel effect" of individual nodes slowing down the overall task progress. Compared with traditional methods that target average latency, it can significantly shorten the overall completion time of collaborative tasks.

[0080] In this embodiment, the joint system model includes an encoding time model and an encoding energy consumption model that consider the semantic encoding process of multiple semantic coding nodes, used to quantify the time and energy consumption of the encoding process. The energy consumption is strongly correlated with the semantic compression rate and image complexity. It also includes a wireless transmission process of the encoded image, used to construct a transmission time model based on Shannon's theorem and a transmission energy consumption model combined with the energy consumption characteristics of the radio frequency front end, quantifying the time and energy consumption of the transmission process. The energy consumption is strongly correlated with the semantic compression rate, bandwidth, transmit power, and channel conditions.

[0081] In this embodiment, wireless communication conditions may include communication bandwidth, transmission power, and other parameters that characterize the communication or computation of semantic coding nodes.

[0082] In this embodiment, based on the above-mentioned computation-communication joint model and full-dimensional constraints, an optimization problem is constructed with minimizing the maximum task completion time of the system as the core. The optimization variables, objective function and constraint set are defined. The goal is to minimize the maximum task completion time of the system, which is the task time of the semantic encoding node that completes the entire task the latest. The constraints include total processing time constraints, semantic compression rate range constraints and transmission task constraints. The transmission task constraints can be constraints that evaluate the semantic quality of transmission, such as semantic reconstruction quality, task success rate, semantic similarity or downstream control error.

[0083] In this embodiment, the objective function can be solved using hierarchical solution algorithms, heuristic search algorithms, genetic algorithms, particle swarm optimization algorithms, Lagrange dual decomposition or alternating iterative optimization methods, as well as approximate decision models based on deep reinforcement learning or neural network regression.

[0084] The embodiments of the present invention are device embodiments corresponding to the above method embodiments. The specific operations of each module processing step can be understood with reference to the description of the method embodiments, and will not be repeated here.

[0085] like Figure 3 As shown, the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the semantic-driven method for dynamic allocation of communication and computing resources for unmanned devices as described in the above embodiments.

[0086] The present invention also provides a computer-readable storage medium storing a computer program thereon. When the computer program is executed by a processor, it implements the semantic-driven dynamic allocation method for communication and computing resources of unmanned equipment as described in the above embodiments. When the computer program is executed by the processor, it implements the following method steps: Step S1: Construct a joint system model for collaborative transmission tasks of multiple semantic coding nodes with heterogeneous communication and computing resources. Determine the wireless communication conditions and semantic compression rate of each semantic coding node based on the system model to achieve dynamic control of the scale of transmitted data. Step S2: The wireless communication conditions and semantic compression rate of the multi-semantic coding nodes are variables that control the transmitted data. The optimization objective function and the constraints of the optimization objective function are constructed by minimizing the overall maximum task completion time of the multi-semantic coding node collaborative task. Step S3: Iteratively solve the objective function of the joint optimization problem using a preset solution strategy to obtain the optimal computing and communication resource allocation scheme that satisfies the semantic availability condition.

[0087] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0088] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for apparatus or system embodiments, since they are basically similar to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. The apparatus and system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0089] Furthermore, the functional modules in the various embodiments of this disclosure can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and the contents not described in detail in the specification of the present invention are well known to those skilled in the art.

Claims

1. A semantically driven method for dynamic allocation of communication and computing resources for unmanned devices, characterized in that, Including the following steps: A joint system model for collaborative transmission tasks among multiple semantic coding nodes with heterogeneous communication and computing resources is constructed. Based on the system model, the wireless communication conditions and semantic compression rate of each semantic coding node are determined to achieve dynamic control over the scale of transmitted data. The wireless communication conditions and semantic compression rate of the multi-semantic coding nodes are used as variables to control the transmitted data. The optimization objective function and its constraints are constructed by minimizing the overall maximum task completion time of the collaborative task of the multi-semantic coding nodes in the system. By iteratively solving the objective function of the joint optimization problem using a pre-defined solution strategy, the optimal allocation scheme of computing and communication resources that satisfies the semantic availability condition is obtained.

2. The semantically driven dynamic allocation method for communication and computing resources of unmanned devices as described in claim 1, characterized in that, The joint system model includes an encoding time model and an encoding energy consumption model that consider the semantic encoding process of multiple semantic coding nodes, which are used to quantify the time and energy consumption of the encoding process, respectively. The energy consumption is strongly correlated with the semantic compression rate and image complexity. It also includes a transmission time model that considers the wireless transmission process of the encoded image. A transmission energy consumption model is constructed by combining the energy consumption characteristics of the radio frequency front end to quantify the time and energy consumption of the transmission process. The energy consumption is strongly correlated with the semantic compression rate, bandwidth, transmission power and channel conditions.

3. The semantically driven dynamic allocation method for communication and computing resources of unmanned devices as described in claim 1, characterized in that, The wireless communication conditions may include communication bandwidth, transmission power, and other parameters that characterize the communication or computation of semantic coding nodes.

4. The semantically driven dynamic allocation method for communication and computing resources of unmanned devices as described in claim 1, characterized in that, The constraints include transmission task constraints that are constraint metrics for evaluating the semantic quality of the transmission.

5. The semantically driven dynamic allocation method for communication and computing resources of unmanned devices as described in claim 1, characterized in that, The objective function is solved using a hierarchical solution algorithm, heuristic search algorithm, genetic algorithm, particle swarm optimization algorithm, Lagrange dual decomposition or alternating iterative optimization method; or an approximate decision model solution method based on deep reinforcement learning or neural network regression.

6. A semantically driven dynamic allocation device for communication and computing resources of unmanned equipment, characterized in that, include: The system model building module is used to construct a joint system model for the collaborative transmission tasks of multiple semantic coding nodes with heterogeneous communication and computing resources. Based on the system model, the wireless communication conditions and semantic compression rate of each semantic coding node are determined to achieve dynamic control of the scale of transmitted data. The optimization function construction module is used to construct the optimization objective function and its constraints by using the wireless communication conditions and semantic compression rate of the multi-semantic coding nodes as variables to control the transmitted data, and minimizing the overall maximum task completion time of the collaborative task of the multi-semantic coding nodes in the system. The solution module is used to iteratively solve the objective function of the joint optimization problem using a preset solution strategy to obtain the optimal computing and communication resource allocation scheme that satisfies the semantic availability conditions.

7. The semantically driven dynamic allocation device for communication and computing resources of unmanned equipment as described in claim 6, characterized in that, The joint system model includes an encoding time model and an encoding energy consumption model that consider the semantic encoding process of multiple semantic coding nodes, which are used to quantify the time and energy consumption of the encoding process, respectively. The energy consumption is strongly correlated with the semantic compression rate and image complexity. It also includes a transmission time model that considers the wireless transmission process of the encoded image. A transmission energy consumption model is constructed by combining the energy consumption characteristics of the radio frequency front end to quantify the time and energy consumption of the transmission process. The energy consumption is strongly correlated with the semantic compression rate, bandwidth, transmission power and channel conditions.

8. The semantically driven dynamic allocation device for communication and computing resources of unmanned equipment as described in claim 6, characterized in that, The constraints include transmission task constraints that are constraint metrics for evaluating the semantic quality of the transmission.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the semantic-driven dynamic allocation method for communication and computing resources of unmanned equipment as described in any one of claims 1 to 5.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the semantic-driven dynamic allocation method for communication and computing resources of unmanned equipment as described in any one of claims 1 to 5.

Images