A heterogeneous task resource allocation method based on multi-vehicle cooperation

Abstract

The application provides a heterogeneous task resource allocation method based on multi-vehicle cooperation, comprising the following steps: S1, constructing a sensing-computing integrated vehicle networking system model; S2, constructing a resource allocation optimization problem with the optimization objective of minimizing a system utility function; S3, decoupling the resource allocation optimization problem into an entertainment task offloading ratio optimization sub-problem, a computing resource allocation optimization sub-problem and a transmission power optimization sub-problem; S4, obtaining optimal values of the transmission power, the offloading ratio and the computing resource respectively based on an alternating iterative optimization algorithm, so as to perform global optimal allocation of system resources for the sensing-computing integrated vehicle networking system model. The beneficial effect is that the application can construct a heterogeneous task resource allocation method which can finely depict heterogeneous task characteristics, jointly optimize multi-dimensional resources, effectively balance the conflict between delay and energy consumption, and has real-time solving capability.

Description

Technical Field

[0001] This invention relates to the technical field of vehicle-to-everything (V2X) communication, and more specifically, to a method for allocating heterogeneous task resources based on multi-vehicle collaboration. Background Technology

[0002] With the large-scale deployment of fifth-generation (5G) mobile communication and the in-depth research on sixth-generation (6G) technology, integrated sensing, communication, and computing (ISCC) has become a core enabling technology for building future intelligent vehicle networks. In a typical urban intersection scenario, connected vehicles perceive their surroundings by emitting radar signals and offload the raw perception data to the mobile edge computing (MEC) server of the roadside unit for multi-view information fusion to eliminate blind spots for single-vehicle perception and improve the safety of autonomous driving. At the same time, vehicle users (VUEs) also generate in-vehicle entertainment tasks such as high-definition video on demand and real-time gaming. These tasks are characterized by computational intensity, large data volume, but relatively high latency tolerance.

[0003] In the aforementioned Integrated Sensor-Computing Vehicle-to-Everything (ISCC-IoV) system, a typical heterogeneous task resource contention problem exists. On the one hand, the perception task, as a safety-related service, has extremely stringent requirements for processing latency (typically requiring data offloading and fusion to be completed within milliseconds), and its perception accuracy is directly limited by the vehicle's transmission power. On the other hand, the entertainment task, as a service to enhance user experience, requires a significant amount of MEC computing power for data processing. These two tasks create intense competition for the system's communication bandwidth, radar transmission power, and edge computing resources. For example, if the system allocates excessive transmission power and MEC computing power to the perception task in pursuit of higher perception accuracy, it will directly lead to a decrease in the transmission rate or a surge in queuing latency for the entertainment task, and vice versa.

[0004] However, existing technologies still have the following shortcomings: First, the existing task model is too simplistic and fails to fully characterize the essential differences between perception tasks and entertainment tasks in terms of latency sensitivity and processing mode (collaborative fusion vs. independent computing), resulting in a lack of targeted resource scheduling strategies. Second, existing resource scheduling schemes are mostly isolated optimizations, that is, optimizing communication, sensing or computing resources separately, lacking joint planning of transmission power, task offloading ratio and computing resources, making it difficult to cope with the strong coupling relationship between the three, and there is a bottleneck in resource utilization efficiency. Third, existing optimization objectives are rather one-sided, usually focusing on minimizing system latency or minimizing system energy consumption as a single objective. In solutions that target latency, system energy consumption is often too high; while solutions that target energy consumption often sacrifice latency performance, failing to meet the millisecond-level latency constraints for perception tasks in autonomous driving scenarios, resulting in obvious performance trade-off defects. Fourth, existing solution algorithms are highly complex. When faced with multivariable, strongly coupled mixed-integer nonlinear programming problems, traditional algorithms struggle to find the optimal solution in real time in the dynamically changing vehicle network environment, thus limiting the engineering practicality of the solution.

[0005] Therefore, how to construct a heterogeneous task resource allocation method that can accurately characterize the features of heterogeneous tasks, jointly optimize multi-dimensional resources, effectively balance the conflict between latency and energy consumption, and has real-time solution capabilities has become a key technical problem that needs to be solved in the current integrated vehicle networking system. Summary of the Invention

[0006] The technical problem to be solved by this invention is how to construct a heterogeneous task resource allocation method that can accurately characterize the features of heterogeneous tasks, jointly optimize multi-dimensional resources, effectively balance the conflict between latency and energy consumption, and have real-time solution capabilities. In order to overcome the defects of the above-mentioned prior art (or related art), this invention provides a heterogeneous task resource allocation method based on multi-vehicle cooperation.

[0007] This invention provides a method for allocating heterogeneous task resources based on multi-vehicle collaboration, comprising the following steps: Step S1: Construct a sensor-computing integrated vehicle network system model. The sensor-computing integrated vehicle network system model includes a base station equipped with a mobile edge computing server and multiple vehicle users. Each vehicle user generates a perception task at the beginning of each time slot and generates at most one entertainment task according to a Poisson distribution. Step S2: Construct a resource allocation optimization problem with minimizing the system utility function as the optimization objective. The system utility function is the weighted sum of the total system latency and total system energy consumption of the integrated sensory computing vehicle network system model. The decision variables of the resource allocation optimization problem include the transmission power of each vehicle user, the offloading ratio of the entertainment task corresponding to each vehicle user, and the computing resources allocated by the mobile edge computing server to the perception task and the entertainment task of each vehicle user. Step S3: Based on the transmission power, the offload ratio, and the computing resources, the resource allocation optimization problem is decoupled into an entertainment task offload ratio optimization sub-problem, a computing resource allocation optimization sub-problem, and a transmission power optimization sub-problem; Step S4: Based on the alternating iterative optimization algorithm, solve the sub-problem of entertainment task offloading ratio optimization, the sub-problem of computing resource allocation optimization, and the sub-problem of transmission power optimization to obtain the optimal values ​​of transmission power, offloading ratio, and computing resources, so as to perform global optimal allocation of system resources for the integrated sensory computing vehicle network system model.

[0008] Compared with existing technologies, the heterogeneous task resource allocation method based on multi-vehicle collaboration proposed in this invention has the following advantages: This invention constructs a sensor-computer integrated vehicle network system model that incorporates heterogeneous tasks of perception and entertainment. It designs a resource allocation optimization problem with the weighted sum of total system latency and total system energy consumption as the optimization objective, achieving global coordination of communication, perception, and computing resources. By jointly using transmission power, task offloading ratio, and edge computing resources as decision variables, it can dynamically balance the low latency requirements of perception tasks with the high computational demands of entertainment tasks, fundamentally solving the resource competition problem in traditional solutions. Finally, through an alternating iterative optimization algorithm, it ensures the convergence and optimality of the resource allocation strategy in real-world complex vehicle network environments, providing highly reliable and low-latency service guarantees for intelligent driving.

[0009] In one possible implementation, the expression for the resource allocation optimization problem constructed in step S2 is: in, This represents the unloading ratio vector for the entertainment task. Represents the computational resource allocation vector. This represents the transmit power vector of the vehicle user. Denotes the system utility function. This represents the perceived mutual message between the received signal and the propagation channel. This represents the minimum mutual information that the echo signal needs to achieve. This represents the total processing time for the perception task. This represents the maximum tolerable delay for the perception task. The vehicle user's identification number, Indicates the first Transmit power for each vehicle user Indicates the maximum transmission power of the vehicle user. Indicates the first The percentage of entertainment tasks uninstalled by each vehicle user. This indicates the CPU computing resources allocated by the mobile edge computing server for entertainment tasks. This indicates the maximum CPU computing resources of the mobile edge computing server. This indicates the CPU computing resources allocated by the mobile edge computing server for information fusion.

[0010] Compared with existing technologies, the above-mentioned technical solution can accurately model the resource allocation optimization problem mathematically and clarify the core constraints of resource allocation. Among them, the mutual information constraint of perception ensures the accuracy of radar perception from the physical layer and avoids environmental detection distortion caused by resource adjustment; the perception task delay constraint ensures the timeliness of collaborative perception data and meets the demanding millisecond-level requirements of autonomous driving; at the same time, the boundary restrictions on transmission power, offloading ratio and computing resources limit the resource allocation optimization problem to the engineering-feasible range, so that the solution is not only theoretically optimal, but also has practical physical meaning and operability.

[0011] In one possible implementation, in step S3, the decoupled entertainment task offloading ratio optimization sub-problem is used to optimize the offloading ratio of the entertainment tasks for each vehicle user under fixed transmission power and computing resource allocation; the decoupled computing resource allocation optimization sub-problem is used to optimize the computing resources allocated by the mobile edge computing server to the perception tasks and entertainment tasks for each vehicle user under fixed transmission power and entertainment task offloading ratio; and the decoupled transmission power optimization sub-problem is used to optimize the transmission power for each vehicle user under fixed entertainment task offloading ratio and computing resource allocation.

[0012] Compared with existing technologies, the above-mentioned technical solution can decouple the highly coupled original resource allocation optimization problem into three independent sub-problems, which greatly reduces the solution complexity of the algorithm and improves the real-time performance of resource allocation. Among them, the offloading ratio optimization sub-problem focuses on the task partitioning strategy, and explores the potential of task-level scheduling under the condition of fixed communication and computing power; the computing resource allocation sub-problem focuses on the computing power scheduling inside the mobile edge computing server, and achieves load balancing between fusion tasks and computing tasks; the transmit power optimization sub-problem focuses on physical layer transmission performance, and improves spectrum and energy efficiency. This decoupling strategy makes the originally difficult-to-solve mixed integer nonlinear programming problem clear in structure and easy to handle, paving the way for the subsequent adoption of specific efficient algorithms.

[0013] In one possible implementation, in step S4, the optimal value of the unloading ratio is obtained by solving the subproblem of the entertainment task unloading ratio using an unreal wild oat optimization algorithm based on an exploration and development balance mechanism.

[0014] Compared with existing technologies, the above-mentioned technical solution can fully leverage the global search advantage of metaheuristic algorithms in nonlinear and multimodal problems by introducing the unrealized oat optimization algorithm when solving the highly non-convex unloading ratio optimization subproblem. The unrealized oat optimization algorithm can effectively avoid getting trapped in local optima and find a near-global optimal combination of unloading ratios in the complex solution space. Compared with traditional convex optimization approximation or greedy algorithms, it significantly improves resource utilization and the upper limit of system performance.

[0015] In one possible implementation, the process of solving the entertainment task unloading ratio optimization subproblem using the false wild oat optimization algorithm in step S4 includes the following steps: Step A1: Initialize the population and set the algorithm parameters, where each individual in the population represents the unloading ratio of the entertainment task. Step A2: Define the fitness function, using the system utility function as the fitness evaluation criterion for each individual; Step A3: Based on the relationship between the random number and the preset threshold, select an exploration mode, a scrolling mode, or a catapult mode for each individual to update its position; Step A4: Based on the constraints of the resource allocation optimization problem, perform boundary truncation on the updated positions of each individual. Step A5: Repeat steps A1 to A4 until the preset maximum number of iterations is reached, and then output the optimal value of the unloading ratio.

[0016] Compared with existing technologies, the above technical solution can realize intelligent parallel search by transforming the optimization process of unloading ratio into the evolution process of AOO population. The initialization of the population ensures the diversity of initial solutions. Using the system utility function as the fitness evaluation standard ensures the consistency between the optimization direction and the overall system goal. Combined with boundary truncation processing, the solution generated by each iteration strictly satisfies the physical constraints. This process ensures the robustness of the algorithm, enabling it to adapt to scenarios with dynamically changing vehicle numbers.

[0017] In one possible implementation, in step A3, for each individual, when the individual enters the exploration mode, the simulated seed spreads over a long distance, and the position update is based on the global optimal solution and the random diffusion vector; when the individual enters the rolling mode, the simulated seed spreads by rolling, and the position update introduces the Lévy flight mechanism; when the individual enters the catapult mode, the simulated seed triggers catapult propagation, and the position update introduces elastic potential energy and projectile motion.

[0018] Compared with existing technologies, the above-mentioned technical solution can achieve a delicate balance between global exploration and local exploitation by simulating three biological behaviors of wild oat seeds. Specifically, the exploration mode enhances population diversity through random diffusion and prevents premature convergence of the algorithm; the rolling mode, combined with the Lévy flight mechanism, performs a fine search near the optimal solution, accelerating the convergence speed; and the ejection mode, by simulating energy accumulation and release, gives the algorithm the ability to escape local extremum traps. This multimodal update mechanism ensures that the algorithm can stably and efficiently approach the optimal solution when facing the task unloading ratio optimization subproblem with complex constraints.

[0019] In one possible implementation, step S4 involves solving the computational resource allocation optimization sub-problem using the projection gradient method, including the following steps: Step B1: Transform the computational resource allocation optimization subproblem into a convex optimization problem; Step B2: Introduce indicator variables to determine the dominant term of the total system delay in the current iteration, and calculate the gradient of the convex optimization problem with respect to computing resources based on the dominant term; Step B3: Update the computing resource allocation vector according to the gradient of computing resources using the projection gradient iteration formula; Step B4: Perform constraint reduction and truncation processing on the updated computing resource allocation vector to meet the total computing resource limit of the mobile edge computing server; Step B5: Reset steps B1 to B4 until the preset maximum number of iterations is reached, and then output the optimal value of the computing resources.

[0020] Compared with existing technologies, the above-mentioned technical solution can solve the subproblem of computing resource allocation optimization by utilizing its convex optimization characteristics and employing the projection gradient method to obtain the theoretically global optimal solution. By introducing an indicator variable to identify the delay-dominant term, it cleverly handles the non-differentiable points in the total delay function, making gradient calculation accurate and effective. The projection gradient iteration form is simple, has low computational overhead, and can converge quickly, making it very suitable for online operation on resource-constrained MEC servers. The constraint reduction and truncation processing ensures that each resource adjustment strictly meets the upper limit of the total computing power of the MEC server, guaranteeing the stable operation of the system.

[0021] In one possible implementation, in step S2, the total system latency of the integrated sensor-computer vehicle network system model is obtained by the processing latency of the perception task and the processing latency of the entertainment task for each vehicle user. The processing latency of the perception task includes the first task unloading latency and the information fusion latency of the mobile edge computing server. The processing latency of the entertainment task includes the maximum value among the local computing latency, the second task unloading latency, and the task computing latency of the mobile edge computing server.

[0022] Compared with existing technologies, the above-mentioned technical solution can accurately identify system latency bottlenecks by performing refined modeling of the latency composition of perception and entertainment tasks. The latency of perception tasks takes into account both transmission and fusion stages, ensuring the real-time convergence of multi-vehicle perception data. The latency of entertainment tasks takes the maximum value of local computing, transmission, and edge computing, accurately reflecting the task completion time under parallel processing mode. This detailed modeling provides an accurate performance evaluation basis for optimization algorithms, enabling computing resources to be accurately allocated to the most scarce links, thereby maximizing the improvement effect of latency performance.

[0023] In one possible implementation, the total energy consumption of the integrated vehicle networking system model consists of the local computing energy consumption of each vehicle user, the task offloading and transmission energy consumption, and the information fusion and task computing energy consumption of the mobile edge computing server.

[0024] Compared with existing technologies, the above-mentioned technical solution can comprehensively cover the main energy-consuming links of the system, including vehicle local computing, wireless transmission and computing and integration of mobile edge computing servers. This allows the energy consumption item in the optimization target to truly reflect the energy expenditure of the system, and effectively control energy consumption while pursuing low latency. Attached Figure Description

[0025] Figure 1 This is a flowchart of the steps of the present invention; Figure 2 This is a schematic diagram of the integrated sensory computing vehicle networking system model of the present invention; Figure 3 This is a schematic diagram illustrating the solution process of the entertainment task unloading ratio optimization subproblem of the present invention; Figure 4 This is a schematic diagram illustrating the solution process of the computational resource allocation optimization subproblem of the present invention; Figure 5 This is a schematic diagram illustrating the impact of different communication sub-channel bandwidths on system utility according to the present invention; Figure 6 This is a schematic diagram illustrating the impact of different communication sub-channel bandwidths on the total system delay according to the present invention; Figure 7 This is a schematic diagram illustrating the impact of different communication sub-channel bandwidths on the total system energy consumption according to the present invention. Detailed Implementation

[0026] First, those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention. Those skilled in the art can make adjustments as needed to adapt to specific application scenarios.

[0027] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0028] See Figure 1 This invention discloses a method for allocating heterogeneous task resources based on multi-vehicle collaboration, comprising the following steps: Step S1: Construct a sensor-computing integrated vehicle network system model. The sensor-computing integrated vehicle network system model includes a base station equipped with a mobile edge computing server and multiple vehicle users. Each vehicle user generates a perception task at the beginning of each time slot and generates at most one entertainment task according to a Poisson distribution. Step S2: Construct a resource allocation optimization problem with minimizing the system utility function as the optimization objective. The system utility function is the weighted sum of the total system latency and total system energy consumption of the integrated sensory computing vehicle network system model. The decision variables of the resource allocation optimization problem include the transmission power of each vehicle user, the offloading ratio of the entertainment tasks corresponding to each vehicle user, and the computing resources allocated by the mobile edge computing server for the perception tasks and entertainment tasks of each vehicle user. Step S3: Based on the transmit power, offload ratio, and computing resources, the resource allocation optimization problem is decoupled into an entertainment task offload ratio optimization subproblem, a computing resource allocation optimization subproblem, and a transmit power optimization subproblem; Step S4: Based on the alternating iterative optimization algorithm, solve the entertainment task offloading ratio optimization subproblem, computing resource allocation optimization subproblem and transmission power optimization subproblem respectively to obtain the optimal values ​​of transmission power, offloading ratio and computing resources, so as to perform global optimal allocation of system resources for the integrated sensory computing vehicle network system model.

[0029] In this embodiment of the invention, the sensor-computer interface vehicle network system model constructed in step S1 considers a vehicle traffic scenario at a crossroads in a city center. Each road at the crossroads consists of two opposing lanes, with several vehicles traveling in each lane. A base station equipped with a Mobile Edge Computing (MEC) server is deployed at the center of the intersection. Figure 2 As shown, assuming that the integrated sensory computing vehicle network system model has I A number of vehicle user equipment (VUE) users are evenly and randomly distributed on roads within the base station's coverage area, labeled as follows: Assuming time is divided into equally spaced and sufficiently small time slots, at the beginning of each time slot, all vehicle users will perceive their surroundings and generate a perception task. Simultaneously, at most one entertainment task will be generated according to a Poisson distribution. and They represent the first The perceptual and entertainment tasks generated by individual vehicle users, among which... and These represent the data sizes for the sensory and entertainment tasks, respectively, in bits. and These represent the complexity of the perception task and the entertainment task, respectively, which are the number of CPU cycles required to process 1 bit of data, measured in cycles / bit. This represents the maximum tolerable delay for the perception task. It's important to note that, as in the... If a vehicle user does not generate an entertainment task at the start of the time slot, then... and .

[0030] In this embodiment of the invention, it is assumed that the integrated vehicle-to-everything (V2I) system model is based on the 5G New Radio standard, and the physical layer of the V2I link adopts Orthogonal Frequency Division Multiple Access (OFDMA) technology. Therefore, when the... When a vehicle user offloads a task to the base station's MEC server via a V2I link, the signal-to-noise ratio (SNR) received by the base station is: (1) in, Indicates the first Transmit power for each vehicle user Indicates the first Channel gain between individual vehicle users and base stations This represents the power spectral density of Gaussian white noise. This represents the communication bandwidth allocated by the system to vehicle users, in the channel gain. This represents the large-scale fading factor, which is the product of path loss and shadow fading. This represents the small-scale fading gain. obey The distribution is such that each link is independent and identically distributed. According to Shannon's formula, the th... The data transmission rate for each vehicle user is: (2) in, Indicates the first Data transmission rate per vehicle user.

[0031] In this embodiment of the invention, during the sensing process, it is assumed that the coherent processing time of the radar is... That is, its scanning waveform is composed of The duration is The frequency-modulated continuous wave signal constitutes the signal, therefore, the first The transmitted signal of an individual vehicle user can be described as follows: (3) in, This represents a rectangular pulse function centered at zero and with a width of 1, used to define the time-domain width of each frequency-modulated continuous wave signal. Represents the waveform slope, here Indicates the vehicle's sweep bandwidth. Indicates the first The center frequency of the sweep bandwidth corresponding to each vehicle user; When the transmitted signal encounters the target, the direct signal will return to the receiver along the reflection path. Therefore, the first... The echo signal received by a vehicle user can be described as follows: (4) in, Indicates the first Transmit power for each vehicle user This indicates the system impact response, including the round-trip path. This represents the round-trip time delay of the target's direct path. This represents the zero-mean Gaussian white noise introduced at the receiving end.

[0032] In this embodiment of the invention, in a practical system, the purpose of sensing is to extract environmental information from the received signal. Sensing Mutual Information (SMI) is the mutual information between the received signal and the propagation channel, which can be expressed as: (5) in, The SINR (Signal-to-Interference-plus-Noise Ratio) of the received echo is related to the sensing duration and can be expressed as: (6) in, This indicates that other vehicle users have views on the first... The attenuation coefficient for interference from individual vehicle users depends on the direction of radar scanning by other vehicle users. For the sake of simplicity, it is assumed to be a fixed constant here. and It is a Fourier transform pair. The perceived channel gain can be expressed as: (7) in, This indicates the gain of the vehicle's transceiver antenna. Indicates the carrier wavelength of the signal. The radar cross section (RCS) represents the target being detected. Indicates the first The distance between each vehicle user and the target; To ensure the sensing performance of the radar receiver, The minimum mutual information required for echo reception should be subject to the following constraints: (8) The above constraints, which are perceptual mutual information constraints, constitute part of the resource allocation optimization problem.

[0033] In this embodiment of the invention, the complete perception task and part of the entertainment task need to be offloaded to the MEC server of the base station for fusion and computation processing respectively. The offloading time of the perception task for vehicle users depends on the data transmission rate of the V2I link, which can be expressed as: (9) The base station's MEC server performs information fusion on all offloaded sensing tasks; therefore, the information fusion time can be expressed as: (10) in, This refers to the CPU computing resources allocated by the MEC server for information fusion, measured in cycles per second. Therefore, ignoring the time required for the sensing information to be fused from the base station broadcast, the total processing time for the sensing task can be expressed as: (11) Considering the stringent latency requirements of the perception task, the following constraints apply: (12) The above constraints, which are time delay constraints for the perception task, constitute part of the resource allocation optimization problem.

[0034] In this embodiment of the invention, since the entertainment tasks of vehicle users are independent and not sensitive to latency, a partial offloading model is adopted. That is, each entertainment task can be divided into two parts: one part is calculated locally, and the other part is offloaded to the MEC server of the base station for processing. Since the result data of the partial entertainment task processed by the MEC server is very small, the latency fed back to the vehicle user is negligible. Therefore, for the first... The local computation time for each vehicle user's entertainment task is: (13) in, Indicates the first The local CPU computing resources of each vehicle user, measured in cycles / second, depend on the number of users. The computing hardware performance of the individual vehicle user; The computational processing time completed on the MEC server side is divided into offload time. and calculation time It is expressed by the following calculation formula: (14) (15) in, This refers to the CPU computing resources allocated by the MEC server for entertainment tasks, measured in cycles per second. Therefore, the calculation time for completing the entertainment task is expressed as: (16) In summary, combining the unloading times of the perception task and the entertainment task, the final total system latency is as follows: (17) Based on the above calculation process, in step S2, the total system latency of the integrated sensory computing vehicle network system model is obtained from the processing latency of the perception task and the processing latency of the entertainment task of each vehicle user. The processing latency of the perception task includes the first task unloading latency and the information fusion latency of the mobile edge computing server. The processing latency of the entertainment task includes the maximum value among the local computing latency, the second task unloading latency and the task computing latency of the mobile edge computing server.

[0035] In this embodiment of the invention, the total energy consumed in processing the perception and entertainment tasks is divided into three parts: the energy consumption for the vehicle's local computation of the entertainment tasks, the transmission energy consumption for the vehicle user to offload the perception and entertainment tasks to the base station MEC server, and the energy consumption for the MEC server to perform information fusion and computation. Based on the dynamic voltage and frequency scaling method, the computational power of the vehicle user and the base station are respectively... and The power of base station perception fusion is ,in It is an effective switched capacitor, which depends on the chip structure; therefore, the local computing power consumption can be expressed as: (18) The transmission energy consumption for unloading user tasks in vehicles can be expressed as: (19) The energy consumption of the base station for processing vehicle sensing and entertainment tasks is expressed as follows: (20) Finally, the total system energy consumption is obtained using the following formula: (twenty one) The total energy consumption of the integrated vehicle networking system model consists of the local computing energy consumption of each vehicle user, the energy consumption of task offloading and transmission, and the energy consumption of information fusion and task computing of the mobile edge computing server.

[0036] In this embodiment of the invention, the system utility function in step S2 is defined as the weighted sum of the total system delay and the total system energy consumption, and the calculation formula is as follows: (twenty two) in, and are positive real numbers, representing the weights of the total system delay and the total system energy consumption, respectively; definition The proportion vector for unloading entertainment tasks for vehicle users. Allocate vectors for computing resources of the MEC server. Given the transmit power vectors of vehicle users, the resulting resource allocation optimization problem is as follows: (twenty three) in, This represents the unloading ratio vector for entertainment tasks. Represents the computational resource allocation vector. This represents the transmit power vector of the vehicle user. Represents the system utility function. This represents the perceived mutual message between the received signal and the propagation channel. This represents the minimum mutual information that the echo signal needs to achieve. This represents the total processing time for the perception task. This represents the maximum tolerable delay for the perception task. The vehicle user's identification number, Indicates the first Transmit power for each vehicle user Indicates the maximum transmission power of the vehicle user. Indicates the first The percentage of entertainment tasks uninstalled by individual vehicle users. This indicates the CPU computing resources allocated by the mobile edge computing server for entertainment tasks. This indicates the maximum CPU computing resources of the mobile edge computing server. This indicates the CPU computing resources allocated by the mobile edge computing server for information fusion. Limit the transmission power of vehicle users. Limit the percentage of entertainment tasks uninstalled. Constraints on the computing resources of the corresponding MEC server.

[0037] In this embodiment of the invention, in step S3, the decoupled entertainment task offloading ratio optimization sub-problem is used to optimize the offloading ratio of entertainment tasks for each vehicle user under the condition of fixed transmission power and computing resource allocation; the decoupled computing resource allocation optimization sub-problem is used to optimize the computing resources allocated by the mobile edge computing server to the perception tasks and entertainment tasks of each vehicle user under the condition of fixed transmission power and entertainment task offloading ratio; and the decoupled transmission power optimization sub-problem is used to optimize the transmission power of each vehicle user under the condition of fixed entertainment task offloading ratio and computing resource allocation.

[0038] See Figure 3 In this embodiment of the invention, step S4 involves using an unrealized wild oat optimization algorithm based on an exploration and development balance mechanism to solve the sub-problem of optimizing the unloading ratio of entertainment tasks and obtain the optimal value of the unloading ratio. Specifically, this includes the following steps: Step A1: Initialize the population and set the algorithm parameters. Each individual in the population represents the unloading ratio of an entertainment task. Step A2: Define the fitness function, using the system utility function as the fitness evaluation criterion for each individual; Step A3: Based on the relationship between the random number and the preset threshold, select either the exploration mode, the scrolling mode, or the catapult mode for each individual to update its position. Step A4: Based on the constraints of the resource allocation optimization problem, perform boundary truncation on the updated positions of each entity. Step A5: Repeat steps A1 to A4 until the preset maximum number of iterations is reached, and then output the optimal value of the unloading ratio.

[0039] In this embodiment of the invention, the calculation formula for solving the first subproblem, "Optimization of Entertainment Task Unloading Ratio," is as follows: ; Sub-problem of optimizing the uninstallation ratio of entertainment tasks The aim is to optimize the first The proportion of entertainment tasks uninstalled by individual vehicle users Minimizing the utility function is a structurally complex non-convex optimization problem, suitable for solving using the Animated Oat Optimization (AOO) algorithm based on an exploration and exploitation balance mechanism. The AOO algorithm is a recently proposed metaheuristic algorithm that simulates the "exploration-rolling-launching" behavior of animated oat seeds in nature. It achieves global exploration through a random diffusion strategy and local exploitation around the current optimal solution through rolling and launching mechanisms. The AOO algorithm framework is as follows: 1) Initialize the population and set parameters: Defined by indivual A00 population is composed of individuals, each individual representing One possible solution is that the dimension of the individual's position vector is... ,correspond The uninstallation rate of each entertainment task, resulting in the number of iterations. In the In this iteration, the group state matrix is ​​denoted as: ; when At that time, according to the constraints Randomly initialize the group state matrix: ; Next, for the first Each individual generates the following three bio-inspired parameters: ; in, This represents the humidity factor, used to control the diffusion of energy from seeds. Indicates the length of the primary awn, used to control the seed ejection stride. This represents the eccentricity, used to control the degree of rolling disturbance. Furthermore, the definition of the first The dynamic factor for the next iteration is: ; Define the fitness function: Define the first In the nth iteration The fitness function for each seed is: ; in, Indicates the first In the nth iteration The utility function value of each seed; 3) Update the location of individuals in the population: In the next iteration, for individuals Generate two independent random numbers Based on its value, three different position update modes are triggered; when At that time, individual Entering exploration mode, the algorithm simulates the biological behavior of wild oat seeds spreading long distances by wind or water flow to enhance its ability to escape local optima and perform global search. Individual positions are updated as follows: ; in, and These represent the element values. and all 1 3D row vectors Indicates the first The diffusion vector of the next iteration. Indicates the first The position of the best-fit individual in a generation; when At that time, individual Entering rolling mode, the algorithm simulates the rolling propagation and ejection behavior of seeds when there are no obstacles. To improve the individual's fine-grained search ability around the current optimal solution, the AOO algorithm introduces the Lévy flight mechanism and humidity rolling perturbation. The Lévy flight mechanism is a random walk strategy that follows a heavy-tailed distribution. Its step size combines the characteristics of short-distance frequent exploration and occasional long-distance jumps, which can effectively avoid the algorithm getting trapped in local limits. The individual position is updated as follows: ; in, Indicates the first A nonlinearly decreasing integer factor in each iteration is used to simulate the natural dissipation of energy as it rolls. In the Lévy flight mechanism, This represents the desired position and is typically used to adjust the step size. It is a random value between 0 and 1, and this randomness helps control the direction and distance of movement during flight. The scale parameter representing the step size distribution is used to control the range of step size variation. This indicates that the elements follow a normal distribution. of A dimensional vector, representing the current velocity vector, reflecting the individual's motion state. The parameters representing the stable distribution determine the shape of the step size distribution, thus affecting the randomness and diversity of the step size. Finally, The gamma function is a continuous extension of the factorial function with non-integer parameters, used to calculate non-integer factorial values. when At that time, individual Entering catapult mode, the algorithm simulates the catapult propagation behavior triggered by a seed encountering an obstacle. Individuals generate sudden displacement by accumulating elastic potential energy, thereby further enhancing the algorithm's ability to escape local optima. The individual position is updated as follows: ; in, , , Indicates the seed ejection amplitude control factor. This represents the elasticity coefficient of the primary awn of a seed. This indicates the change in the length of the energy stored during the ejection of the main beam. Indicates the angle between the launch trajectory and the ground. This represents the air resistance coefficient in projectile motion; 4) Constraint Reduction and Individual Position Truncation: To reduce the search space, a comprehensive approach is adopted. It can shrink the first The upper limit for the percentage of vehicle users who uninstall the software is: ; Therefore, to ensure the validity of the solution, the vector that has been updated with individual positions needs to be truncated as follows: ; The optimal value of the unloading ratio can be obtained through the above calculation process.

[0040] See Figure 4 In this embodiment of the invention, step S4 involves using the projection gradient method to solve the sub-problem of resource allocation optimization, specifically including the following steps: Step B1: Transform the computational resource allocation optimization subproblem into a convex optimization problem; Step B2: Introduce indicator variables to determine the dominant term of the total system delay in the current iteration, and calculate the gradient of the convex optimization problem with respect to computing resources based on the dominant term; Step B3: Update the computing resource allocation vector according to the gradient of computing resources using the projection gradient iteration formula; Step B4 involves constraining and truncating the updated computing resource allocation vector to meet the total computing resource limit of the mobile edge computing server. Step B5: Reset steps B1 to B4 until the preset maximum number of iterations is reached, and then output the optimal value of computing resources.

[0041] In this embodiment of the invention, the calculation formula for solving the second subproblem, "the subproblem of optimizing resource allocation," is expressed as follows: (48) First, The objective function is expanded as follows: (49) As can be seen, in , On the domain of definition, the function and Both are convex functions. Since the non-negative linear combination and pointwise maximum operation of convex functions still maintain convexity, the time delay function... It has strict convex properties, and similarly, and It is also a convex function, therefore the energy consumption function It is a convex function, and at the same time, it is constrained. Since the inequality is linear, the resulting feasible region is a simple convex polyhedron. Therefore... This is a convex optimization problem, and the global optimum can be found using the projected gradient method. The form of the first derivative of the objective function and the iterative steps are given below. because Contains The function is not differentiable at the boundary point but is differentiable at other points in its domain. In solving the problem, the dominant term of the total system delay can be determined using the subgradient principle, and its derivative can be calculated. For simplicity, an indicator variable is introduced. and Perform the following calculations: (50) (51) Among them, when When it is not the dominant term of total delay, the corresponding ,at this time Only the energy consumption term of the objective function is affected; when When it is the dominant term in the total system delay, the corresponding , The time delay and energy consumption terms that affect the objective function The definitions and effects are similar; Therefore, the objective function is related to and The gradient can be expressed as follows: (52) (53) remember , The iterative form of the projection gradient is: (54) in, Step size, for Defined feasible domain For projection operators, Indicates to Take the gradient, since For convex optimization problems, when the step size is appropriate, the iterative formula (55) can guarantee that the objective function is monotonically non-increasing and converges to the global optimum. In addition, the constraint reduction is performed according to the following steps: according to and ,right and By interval Cut off, if ,but satisfy; Otherwise, for and Reduce by a factor of 1 to make the sum equal to 2. This yields the following expression: (55) (56) The optimal value of computing resources can be obtained through the above calculation process.

[0042] In this embodiment of the invention, the calculation formula for solving the second subproblem, "transmission power optimization subproblem," is expressed as follows: (57) The specific solution steps are similar to those for the first subproblem, and will not be repeated in this invention specification. Set a maximum number of iterations and perform alternating optimizations. , and In each iteration, the optimal value of the corresponding optimization variable is calculated, and the optimization stops after the maximum number of iterations is reached.

[0043] Example 1 In this embodiment, the following was constructed: Figure 2 The urban traffic simulation system shown includes a central city intersection, a base station, and several vehicle users. Table 1 below shows the important simulation parameters and default settings: Table 1 Simulation Parameter Settings To comprehensively evaluate the performance of the method of the present invention, the following common unloading strategies were selected for comparison in this embodiment: Local computing solution: Entertainment tasks are calculated locally by the vehicle user; Uniform distribution scheme: Vehicle users offload half of their entertainment tasks to the MEC server for computation, i.e. Furthermore, the latter allocates the same computing resources to all entertainment tasks; MEC computing solution: Vehicle users offload the entire entertainment task to the MEC server for computing; In addition, to further evaluate the effectiveness and applicability of the adopted AOO algorithm, the following common crowd intelligence optimization algorithms were selected for performance comparison in this embodiment: PSO algorithm: and The particle swarm optimization algorithm is used to solve this problem. The solution remains unchanged; GWO algorithm: and The Grey Wolf optimization algorithm is used to solve this problem. The solution remains unchanged; DBO algorithm: and The dung beetle algorithm is used to solve this problem. The solution remains unchanged; Figure 5-7 The impact of the communication sub-channel bandwidth allocated to vehicle users on system performance is shown. It can be seen that when the communication bandwidth increases, the total latency, total energy consumption, and system utility value of all schemes decrease. This is because high communication bandwidth improves the V2I link data transmission rate, thereby reducing the transmission latency and energy consumption of perception and entertainment tasks during the offloading process. Since the system utility comprehensively considers energy consumption and latency, its trend is consistent with the performance of the two tasks. Compared with other computation offloading schemes, the method of this invention achieves optimal performance by balancing the transmission latency and MEC server and vehicle computing resources according to the allocated bandwidth resources through a dynamic offloading mechanism. In addition, the latency of the three comparative schemes, from highest to lowest, is local computing, uniform allocation, and MEC computing. This is because the MEC server has more powerful computing resources than the local server, effectively reducing the computation latency.

[0044] In the description of this invention, the references to "one embodiment," "some embodiments," "in this embodiment," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0045] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for allocating heterogeneous task resources based on multi-vehicle collaboration, characterized in that, Includes the following steps: Step S1: Construct a sensor-computing integrated vehicle network system model. The sensor-computing integrated vehicle network system model includes a base station equipped with a mobile edge computing server and multiple vehicle users. Each vehicle user generates a perception task at the beginning of each time slot and generates at most one entertainment task according to a Poisson distribution. Step S2: Construct a resource allocation optimization problem with minimizing the system utility function as the optimization objective. The system utility function is the weighted sum of the total system latency and total system energy consumption of the integrated sensory computing vehicle network system model. The decision variables of the resource allocation optimization problem include the transmission power of each vehicle user, the offloading ratio of the entertainment task corresponding to each vehicle user, and the computing resources allocated by the mobile edge computing server to the perception task and the entertainment task of each vehicle user. Step S3: Based on the transmission power, the offload ratio, and the computing resources, the resource allocation optimization problem is decoupled into an entertainment task offload ratio optimization sub-problem, a computing resource allocation optimization sub-problem, and a transmission power optimization sub-problem; Step S4: Based on the alternating iterative optimization algorithm, solve the sub-problem of entertainment task offloading ratio optimization, the sub-problem of computing resource allocation optimization, and the sub-problem of transmission power optimization to obtain the optimal values ​​of transmission power, offloading ratio, and computing resources, so as to perform global optimal allocation of system resources for the integrated sensory computing vehicle network system model.

2. The heterogeneous task resource allocation method according to claim 1, characterized in that, The expression for the resource allocation optimization problem constructed in step S2 is: in, This represents the unloading ratio vector for the entertainment task. Represents the computational resource allocation vector. This represents the transmit power vector of the vehicle user. Denotes the system utility function. This represents the perceived mutual message between the received signal and the propagation channel. This represents the minimum mutual information that the echo signal needs to achieve. This represents the total processing time for the perception task. This represents the maximum tolerable delay for the perception task. The vehicle user's identification number, Indicates the first Transmit power for each vehicle user Indicates the maximum transmission power of the vehicle user. Indicates the first The percentage of entertainment tasks uninstalled by each vehicle user. This indicates the CPU computing resources allocated by the mobile edge computing server for entertainment tasks. This indicates the maximum CPU computing resources of the mobile edge computing server. This indicates the CPU computing resources allocated by the mobile edge computing server for information fusion.

3. The heterogeneous task resource allocation method according to claim 1, characterized in that, In step S3, the decoupled entertainment task offloading ratio optimization sub-problem is used to optimize the offloading ratio of the entertainment task for each vehicle user under the condition of fixed transmission power and computing resource allocation; the decoupled computing resource allocation optimization sub-problem is used to optimize the computing resources allocated by the mobile edge computing server to the perception task and the entertainment task for each vehicle user under the condition of fixed transmission power and the offloading ratio of the entertainment task. The decoupled transmit power optimization subproblem is used to optimize the transmit power of each vehicle user while keeping the offload ratio of the entertainment task and the allocation of computing resources fixed.

4. The heterogeneous task resource allocation method according to claim 1, characterized in that, In step S4, the optimal value of the unloading ratio is obtained by solving the subproblem of the entertainment task unloading ratio optimization using the unreal wild oat optimization algorithm based on the exploration and development balance mechanism.

5. The heterogeneous task resource allocation method according to claim 4, characterized in that, The process of solving the entertainment task unloading ratio optimization sub-problem using the false wild oat optimization algorithm in step S4 includes the following steps: Step A1: Initialize the population and set the algorithm parameters, where each individual in the population represents the unloading ratio of the entertainment task. Step A2: Define the fitness function, using the system utility function as the fitness evaluation criterion for each individual; Step A3: Based on the relationship between the random number and the preset threshold, select an exploration mode, a scrolling mode, or a catapult mode for each individual to update its position; Step A4: Based on the constraints of the resource allocation optimization problem, perform boundary truncation on the updated positions of each individual. Step A5: Repeat steps A1 to A4 until the preset maximum number of iterations is reached, and then output the optimal value of the unloading ratio.

6. The heterogeneous task resource allocation method according to claim 5, characterized in that, In step A3, for each individual, when the individual enters the exploration mode, the simulated seed spreads over a long distance, and the position update is based on the global optimal solution and the random diffusion vector; when the individual enters the rolling mode, the simulated seed spreads by rolling, and the position update introduces the Lévy flight mechanism; when the individual enters the catapult mode, the simulated seed triggers catapult propagation, and the position update introduces elastic potential energy and projectile motion.

7. The heterogeneous task resource allocation method according to claim 1, characterized in that, In step S4, the projection gradient method is used to solve the computational resource allocation optimization sub-problem, which includes the following steps: Step B1: Transform the computational resource allocation optimization subproblem into a convex optimization problem; Step B2: Introduce indicator variables to determine the dominant term of the total system delay in the current iteration, and calculate the gradient of the convex optimization problem with respect to computing resources based on the dominant term; Step B3: Update the computing resource allocation vector according to the gradient of computing resources using the projection gradient iteration formula; Step B4: Perform constraint reduction and truncation processing on the updated computing resource allocation vector to meet the total computing resource limit of the mobile edge computing server; Step B5: Reset steps B1 to B4 until the preset maximum number of iterations is reached, and then output the optimal value of the computing resources.

8. The heterogeneous task resource allocation method according to claim 1, characterized in that, In step S2, the total system latency of the integrated sensor-computer vehicle network system model is obtained by the processing latency of the perception task and the processing latency of the entertainment task of each vehicle user. The processing latency of the perception task includes the first task unloading latency and the information fusion latency of the mobile edge computing server. The processing latency of the entertainment task includes the maximum value of the local computing latency, the second task unloading latency and the task computing latency of the mobile edge computing server.

9. The heterogeneous task resource allocation method according to claim 1, characterized in that, The total energy consumption of the integrated vehicle networking system model consists of the local computing energy consumption of each vehicle user, the task offloading and transmission energy consumption, and the information fusion and task computing energy consumption of the mobile edge computing server.

Images