Task unloading and resource allocation method and device oriented to V2V assistance in Internet of Vehicles, and medium

Abstract

The invention belongs to the technical field of edge computing, Internet of Vehicles, resource allocation and task unloading, and discloses a V2V assistance-oriented task unloading and resource allocation method and device in the Internet of Vehicles, and a medium, and the method comprises the steps: constructing a task queue model; determining an energy consumption model based on the task queue model; based on the energy consumption model, constructing an optimization problem by taking minimization of long-term energy consumption as a target; and solving the optimization problem to determine a task unloading and resource allocation scheme. According to the method, unloading link selection, bandwidth resource allocation, computing resource allocation, unloading proportion and transmission power allocation are jointly optimized. In consideration of different expenditures of decisions, a double-time-scale method is adopted, that is, decisions with large expenditures are optimized on a large time scale; the decision with low overhead is optimized on a small time scale; in this way, by combining V2I and V2V unloading models, the system energy consumption is reduced to the maximum extent under the condition that the system queue is stable.

Description

Technical Field

[0001] The present invention relates to the technical fields of edge computing, Internet of Vehicles, resource allocation and task offloading, and more specifically, to a method, device and medium for task offloading and resource allocation for V2V assistance in an Internet of Vehicles. Background Art

[0002] With the development of 5G technology, the Internet of Vehicles has become a research focus in the field of the Internet of Things. With the help of vehicle-to-infrastructure (V2I) and vehicle-to-vehicle (V2V) communication technologies, vehicles or infrastructure are connected to the network to achieve information sharing and service support. The Internet of Vehicles is an effective way to achieve intelligent transportation. In order to achieve safe and convenient driving, a large number of resource-intensive in-vehicle network applications such as collision detection and real-time navigation are emerging. The above applications need to be completed within strict deadlines, which is a huge challenge for vehicle servers with insufficient computing resources. Offloading tasks to cloud servers with sufficient computing resources is a computing mode. However, the transmission distance between vehicles and cloud servers is long, which inevitably generates high propagation delay and jitter. As a more promising computing paradigm, vehicle edge computing (VEC) provides users with high reliability and low latency services by offloading vehicle tasks to fixed edge servers (FES) or nearby vehicles with spare computing resources. Considering the random mobility of vehicles and the rapid changes in the network environment, it is challenging to explore the optimal resource allocation and task offloading schemes in VEC under dynamic topology.

[0003] Many existing works are devoted to exploring resource allocation and task offloading based on the V2I mode, that is, vehicles can offload tasks to FES through V2I links. However, the V2I mode relies heavily on many FES nodes, such as road side units (RSU), which inevitably leads to high deployment and operation costs. At the same time, with the increase of traffic density within the coverage area of ​​RSU, FES with limited resources can hardly meet the user's delay requirements. Considering that there are some vehicles with surplus computing resources in the vehicle network, some research works have studied resource allocation and task offloading schemes based on the V2V mode. In the V2V mode, task vehicles (TV) with limited computing resources can offload tasks to collaborative vehicles (CV) equipped with mobile computing servers (MCS) without FES nodes. The V2V offloading mode can serve as a supplement to the V2I offloading mode, reduce the computing burden of RSU, and improve the utilization of vehicle computing resources. At present, the exploration of V2I and V2V offloading modes is still limited. Vehicles can only choose fixed unloading links and cannot flexibly choose V2I or V2V unloading modes. In addition, computing and communication resources, which are core resources, are not fully exploited during the unloading process.

[0004] To achieve the optimal resource allocation and task offloading solution, it is usually necessary to solve mixed integer nonlinear programming (MINLP), which is generally considered to be NP-hard. Some studies have designed heuristic intelligent optimization algorithms and game-based algorithms. However, the above traditional optimization algorithms usually require a large number of iterations and high computational complexity, and are not suitable for rapidly changing VECs.

[0005] To this end, some researchers apply data-driven deep reinforcement learning (DRL) algorithms to develop optimal resource allocation and task offloading schemes. DRL uses agents to interact with the environment and obtain rewards, thereby continuously improving action functions to adapt to the rapidly changing VEC environment. In the process of optimizing energy consumption or delay for VEC, the long-term performance of the system is equally important. Existing DRL algorithms cannot directly handle long-term constraints. To this end, some researchers proposed a Lyapunov-based DRL algorithm. The algorithm first applies the Lyapunov framework to transform the multi-slot optimization problem into a deterministic subproblem for each slot to ensure system stability. Then, the corresponding DRL algorithm is proposed to solve each subproblem to obtain an efficient solution. However, the above algorithms usually operate at a single time scale, that is, the timeline is divided into multiple discrete time slots, and resource and / or offloading allocation decisions are made at the beginning of each time slot. This single time scale framework ignores the overhead of decision making. Specifically, the decision overhead that requires global information is higher than the decision overhead that requires local information. In order to achieve the best performance indicators, the update frequency of the above two types of decisions should also be different. However, existing strategies rarely take this into account. Summary of the invention

[0006] The present invention is provided to solve the above-mentioned problems existing in the prior art. Therefore, there is a need for a method, device and medium for task offloading and resource allocation for V2V assistance in a connected vehicle network, the purpose of which is to combine V2I and V2V offloading models to minimize system energy consumption under the condition of stable system queues. The present invention jointly optimizes offloading link selection, bandwidth re-resource allocation, computing resource allocation, offloading ratio and transmission power allocation. Taking into account the different overheads of decisions, a dual time scale approach is adopted, that is, decisions with large overhead are optimized on a large time scale; decisions with small overhead are optimized on a small time scale. In order to obtain online decisions, a Lyapunov-based optimization scheme is proposed. The optimization scheme applies the Lyapunov framework to decouple the formulated problem into two sub-problems on two time scales.

[0007] According to a first aspect of the present invention, a method for task offloading and resource allocation for V2V assistance in a connected vehicle network is provided, the method comprising:

[0008] Build a task queue model;

[0009] Based on the task queue model, determining an energy consumption model;

[0010] Based on the energy consumption model, construct an optimization problem with the goal of minimizing long-term energy consumption;

[0011] Solving the optimization problem to determine a task offloading and resource allocation solution;

[0012] Among them, the task queue model is constructed by the following method:

[0013] With binary variable o m,n (k) = {0, 1} represents the unloading link selection between the task vehicle TV m and the cooperative vehicle CV n;

[0014] In each time frame, TV can only establish a connection with one CV and one CV can only provide services to one TV. Where N represents the CV set and M represents the TV set;

[0015] The amount of data generated by the mth task vehicle TV m in time slot t is A m (t);

[0016] The local processing part of TV is The part that is unloaded to the RSU for processing is The part unloaded to the cooperative vehicle CV for processing is is the task offloading ratio of the mth cooperative vehicle CVm in time slot t, and in, Indicates the local processing ratio of TV m, Indicates the proportion of offloading to RSU for processing. Indicates the proportion of unloading to cooperative vehicles;

[0017] At time slot t, the length of TVm's local processing queue is The update formula is:

[0018]

[0019] in represents the amount of data processed locally in time slot t; max represents the maximum value function; τ represents the duration of a time slot; ω represents the CPU cycle required to process a 1-bit task; Indicates the CPU frequency of TV m;

[0020] The transmission rate from TVm to RSU is:

[0021]

[0022] in, Indicates the transmission rate from TVm to RSU; is the number of subcarriers allocated to the V2I link in time frame k, is the V2I link transmission power, represents the channel gain of the V2I link, σ 2 Represents the noise power.

[0023] At time slot t, the length of the V2I link task sending queue of TVm is The update formula is:

[0024]

[0025] in represents the amount of data transmitted in time slot t;

[0026] The V2V link transmission rate from TVm to CVn is:

[0027]

[0028] in, represents the V2V link transmission rate from TVm to CVn; is the V2V link transmission power, is the channel gain of the V2V link; b represents the bandwidth of a subchannel;

[0029] At time slot t, the length of TVm’s V2V task sending queue is The update formula is:

[0030]

[0031] in represents the amount of data sent to CV in time slot t. The number of tasks offloaded from TVm to RSU is:

[0032]

[0033] In the formula, represents the number of tasks offloaded from TVm to RSU; min represents the minimum function; represents the amount of data sent to RSU in time slot t;

[0034] At time slot t, the length of the RSU task processing queue is The update formula is:

[0035]

[0036] in Indicates the amount of data processed by the RSU; It represents the calculation frequency of RSU allocated to TV m.

[0037] The number of tasks offloaded to CVn is:

[0038]

[0039] Among them, the binary variable o m,n =1 means TVm can offload tasks to CVn; Indicates the number of tasks offloaded to CVn;

[0040] At time slot t, CVn calculates the queue length as The update formula is:

[0041]

[0042] in Indicates the amount of data sent to CV n; Indicates the calculation frequency of CV n.

[0043] According to a second aspect of the present invention, a task offloading and resource allocation device for V2V assistance in a connected vehicle network is provided, the device comprising:

[0044] A first model building unit is configured to build a task queue model;

[0045] A second model building unit is configured to determine an energy consumption model based on the task queue model;

[0046] An optimization problem construction unit is configured to construct an optimization problem based on the energy consumption model with the goal of minimizing long-term energy consumption;

[0047] an optimization problem solving unit, configured to solve the optimization problem to determine a task offloading and resource allocation solution;

[0048] The first model building unit is further configured to build a task queue model by the following method:

[0049] With binary variable o m,n (k) = {0, 1} represents the unloading link selection between the task vehicle TV m and the cooperative vehicle CV n;

[0050] In each time frame, TV can only establish a connection with one CV and one CV can only provide services to one TV. Where N represents the CV set and M represents the TV set;

[0051] The amount of data generated by the mth task vehicle TV m in time slot t is A m (t);

[0052] The local processing part of TV is The part that is unloaded to the RSU for processing is The part unloaded to the cooperative vehicle CV for processing is is the task offloading ratio of the mth cooperative vehicle CVm in time slot t, and in, Indicates the local processing ratio of TV m, Indicates the proportion of offloading to RSU for processing. Indicates the proportion of unloading to cooperative vehicles;

[0053] At time slot t, the length of TVm's local processing queue is The update formula is:

[0054]

[0055] in represents the amount of data processed locally in time slot t; max represents the maximum value function; τ represents the duration of a time slot; ω represents the CPU cycle required to process a 1-bit task; Indicates the CPU frequency of TV m;

[0056] The transmission rate from TVm to RSU is:

[0057]

[0058] in, Indicates the transmission rate from TVm to RSU; is the number of subcarriers allocated to the V2I link in time frame k, is the V2I link transmission power, represents the channel gain of the V2I link, σ 2 Represents the noise power.

[0059] At time slot t, the length of the V2I link task sending queue of TVm is The update formula is:

[0060]

[0061] in represents the amount of data transmitted in time slot t;

[0062] The V2V link transmission rate from TVm to CVn is:

[0063]

[0064] in, represents the V2V link transmission rate from TVm to CVn; is the V2V link transmission power, is the channel gain of the V2V link; b represents the bandwidth of a subchannel;

[0065] At time slot t, the length of TVm’s V2V task sending queue is The update formula is:

[0066]

[0067] in represents the amount of data sent to CV in time slot t. The number of tasks offloaded from TVm to RSU is:

[0068]

[0069] In the formula, represents the number of tasks offloaded from TVm to RSU; min represents the minimum function; Represents the amount of data sent to the RSU in time slot t.

[0070] At time slot t, the length of the RSU task processing queue is The update formula is:

[0071]

[0072] in Indicates the amount of data processed by the RSU; It represents the calculation frequency of RSU allocated to TV m.

[0073] The number of tasks offloaded to CVn is:

[0074]

[0075] Among them, the binary variable o m,n =1 means TVm can offload tasks to CVn; Indicates the number of tasks offloaded to CVn;

[0076] At time slot t, CVn calculates the queue length as The update formula is:

[0077]

[0078] in Indicates the amount of data sent to CV n; Indicates the calculation frequency of CV n.

[0079] According to a third aspect of the present invention, there is provided a readable storage medium storing one or more programs, wherein the one or more programs can be executed by one or more processors to implement the method as described above.

[0080] The present invention has at least the following beneficial effects:

[0081] The present invention considers two offloading modes, V2I and V2V, and proposes an online dual-time scale strategy to jointly optimize resource allocation and task offloading. Specifically, offloading link selection, bandwidth resource allocation and computing resource allocation are performed on a large time scale, and offloading ratio and transmission power allocation are performed on a small time scale. In order to obtain the optimal solution, an optimization problem is proposed to minimize the system energy consumption and ensure the stability of the task queue. In order to solve the proposed problem, the Lyapunov optimization framework is used to decompose the original problem into two time scale sub-problems. On this basis, a DRL algorithm and a generalized benders decomposition algorithm based on continuous convex approximation are proposed to solve the two sub-problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 A schematic diagram of a vehicle network architecture provided by an embodiment of the present invention.

[0083] Figure 2 A schematic diagram of a dual time scale provided by an embodiment of the present invention.

[0084] Figure 3 A flowchart of a method for task offloading and resource allocation for V2V assistance in a connected vehicle network provided by an embodiment of the present invention.

[0085] Figure 4 A structural diagram of a task offloading and resource allocation device for V2V assistance in a connected vehicle network provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0086] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. The embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings and specific embodiments, but are not intended to limit the present invention. For the various steps described herein, if there is no necessity for a causal relationship between each other, the order in which they are described as examples herein should not be regarded as a limitation, and those skilled in the art should know that they can be adjusted in order, as long as the logic between them is not destroyed, resulting in the inability to implement the entire process.

[0087] The embodiment of the present invention provides a method for task offloading and resource allocation for V2V assistance in a connected vehicle network. Figure 1 As shown in FIG. 1 , a schematic diagram of a vehicle network architecture provided by an embodiment of the present invention is shown. The task offloading and resource allocation method for V2V assistance in the vehicle network can be applied to the following examples: Figure 1 It should be noted that the vehicle network architecture provided here is only an example of a specific scenario in which the method proposed in the present invention can be applied, and does not mean that the present invention must be applied to such a specific scenario. Figure 1 The vehicle network architecture shown. Figure 1The vehicle network architecture shown includes a mission vehicle TV, a cooperative vehicle CV and a roadside unit RSC. The mission vehicle can have up to three processing modes, namely local processing, RSU processing and cooperative vehicle processing.

[0088] This method introduces a dual time scale model and designs an online optimization strategy to solve the resource allocation and task offloading problems in V2I and V2V while ensuring the stability of the system. Figure 2 As shown, it is a schematic diagram of the dual time scale provided by an embodiment of the present invention. The dual time scale specifically refers to: T consecutive time slots constitute a time frame; a large time scale decision is made at the beginning of each time frame, and a small time scale decision is made at the beginning of each time slot.

[0089] Combination Figure 1 The vehicle network architecture shown and Figure 2 The dual time scale shown, Figure 3 The flowchart of the method for task offloading and resource allocation for V2V assistance in the connected vehicle network provided by the embodiment of the present invention can be implemented by the following steps S10 to S40 when the method for task offloading and resource allocation for V2V assistance in the connected vehicle network is specifically implemented.

[0090] S10: Build a task queue model.

[0091] In this embodiment, the task queue model is constructed by the following method:

[0092] With binary variable o m,n (k) = {0, 1} represents the unloading link selection between the task vehicle TV m and the cooperative vehicle CV n;

[0093] In each time frame, TV can only establish a connection with one CV and one CV can only provide services to one TV. Where N represents the CV set and M represents the TV set;

[0094] The amount of data generated by the mth task vehicle TV m in time slot t is A m (t);

[0095] The local processing part of TV is The part that is unloaded to the RSU for processing is The part unloaded to the cooperative vehicle CV for processing is is the task offloading ratio of the mth cooperative vehicle CVm in time slot t, and in, Indicates the local processing ratio of TV m, Indicates the proportion of offloading to RSU for processing. Indicates the proportion of unloading to cooperative vehicles;

[0096] At time slot t, the length of TVm's local processing queue is The update formula is:

[0097]

[0098] in represents the amount of data processed locally in time slot t; max represents the maximum value function; τ represents the duration of a time slot; ω represents the CPU cycle required to process a 1-bit task; Indicates the CPU frequency of TV m;

[0099] The transmission rate from TVm to RSU is:

[0100]

[0101] in, Indicates the transmission rate from TVm to RSU; is the number of subcarriers allocated to the V2I link in time frame k, is the V2I link transmission power, represents the channel gain of the V2I link, σ 2 Represents the noise power.

[0102] At time slot t, the length of the V2I link task sending queue of TVm is The update formula is:

[0103]

[0104] in represents the amount of data transmitted in time slot t;

[0105] The V2V link transmission rate from TVm to CVn is:

[0106]

[0107] in, represents the V2V link transmission rate from TVm to CVn; is the V2V link transmission power, is the channel gain of the V2V link; b represents the bandwidth of a subchannel;

[0108] At time slot t, the length of TVm’s V2V task sending queue is The update formula is:

[0109]

[0110] in represents the amount of data sent to CV in time slot t. The number of tasks offloaded from TVm to RSU is:

[0111]

[0112] In the formula, represents the number of tasks offloaded from TVm to RSU; min represents the minimum function; Represents the amount of data sent to the RSU in time slot t.

[0113] At time slot t, the length of the RSU task processing queue is The update formula is:

[0114]

[0115] in Indicates the amount of data processed by the RSU; It represents the calculation frequency of RSU allocated to TV m.

[0116] The number of tasks offloaded to CVn is:

[0117]

[0118] Among them, the binary variable o m,n =1 means TVm can offload tasks to CVn; Indicates the number of tasks offloaded to CVn;

[0119] At time slot t, CVn calculates the queue length as The update formula is:

[0120]

[0121] in Indicates the amount of data sent to CV n; Indicates the calculation frequency of CV n.

[0122] S20: Determine an energy consumption model based on the task queue model.

[0123] In some embodiments, based on the task queue model, the energy consumption model is determined by the following method:

[0124] In time slot t, TVm local processing energy consumption for:

[0125]

[0126] in, is the CPU frequency of TV;

[0127] Energy consumption uploaded from TVm to RSU for:

[0128]

[0129] Energy consumption of TVm uploading to CV for:

[0130]

[0131] Energy consumed by RSU in processing tasks uploaded by TVm for:

[0132]

[0133] Where e represents the energy consumption of RSU processing 1-bit task;

[0134] Energy consumed by CVn processing tasks for:

[0135]

[0136] The total energy consumption in time slot t is:

[0137]

[0138] Where E(t) is the total energy consumption in time slot t.

[0139] S30: Based on the energy consumption model, an optimization problem is constructed with the goal of minimizing long-term energy consumption. In some embodiments, the optimization problem is constructed with the goal of minimizing long-term energy consumption. For the convenience of problem description, let O(k) = {o m,n (k)} m∈M,n∈N , α(t)={α m (t)} m∈M ,

[0140] B(k)={B m (k)} m∈M , P(t)={P m (t)} m∈M ,

[0141] Then the energy consumption minimization problem P0 is:

[0142]

[0143] Among them, P0 represents the problem of minimizing energy consumption, st represents being subject to constraints, C1, C2, C3, C4, C5, C6, C7, C8, C9, C10 and C11 all represent constraints; B max represents the subcarrier overview, f maxIndicates the maximum operating frequency of RSU. Indicates the maximum transmission power of TV m.

[0144] To ensure the stable operation of the system, define the queue backlog Q(t) = {Q TV (t),Q CV (t)}, where

[0145]

[0146] Define the Lyapunov function L(Q(t)):

[0147]

[0148] Define the T-slot Lyapunov drift function Δ T (Q(t)):

[0149] Δ T (Q(t))=E{L(Q(t+T))-L(Q(t))|Q(t)}

[0150] Define drift plus penalty function

[0151]

[0152] Where V represents the trade-off parameter; T k represents the set of time slots of time frame k.

[0153] By minimizing the upper bound of the drift plus penalty function, P0 is transformed into the first problem P1 as follows:

[0154]

[0155] in

[0156]

[0157] The first problem P1 can be further decomposed into the per-frame sub-problem P2 at a large time scale and the per-time slot sub-problem P3 at a small time scale as shown below:

[0158]

[0159] S40: Solve the optimization problem to determine a task offloading and resource allocation solution.

[0160] In some embodiments, for the large time scale optimization problem (i.e., the per-frame subproblem P2 at a large time scale), it is modeled as a Markov decision process, and a reinforcement learning algorithm with a deep deterministic strategy is proposed to solve the problem. An actor network combining a Softmax activation function and a heuristic algorithm is designed to ensure that its output decision satisfies the constraints. For the small time scale optimization problem (i.e., the per-time slot subproblem P3 at a small time scale), a generalized Benders decomposition algorithm based on continuous convex approximation is designed to solve it. First, the objective function is linearized using the big-M method and converted into a mixed integer nonlinear problem. Then, the successive convex approximation (SCA) algorithm is applied to convert the non-convex components in the MINLP problem into convex components. Finally, for the converted mixed integer nonlinear problem, the generalized Benders algorithm is applied to solve it.

[0161] The embodiment of the present invention also provides a task offloading and resource allocation device for V2V assistance in a vehicle network, such as Figure 4 As shown, the device comprises:

[0162] A first model building unit 401 is configured to build a task queue model;

[0163] A second model building unit 402 is configured to determine an energy consumption model based on the task queue model;

[0164] An optimization problem construction unit 403 is configured to construct an optimization problem based on the energy consumption model with the goal of minimizing long-term energy consumption;

[0165] The optimization problem solving unit 404 is configured to solve the optimization problem to determine a task offloading and resource allocation solution;

[0166] The first model building unit 401 is further configured to build a task queue model by the following method:

[0167] With binary variable o m,n (k) = {0, 1} represents the unloading link selection between the task vehicle TV m and the cooperative vehicle CV n;

[0168] In each time frame, TV can only establish a connection with one CV and one CV can only provide services to one TV. Where N represents the CV set, and M represents the TV set;

[0169] The amount of data generated by the mth task vehicle TV m in time slot t is A m (t);

[0170] The local processing part of TV is The part that is unloaded to the RSU for processing is The part unloaded to the cooperative vehicle CV for processing is is the task offloading ratio of the mth cooperative vehicle CVm in time slot t, and in, Indicates the local processing ratio of TV m, Indicates the proportion of offloading to RSU for processing. Indicates the proportion of unloading to cooperative vehicles;

[0171] At time slot t, the length of TVm's local processing queue is The update formula is:

[0172]

[0173] in represents the amount of data processed locally in time slot t; max represents the maximum value function; τ represents the duration of a time slot; ω represents the CPU cycle required to process a 1-bit task; Indicates the CPU frequency of TV m;

[0174] The transmission rate from TVm to RSU is:

[0175]

[0176] in, Indicates the transmission rate from TVm to RSU; is the number of subcarriers allocated to the V2I link in time frame k, is the V2I link transmission power, represents the channel gain of the V2I link, σ 2 Represents the noise power.

[0177] At time slot t, the length of the V2I link task sending queue of TVm is The update formula is:

[0178]

[0179] in represents the amount of data transmitted in time slot t;

[0180] The V2V link transmission rate from TVm to CVn is:

[0181]

[0182] in, represents the V2V link transmission rate from TVm to CVn; is the V2V link transmission power, is the channel gain of the V2V link; b represents the bandwidth of a subchannel;

[0183] At time slot t, the length of TVm’s V2V task sending queue is The update formula is:

[0184]

[0185] in represents the amount of data sent to CV in time slot t. The number of tasks offloaded from TVm to RSU is:

[0186]

[0187] In the formula, represents the number of tasks offloaded from TVm to RSU; min represents the minimum function; Represents the amount of data sent to the RSU in time slot t.

[0188] At time slot t, the length of the RSU task processing queue is The update formula is:

[0189]

[0190] in Indicates the amount of data processed by the RSU; It represents the calculation frequency of RSU allocated to TV m.

[0191] The number of tasks offloaded to CVn is:

[0192]

[0193] Among them, the binary variable o m,n =1 means TVm can offload tasks to CVn; Indicates the number of tasks offloaded to CVn;

[0194] At time slot t, CVn calculates the queue length as The update formula is:

[0195]

[0196] in Indicates the amount of data sent to CV n; Indicates the calculation frequency of CV n.

[0197] It should be noted that the various device structures described in this embodiment belong to the same technical concept as the previously described method, and achieve the same technical effect through the same principle, which will not be repeated here.

[0198] An embodiment of the present invention further provides a readable storage medium, which stores one or more programs. The one or more programs can be executed by one or more processors to implement the method described in any of the above embodiments.

[0199] In addition, although exemplary embodiments have been described herein, the scope includes any and all embodiments based on the present invention with equivalent elements, modifications, omissions, combinations (e.g., various embodiments intersecting schemes), adaptations or changes. The elements in the claims will be interpreted broadly based on the language adopted in the claims, and are not limited to the examples described in this specification or during the implementation of this application, and the examples will be interpreted as non-exclusive. Therefore, this specification and examples are intended to be considered as examples only, and the true scope and spirit are indicated by the following claims and the full scope of their equivalents.

[0200] The above description is intended to be illustrative rather than restrictive. For example, the above examples (or one or more of them) can be used in combination with each other. For example, those of ordinary skill in the art can use other embodiments when reading the above description. In addition, in the above-mentioned specific embodiments, various features can be grouped together to simplify the present invention. This should not be interpreted as a feature of an invention that is not claimed for protection being necessary for any claim. On the contrary, the subject matter of the present invention may be less than all the features of the embodiments of a specific invention. Thus, the following claims are incorporated into the specific embodiments as examples or embodiments, wherein each claim is independently used as a separate embodiment, and it is considered that these embodiments can be combined with each other in various combinations or arrangements. The scope of the present invention should be determined with reference to the full scope of the equivalent forms of the attached claims and these claims.

Claims

1. A method for task offloading and resource allocation for V2V assistance in a connected vehicle network, characterized in that: The method comprises: Build a task queue model; Based on the task queue model, determining an energy consumption model; Based on the energy consumption model, construct an optimization problem with the goal of minimizing long-term energy consumption; Solving the optimization problem to determine a task offloading and resource allocation solution; Among them, the task queue model is constructed by the following method: With binary variable o m,n (k) = {0, 1} represents the unloading link selection between the task vehicle TV m and the cooperative vehicle CV n; In each time frame, TV can only establish a connection with one CV and one CV can only provide services to one TV. Where N represents the CV set, and M represents the TV set; The amount of data generated by the mth task vehicle TV m in time slot t is A m (t); The local processing part of TV is The part that is unloaded to the RSU for processing is The part unloaded to the cooperative vehicle CV for processing is is the task offloading ratio of the mth cooperative vehicle CVm in time slot t, and in, Indicates the local processing ratio of TV m, Indicates the proportion of offloading to RSU for processing. Indicates the proportion of unloading to cooperative vehicles; At time slot t, the length of TVm's local processing queue is The update formula is: in represents the amount of data processed locally in time slot t; max represents the maximum value function; τ represents the duration of a time slot; ω represents the CPU cycle required to process a 1-bit task; Indicates the CPU frequency of TV m; The transmission rate from TVm to RSU is: in, Indicates the transmission rate from TVm to RSU; is the number of subcarriers allocated to the V2I link in time frame k, is the V2I link transmission power, represents the channel gain of the V2I link, σ 2 represents the noise power; At time slot t, the length of the V2I link task sending queue of TVm is The update formula is: in represents the amount of data transmitted in time slot t; The V2V link transmission rate from TVm to CVn is: in, represents the V2V link transmission rate from TVm to CVn; is the V2V link transmission power, is the channel gain of the V2V link; b represents the bandwidth of a subchannel; At time slot t, the length of TVm’s V2V task sending queue is The update formula is: in represents the amount of data sent to CV in time slot t; The number of tasks offloaded from TVm to RSU is: In the formula, represents the number of tasks offloaded from TVm to RSU; min represents the minimum function; represents the amount of data sent to RSU in time slot t; At time slot t, the length of the RSU task processing queue is The update formula is: in Indicates the amount of data processed by the RSU; Indicates the calculation frequency of RSU allocated to TVm; The number of tasks offloaded to CVn is: Among them, the binary variable o m,n =1 means TVm can offload tasks to CVn; Indicates the number of tasks offloaded to CVn; At time slot t, CVn calculates the queue length as The update formula is: in Indicates the amount of data sent to CV n; Indicates the calculation frequency of CV n.

2. The method according to claim 1, characterized in that: Based on the task queue model, an energy consumption model is determined, specifically including: In time slot t, TVm local processing energy consumption for: in, is the CPU frequency of TV; Energy consumption uploaded from TVm to RSU for: Energy consumption of TVm uploading to CV for: Energy consumed by RSU in processing tasks uploaded by TVm for: Where e represents the energy consumption of RSU processing 1-bit task; Energy consumed by CVn processing tasks for: The total energy consumption in time slot t is: Where E(t) is the total energy consumption in time slot t.

3. The method according to claim 1, characterized in that Based on the energy consumption model, an optimization problem is constructed with the goal of minimizing long-term energy consumption, including: To facilitate the description of the problem, let O(k) = { m,n (k)} m∈M,n∈N , α(t)={α m (t)} m∈M , B(k)={B m (k)} m∈M , P(t)={P m (t)} m∈M , Then the problem of minimizing energy consumption can be expressed as: Among them, P0 represents the problem of minimizing energy consumption, st represents being subject to constraints, C1, C2, C3, C4, C5, C6, C7, C8, C9, C10 and C11 all represent constraints; B max represents the subcarrier overview, f max Indicates the maximum operating frequency of RSU. represents the maximum transmission power of TV m; The queue backlog in time slot t is expressed as: Q(t)={Q TV (t),Q CV (t)} in Determine the Lyapunov function, T-slotLyapunov drift function, and drift plus penalty function; Transform P0 into the first problem P1 by minimizing the upper bound of the drift plus penalty function; The first problem P1 is decomposed into a per-frame sub-problem P2 at a large time scale and a per-time slot sub-problem P3 at a small time scale.

4. The method according to claim 3, characterized in that The Lyapunov function is expressed as: Wherein, L(Q(t)) represents the Lyapunov function.

5. The method according to claim 4, characterized in that The T-slot Lyapunov drift function is expressed as: Among them, Δ T (Q(t)) represents the T-slot Lyapunov drift function.

6. The method according to claim 5, characterized in that The drift plus penalty function It is expressed as: Where V represents the trade-off parameter; T k represents the set of time slots of time frame k.

7. The method according to claim 6, characterized in that The first problem P1 is expressed as: stC1-C11, in The per-frame subproblem P2 at a large time scale and the per-time-slot subproblem P3 at a small time scale are expressed as: stC1-C6,C11, stC7,C8,C10.

8. The method according to claim 7, characterized in that Solving the optimization problem to determine a task offloading and resource allocation solution includes: For the per-frame subproblem P2 at a large time scale, it is modeled as a Markov decision process, using an actor network that combines a Softmax activation function and a heuristic algorithm to ensure that the output decision satisfies the constraints; For the per-time-slot subproblem P3 at a small time scale, the big-M method is first used to linearize the objective function and transform it into a mixed integer nonlinear problem. Then, the successive convex approximation algorithm is used to transform the non-convex components in the MINLP problem into convex components. Finally, the generalized Benders algorithm is used to solve the transformed mixed integer nonlinear problem.

9. A task offloading and resource allocation device for V2V assistance in a vehicle network, characterized in that: The device comprises: A first model building unit is configured to build a task queue model; A second model building unit is configured to determine an energy consumption model based on the task queue model; An optimization problem construction unit is configured to construct an optimization problem based on the energy consumption model with the goal of minimizing long-term energy consumption; an optimization problem solving unit, configured to solve the optimization problem to determine a task offloading and resource allocation solution; The first model building unit is further configured to build a task queue model by the following method: With binary variable o m,n (k) = {0, 1} represents the unloading link selection between the task vehicle TV m and the cooperative vehicle CV n; In each time frame, TV can only establish a connection with one CV and one CV can only provide services to one TV. Where N represents the CV set, and M represents the TV set; The amount of data generated by the mth task vehicle TV m in time slot t is A m (t); The local processing part of TV is The part that is unloaded to the RSU for processing is The part unloaded to the cooperative vehicle CV for processing is is the task offloading ratio of the mth cooperative vehicle CVm in time slot t, and in, Indicates the local processing ratio of TV m, Indicates the proportion of offloading to RSU for processing. Indicates the proportion of unloading to cooperative vehicles; At time slot t, the length of TVm's local processing queue is The update formula is: in represents the amount of data processed locally in time slot t; max represents the maximum value function; τ represents the duration of a time slot; ω represents the CPU cycle required to process a 1-bit task; Indicates the CPU frequency of TV m; The transmission rate from TVm to RSU is: in, Indicates the transmission rate from TVm to RSU; is the number of subcarriers allocated to the V2I link in time frame k, is the V2I link transmission power, represents the channel gain of the V2I link, σ 2 represents the noise power; At time slot t, the length of the V2I link task sending queue of TVm is The update formula is: in represents the amount of data transmitted in time slot t; The V2V link transmission rate from TVm to CVn is: in, represents the V2V link transmission rate from TVm to CVn; is the V2V link transmission power, is the channel gain of the V2V link; represents the number of subcarriers allocated to the V2V link from TVm to CVn; b represents the bandwidth of a subchannel; At time slot t, the length of TVm’s V2V task sending queue is The update formula is: in The amount of data TVm sent to CV at time slot t is offloaded to the number of tasks of RSU: In the formula, represents the number of tasks offloaded from TVm to RSU; min represents the minimum function; represents the amount of data sent to RSU in time slot t; At time slot t, the length of the RSU task processing queue is The update formula is: in Indicates the amount of data processed by the RSU; Indicates the calculation frequency of RSU allocated to TVm The number of tasks offloaded to CVn is: Among them, the binary variable o m,n =1 means TVm can offload tasks to CVn; Indicates the number of tasks offloaded to CVn; At time slot t, CVn calculates the queue length as The update formula is: in Indicates the amount of data sent to CV n; Indicates the calculation frequency of CV n. 10 . A non-transitory computer-readable storage medium storing instructions, which, when executed by a processor, perform the method according to claim 1 .

Images