Idle vehicle resource assisted drone cooperative computing method

Abstract

The application discloses an idle vehicle resource auxiliary unmanned aerial vehicle cooperative computing method. First, an idle vehicle auxiliary unmanned aerial vehicle edge computing architecture is constructed, aiming at reducing the work load of the edge server and processing emergency. Secondly, an incentive mechanism based on game is established, and a multi-stage Stackelberg game model is constructed to coordinate the interests of participants in each stage. In addition, in order to solve the problem of information asymmetry between VOPs and vehicles, we introduce the contract theory, establish a cooperation mechanism between the two parties, and finally ensure the feasibility of the contract. Finally, a multi-stage multi-master multi-slave Nash equilibrium search algorithm is proposed to obtain the resource allocation and pricing scheme.

Description

Technical Field

[0001] This invention relates to the field of edge computing technology for unmanned aerial vehicles (UAVs), and in particular to collaborative computing, resource allocation, and extended utilization of idle resources between edge devices and cloud-edge-device servers. Background Technology

[0002] As data volume and computing demands grow, the resource supply of this cloud computing and edge computing paradigm often cannot meet the demand, leading to resource scarcity and competition. This can affect service quality and efficiency, and even cause business interruptions or failures. Especially in situations such as traffic surges, equipment failures, and natural disasters, the problem of resource overload or unavailability becomes more prominent, causing significant difficulties and losses for drones and operators. Due to the unpredictability of traffic surges, adding additional edge servers can help alleviate resource overload or unavailability to some extent. However, it can also lead to a waste of computing resources. This is because when traffic decreases, the utilization rate of edge servers may drop, while maintenance costs remain constant. This can increase operators' investment and expenditure, reducing resource efficiency. Researchers have found that in daily life, a large number of vehicles are parked in parking lots for extended periods or drive slowly on streets. Statistics show that approximately 70% of private cars are parked for more than 20 hours per day. Due to the large number of vehicles, long parking times, and slow movement speeds, they can be ideal candidates for building a city-wide static backbone network and edge servers. Therefore, we can utilize the idle computing, storage, and bandwidth resources of vehicles to provide edge computing services to other vehicles or networked drones. However, several challenging problems still need to be solved to achieve this goal:

[0003] 1) Vehicles are self-interested and have no obligation to share their idle resources. Therefore, a well-designed incentive mechanism that provides reasonable rewards is essential.

[0004] 2) Vehicles are often unwilling to disclose their private information, which leads to information asymmetry between vehicles and organizers in the VEC network (such as roadside units).

[0005] 3) Since computing resources at the network edge are provided by multiple different service providers (SPs), such as base stations, roadside units, and vehicle units, competition or cooperation may exist among these service providers. Their interactions are complex and difficult to model due to their rationality, self-interest, and the tight coupling between resource demand and supply. Furthermore, resource allocation issues involve coordinating the interests of multiple parties, including operators and drones.

[0006] Inspired by the above, this paper considers designing a reasonable resource allocation mechanism that can meet the computing needs of the Internet of Vehicles (IoV), balance the utilization and quality of service of network edge computing resources, and take into account the interests and strategies of service providers. This paper proposes the concept of idle vehicle assistance, treating parked or unused vehicles as available extended resources. To ensure that requesting drones, ESCPs (Electronic Component Providers), and idle vehicles are satisfied and simultaneously make decisions that meet their respective constraints, we model the resource allocation and pricing problem as a three-stage Stackelberg game model, exploring how to find the optimal strategy for each participant, transforming the problem into a convex optimization problem, and theoretically proving that the game has a unique Nash equilibrium. In the third stage, the original problem is transformed into a convex optimization problem and solved using Karush-Kuhn-Tucker (KKT) conditions. In the second stage, we first prove that the price Nash point between ECSPs exists and is unique, and solve the optimal resource pricing problem. In the first stage, we use KKT conditions to transform the original problem into a Lagrange dual problem. Simultaneously, we design an optimal contract between idle vehicles and VOPs, using contract theory to incentivize nearby idle vehicles to contribute resources, allocating resources to drones with limited computing resources. Summary of the Invention

[0007] This invention addresses the resource allocation and pricing issues in edge environments for budget-constrained drones by providing a method for allocating and pricing idle vehicle-assisted resources.

[0008] This invention is achieved using the following technical solution:

[0009] A method for collaborative computing of unmanned aerial vehicles (UAVs) assisted by idle vehicle resources is characterized by the following steps:

[0010] 1) An incentive-driven idle vehicle-assisted cloud edge computing (VCEC-IDM) framework was established.

[0011] 2) A multi-stage, multi-master, multi-slave (MSMLMF) Stackelberg game model is proposed for resource allocation and pricing in budget-constrained UAV edge environments.

[0012] 3) Extending to situations with information asymmetry, we propose an optimal contract design and a multi-stage, multi-master, multi-slave Nash equilibrium search algorithm to obtain resource allocation and pricing schemes.

[0013] In the above technical solution, further, the VCE-IDM (VCEC-IDM) architecture for idle vehicle-assisted drones described in step 1) comprises four parts:

[0014] (1) Information monitoring: used to collect real-time information from drone purchase resources and monitor the idle status of surrounding vehicles.

[0015] (2) Resource allocation: Based on the MSMLMF Stackelberg game model, the resource allocation strategy is determined and a reasonable resource price is calculated.

[0016] (3) Strategy implementation: Allocate resources to drones according to the proposed resource allocation strategy.

[0017] (4) Idle vehicle auxiliary resource allocation: Recruit a certain number of idle vehicles to expand the computing power of providers with edge / cloud server resources (ECSP), and use contract theory to incentivize nearby idle vehicles to contribute resources to allocate resources to drones with limited computing resources.

[0018] Furthermore, the MSMLMF Stackelberg game model for resource allocation and pricing in budget-constrained drone edge environments, as described in step 2), is defined as follows:

[0019] The interaction process between Vehicle Operators (VOPs), Edge / Cloud Server Resource Providers (ECSPs), and Unmanned Aerial Vehicles (UAVs) is designed as a three-stage, multi-leader, multi-follower Stackelberg game model as follows:

[0020] (1) Third stage: UAVs are followers of the UAV-ECSP stage. After understanding the unit resource price and total capacity of each member in the ECSP, they need to decide to transfer their computing needs to the ECSP. Their utility function is defined as:

[0021]

[0022] Among them, R i,j This represents the satisfaction reward that drone i receives by purchasing resources, denoted by R. i,j =ln(1+β) j f i,j To calculate, here β j It is a constant used to control the sensitivity to rewards. B represents the set of numbers of drones. i This indicates the drone's own budget when purchasing resources.

[0023] At this stage, follower UAVs formulate optimal purchasing strategies based on the leader's ECSP pricing policy, aiming to maximize satisfaction while meeting maximum budget constraints. Mathematically, the problem of maximizing the benefits of drones (followers) can be expressed as:

[0024]

[0025] stf i,j ≥0 (2b)

[0026]

[0027] Among them, (2b) is to ensure the validity of the decision variables, while (2c) represents the drone's own budget B when purchasing resources. i Constraints.

[0028] Phase Two: The ECSP is a follower of the VOP-ECSP phase and determines the quantity of expanded resources to be purchased. Simultaneously, it is a leader in the ECSP-UAV phase and determines the pricing of resources sold to UAVs. The utility function of the j-th ECSP can be expressed as the cost paid by the UAVs when purchasing resources minus their respective costs. These costs consist of three parts: first, the cost of their own resources; second, the energy consumption cost generated by these resources; and third, the cost of the VOP utilizing idle vehicles to expand resources.

[0029]

[0030] in, c j Indicates server s j Unit resource cost It is a VOP to s j The pricing. j The energy consumption formula is expressed as follows: Where 'a' represents the energy consumption discount factor, and e j and k j They represent the components of s respectively. j The constants determined by the switching capacitor and energy cost factor.

[0031] Mathematically speaking, the ECSP profit maximization problem can be expressed as:

[0032]

[0033] stp j ≥c j (4b)

[0034]

[0035] Among them, (4b) ensures that the pricing decisions made by the ECSP are accurate and reasonable. j It will not be lower than the server's s j Unit resource cost c j This prevents losses. (4c) and (4d) ensure server s j Its total computing resources Q j And the total resources extended to VOP Total resources F sufficient to cover drone requests jIt is worth noting that since cloud server s1 has abundant computing resources, we do not impose any restrictions on its resource capacity.

[0036] Phase 1: The VOP is the leader of the VOP-ECSP phase, determining the unit price of computing resources and providing computing services to the ECSP. Its utility function is defined as the fee charged minus its cost.

[0037]

[0038] The fees charged here refer to the fees that VOPs charge to ECSPs, while the costs refer to the actual costs incurred by VOPs when providing computing services. m VOP represents the value of the m-th vehicle. m Payment. Indicates a VOP type θ m The number of vehicles, of which This is the total number of vehicles.

[0039] Mathematically speaking, the problem of maximizing VOP profits can be expressed as:

[0040]

[0041] Among them, (6b) will include vehicle v m The amount of resources that can be contributed, f m Limited to Within the range, Indicates v m Total available resources. In addition, resources purchased by ECSP from VOP. The total shared capacity available to all vehicles cannot exceed the limit (6c). (6d) ensures the feasibility and universality of the decision. It is worth noting the price at which VOPs sell resources. The price must be lower than ECSP's p j Otherwise, ECSP would have no reason to purchase resources from VOP.

[0042] Furthermore, the optimal contract model design described in step 3) is as follows:

[0043] 1) Optimization problem and its simplification: Vehicle utility vehicle v m utility function U m It is expressed as the profit obtained from selling its resources minus its energy consumption costs:

[0044] U m =p m -s m E m (7)

[0045] Energy consumption can be expressed as: Among them, sm f is the energy consumption discount factor. m Indicates vehicle v m The amount of computing resources contributed. and Represent the energy cost coefficient and the coefficient derived from v, respectively. m The constant is determined by the switched capacitor.

[0046] Since VOPs lack accurate type information for each vehicle, we utilize the discrete uniform distribution θ of vehicle types in historical data. m and probability distribution λ m To optimize the expected utility of VOP. Specifically, based on equation (7), the vehicle v is defined. m The type is:

[0047]

[0048] So U m The expression can be represented as:

[0049]

[0050] 2) To attract temporary vehicle cooperation and ensure the authenticity of the information provided, we introduce contract theory to model the relationship between VOPs and vehicles, ensuring that all vOPs... m Accepting true feasibility θ m The contract should meet the following IR and IC conditions:

[0051] IR condition: Ensure that the vehicle obtains non-negative utility when accepting any contract (f m ,p m ):

[0052]

[0053] IC conditions:

[0054]

[0055] Among them, f m′ and p m′ They represent v respectively m Dishonest CPU contributions and rewards. We note that these values ​​are reported by the vehicle model θ, representing the error type parameter. m′ Determined by VOP, This refers to a collection of idle vehicles.

[0056] From an incentive perspective, the reward amount available to vehicles should exhibit a monotonically increasing trend. Generally, we can assume that the more types of vehicles available, the more willing the user is to contribute resources, and therefore the corresponding reward will increase accordingly. Meanwhile, in the context of asymmetric information contract design, VOP needs to design a set of contractual terms (f m ,p m To maximize U vop This ensures compliance with both IR and IC constraints. Therefore, the optimization problem (P3) can be expressed as a design contract satisfying the following conditions:

[0057]

[0058] st(6b)-(6d)

[0059]

[0060] Monotonic: f0≤f1<... <f m (12d)

[0061]

[0062] Constraints (12b) and (12c) guarantee the feasibility of the contract. Here, f0 represents a contract of type θ0, denoted as (f0,p0) = (0,0). Furthermore, constraint (12d) guarantees that the ability or willingness to share resources varies with type θ. m It increases with the increase of.

[0063] In (P3'), we note that constraints (12b) and (12c) exhibit high dimensionality and coupling (with IR constraints and This complicates the solution process. To further simplify the analysis, we use the following lemma to simplify the constraints:

[0064] Lemma 1: Due to the monotonicity of the IC condition in (P3'), it can be simplified to locally upward excitation compatibility (LUIC) and locally downward excitation compatibility (LDIC):

[0065]

[0066] Lemma 2: Assume that in (P3'), the vehicle obtains the optimal contract. The IC constraint can be replaced with:

[0067]

[0068] Lemma 3: Given any feasible contract (f) m ,p mThe IR condition holds if and only if the utility of vehicle type 1 (θ1) is sufficiently small (close to 0). Therefore, the IR condition can be replaced by:

[0069]

[0070] Lemma 4: Based on Lemmas 2 and 3, we can eliminate constraints (12b) and (12c), further simplifying (P3') to obtain the optimal solution:

[0071]

[0072] Using the above lemma, we can eliminate the IC and IR constraints, and by substituting equation (17) into (P3'), we obtain the final simplified problem (P3”):

[0073]

[0074] st(6b)-(6d),(12d).

[0075] Considering the different needs and interests of tasks and vehicles, we propose a multi-stage, multi-master, multi-slave nash equilibrium search algorithm. The specific steps are as follows:

[0076] ① Initialization: Set the convergence tolerance ∈.

[0077] ②Fixed Obtaining the optimal strategy from unmanned aerial vehicles (UAVs)

[0078] ③ Fix Obtain the optimal strategy from vehicle operators (ECSPs)

[0079] ④ Fix Obtain the optimal strategy from edge computing service providers (ECSPs)

[0080] ⑤ Calculate U according to formulas (1), (3) and (5) respectively. i ,

[0081] ⑥ Repeat steps 2 to 5 until the convergence tolerance is reached.

[0082] ⑦ End: Output the optimal strategy for all entities.

[0083] The inventive principle of this invention:

[0084] This invention primarily addresses the resource allocation and pricing problem in edge computing environments for unmanned aerial vehicles (UAVs). It designs an idle vehicle-assisted resource allocation and pricing system. Under the premise of satisfying their respective constraints, the system aims to maximize the satisfaction of resource-constrained UAVs and the highest profit of the ESCP (Enhanced Resource Conversion Rate). It seeks the Nash equilibrium point of the Stackelberg game model between the pricing given by the VOP (Vehicle Operator) and the corresponding resource allocation decisions. Extending this to situations of information asymmetry, the invention proposes an optimal contract design and a multi-stage, multi-master, multi-slave Nash equilibrium search algorithm to obtain a stable resource allocation and pricing scheme.

[0085] The beneficial effects of this invention are as follows:

[0086] This invention proposes for the first time a collaborative computing method for UAVs assisted by idle vehicle resources. This method can respond to any resource requests issued by resource-constrained UAVs and attract surrounding idle vehicle resources to expand the computing power of the VEC server. Under the premise of satisfying their respective constraints, with the goal of maximizing the satisfaction of resource-constrained UAVs and the highest profit of ESCP, the method seeks the Nash equilibrium point of the Stackelberg game model between the pricing given by VOP and the corresponding resource allocation decisions. It extends to the case of information asymmetry, proposes an optimal contract design, and proposes a multi-stage, multi-master, multi-slave Nash equilibrium search algorithm to obtain a stable resource allocation and pricing scheme. Attached Figure Description

[0087] Figure 1 An edge computing system architecture for assisting unmanned aerial vehicles in idle vehicles;

[0088] Figure 2 The process of the MSMLMF Stackelberg game model;

[0089] Figure 3 The relationship between the convergence of the MSMLMF Stackelbergbo game and the number of iterations;

[0090] Figure 4 The impact of different numbers of drones on the utility, resource pricing, and resource purchasing of participating entities within the system;

[0091] Figure 5 The impact of different numbers of ESCPs on the utility, resource pricing, and resource purchasing behavior of participating entities within the system;

[0092] Figure 6 The impact of different numbers of idle vehicles on the utility, resource pricing, and resource purchasing of participating entities within the system;

[0093] Figure 7 A comparison of social welfare under different models and with different numbers of drones, ESCPs, and idle vehicles;

[0094] Figure 8 To compare overall social welfare with a model without vehicle assistance.

[0095] Figure 9 Optimal contract revenue for different types of idle vehicles. Detailed Implementation

[0096] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0097] A method for collaborative computing of unmanned aerial vehicles (UAVs) assisted by idle vehicle resources is characterized by the following steps:

[0098] 1) An incentive-driven, idle vehicle-assisted cloud edge computing (VCEC-IDM) framework was established.

[0099] 2) A multi-stage, multi-master, multi-slave (MSMLMF) Stackelberg game model is proposed for resource allocation and pricing in budget-constrained UAV edge environments;

[0100] 3) Extending to situations with information asymmetry, we propose an optimal contract design and a multi-stage, multi-master, multi-slave Nash equilibrium search algorithm to obtain resource allocation and pricing schemes.

[0101] The proposed architecture for edge computing of idle vehicles-assisted drones (VCEC-IDM) in this invention is as follows: Figure 1 As shown, it includes four parts:

[0102] (1) Information monitoring: used to collect real-time information from drone purchase resources and monitor the idle status of surrounding vehicles.

[0103] (2) Resource allocation: Based on the MSMLMF Stackelberg game model, determine the resource allocation strategy and calculate a reasonable resource price;

[0104] (3) Strategy implementation: Allocate resources to drones according to the proposed resource allocation strategy.

[0105] (4) Idle vehicle auxiliary resource allocation: Recruit a certain number of idle vehicles to expand the computing power of providers with edge / cloud server resources (ECSP), and use contract theory to incentivize nearby idle vehicles to contribute resources to allocate resources to drones with limited computing resources.

[0106] Figure 2A flowchart of the MSMLMF Stackelberg game model is provided. In the first stage of the game, we designed a contract-based incentive recruitment mechanism. In this mechanism, the VOP, as the employer, offers contracts to various types of parked vehicles (employees) for them to choose from. Considering that the types of vehicles may differ, each contract design should fully consider the incentive for idle vehicles to contribute idle resources. Simultaneously, the VOP, as the leader in the VOP-ECSP game stage, determines the final selling price based on the resource purchase volume of its subordinate ECSPs and considering the available idle resources. In the second stage, the edge / cloud servers managed by the ECSP are both followers in the VOP-ECSP game stage and leaders in the ECSP-UAV stage. The ECSP determines its unit resource selling price based on the quantity of resources purchased from the VOP and the quantity of resources sold to lower-level drones. In the third stage, all drones are followers in the ECSP-UAV stage. Drones decide the quantity of resources to purchase from the ECSP based on the price offered by the ECSP and considering that they should not exceed their own budget.

[0107] (1) Third stage: UAVs are followers of the UAV-ECSP stage. After understanding the unit resource price and total capacity of each member in the ECSP, they need to decide to transfer their computing needs to the ECSP. Their utility function is defined as:

[0108]

[0109] Among them, R i,j This represents the satisfaction reward that drone i receives by purchasing resources, denoted by R. i,j =ln(1+β) j f i,j To calculate, here β j It is a constant used to control the sensitivity to rewards. B represents the set of numbers of drones. i This indicates the drone's own budget when purchasing resources.

[0110] At this stage, follower UAVs formulate optimal purchasing strategies based on the leader's ECSP pricing policy, aiming to maximize satisfaction while meeting maximum budget constraints. Mathematically, the problem of maximizing the benefits of drones (followers) can be expressed as:

[0111]

[0112] stf i,j ≥0 (2b)

[0113]

[0114] Among them, (2b) is to ensure the validity of the decision variables, while (2c) represents the drone's own budget B when purchasing resources. i Constraints.

[0115] (2) Second stage: ECSP is a follower of the VOP-ECSP stage and determines the amount of expanded resources to be purchased. It is also a leader of the ECSP-UAV stage and determines the pricing of resources sold to UAVs. The utility function of the j-th ECSP can be expressed as the cost paid by the UAVs when purchasing resources minus their respective costs. The costs consist of three parts: first, the cost of their own resources; second, the energy consumption cost generated by these resources; and third, the cost of the VOP using idle vehicles to expand resources.

[0116]

[0117] in, c j Indicates server s j Unit resource cost It is a VOP to s j The pricing. j The energy consumption formula is expressed as follows: Where 'a' represents the energy consumption discount factor, and e j and k j They represent the components of s respectively. j The constants determined by the switching capacitor and energy cost factor.

[0118] Mathematically speaking, the ECSP profit maximization problem can be expressed as:

[0119]

[0120] stp j ≥c j (4b)

[0121]

[0122] Among them, (4b) ensures that the pricing decisions made by the ECSP are accurate and reasonable. j It will not be lower than the server's s j Unit resource cost c j This prevents losses. (4c) and (4d) ensure server s j Its total computing resources Q j And the total resources extended to VOP Total resources F sufficient to cover drone requests j It is worth noting that since cloud server s1 has abundant computing resources, we do not impose any restrictions on its resource capacity.

[0123] (3) First stage: VOP is the leader of the VOP-ECSP stage, determining the unit price of computing resources and providing computing services to ECSP. Its utility function is defined as the fee charged minus its cost:

[0124]

[0125] The fees charged here refer to the fees that VOPs charge to ECSPs, while the costs refer to the actual costs incurred by VOPs when providing computing services. m VOP represents the value of the m-th vehicle. m Payment. Indicates a VOP type θ m The number of vehicles, of which This is the total number of vehicles.

[0126] Mathematically speaking, the problem of maximizing VOP profits can be expressed as:

[0127]

[0128] Among them, (6b) will include vehicle v m The amount of resources that can be contributed, f m Limited to Within the range, Indicates v m Total available resources. In addition, resources purchased by ECSP from VOP. The total shared capacity available to all vehicles cannot exceed the limit (6c). (6d) ensures the feasibility and universality of the decision. It is worth noting the price at which VOPs sell resources. The price must be lower than ECSP's p j Otherwise, ECSP would have no reason to purchase resources from VOP.

[0129] The process of resource allocation and pricing using the method of this invention is as follows:

[0130] 1) Information monitoring: Used to collect real-time information from drone purchase resources and monitor the idle status of surrounding vehicles;

[0131] 2) Resource Allocation: Based on the MSMLMF Stackelberg game model, a resource allocation strategy is determined, and a reasonable resource price is calculated. The specific steps for finding the Nash equilibrium point of the MSMLMF Stackelberg game model are as follows:

[0132] ① Initialization: Set the convergence tolerance ∈.

[0133] ②Fixed Obtaining the optimal strategy from unmanned aerial vehicles (UAVs)

[0134] ③ Fix Obtain the optimal strategy from vehicle operators (ECSPs)

[0135] ④ Fix Obtain the optimal strategy from edge computing service providers (ECSPs)

[0136] ⑤ Calculate U according to formulas (1), (3) and (5) respectively. i , U vop .

[0137] ⑥ Repeat steps 2 to 5 until the convergence tolerance is reached.

[0138] ⑦ End: Output the optimal strategy for all entities.

[0139] 3) Strategy implementation: Allocate resources to drones according to the established resource allocation strategy.

[0140] 4) Repeat steps 2)-3) until all drone resource requests are met.

[0141] 5) Simulation Results

[0142] Figure 3 The algorithm demonstrates a multi-stage, multi-master, multi-slave Nash equilibrium search algorithm, in which the utility of each participating entity gradually converges to the Nash equilibrium point as the number of game iterations increases. Figure 4 , Figure 5 and Figure 6 The impact of varying numbers of drones, ESCPs, and idle vehicles on the utility, resource pricing, and resource purchasing behavior of participating entities within the system is presented. Data results are compared between the contract-based information asymmetry model (IA-C) proposed in this invention and conventional solutions (the Stackelberg-based information-free asymmetry model (NIA-S) and linear pricing model (LP)). Figure 7 It can be seen that, under different numbers of drones, ESCPs, and idle vehicles, our model IA-C has better average social welfare than LP, and is relatively close to NIA-S in terms of maximizing social welfare. Figure 8 The differences between this model and the model without vehicle assistance are demonstrated. Compared with the ECSP without vehicle assistance, our method improves the average overall social welfare. Figure 9 Five different types of idle vehicles are displayed (θ2, θ4, θ6, θ8, θ...). 10 The optimal contract payoff curve is obtained to verify the feasibility of contract-based solutions under information asymmetry (i.e., IR and IC).

Claims

1. A method for collaborative computing of unmanned aerial vehicles (UAVs) assisted by idle vehicle resources, characterized in that, Includes the following steps: 1) An incentive-driven idle vehicle-assisted cloud edge computing (VCEC-IDM) framework was established. 2) A multi-stage, multi-master, multi-slave (MSMLMF) Stackelberg game model is proposed for the allocation and pricing of airborne edge computing resources for budget-constrained UAVs. 3) Extending to situations with information asymmetry, we propose an optimal contract design and a multi-stage, multi-master, multi-slave Nash equilibrium search algorithm to obtain resource allocation and pricing schemes.

2. The method for assisting UAV collaborative computing using idle vehicle resources as described in claim 1, characterized in that, The VCE-IDM (Vehicle-Assisted Drone) edge computing architecture described in step 1) consists of four parts: (1) Information monitoring: used to collect real-time information from drone purchase resources and monitor the idle status of surrounding vehicles. (2) Resource allocation: Based on the MSMLMF Stackelberg game model, the resource allocation strategy is determined and a reasonable resource price is calculated. (3) Strategy implementation: Allocate resources to drones according to the proposed resource allocation strategy. (4) Idle vehicle auxiliary resource allocation: Recruit a certain number of idle vehicles to expand the computing power of providers with edge / cloud server resources (ECSP), and use contract theory to incentivize nearby idle vehicles to contribute resources to allocate resources to drones with limited computing resources.

3. The idle vehicle-assisted ESCP resource allocation and pricing method based on game theory and contract incentive mechanism as described in claim 2, characterized in that, The MSMLMF Stackelberg game model for resource allocation and pricing of budget-constrained drones in edge environments, as described in step 2), is defined as follows: The interaction process between vehicle operators (VOPs), edge / cloud server resource providers (ECSPs), and unmanned aerial vehicles (UAVs) is designed as a three-stage, multi-leader, multi-follower Stackelberg game model as follows: (1) Third stage: UAVs are followers of the UAV-ECSP stage. After understanding the unit resource price and total capacity of each member in the ECSP, they need to decide to transfer their computing needs to the ECSP. Their utility function is defined as: Among them, R i,j This represents the satisfaction reward that drone i receives by purchasing resources, denoted by R. i,j =ln(1+β) j f i,j To calculate, here β j It is a constant used to control the sensitivity to rewards. B represents the set of numbers of drones. i This indicates the drone's own budget when purchasing resources. At this stage, follower UAVs formulate optimal purchasing strategies based on the leader's ECSP pricing policy, aiming to maximize satisfaction while meeting maximum budget constraints. Mathematically, the problem of maximizing the benefits of drones (followers) can be expressed as: s.t.f i,j ≥0 (2b) where (2b) is to guarantee the validity of decision variables, and (2c) represents the constraint of the UAV's own budget B when purchasing resources. i . (2) Second stage: ECSP is a follower of the VOP-ECSP stage and determines the amount of expanded resources to be purchased. It is also a leader of the ECSP-UAV stage and determines the pricing of resources sold to UAVs. The utility function of the j-th ECSP can be expressed as the cost paid by the UAVs when purchasing resources minus their respective costs. The costs consist of three parts: first, the cost of their own resources; second, the energy consumption cost generated by these resources; and third, the cost of the VOP using idle vehicles to expand resources. in, c j Indicates server s j Unit resource cost It is a VOP to s j The pricing. j The energy consumption formula is expressed as follows: Where 'a' represents the energy consumption discount factor, and e j and k j They represent the components of s respectively. j The constants determined by the switching capacitor and energy cost factor. Mathematically speaking, the ECSP profit maximization problem can be expressed as: s.t.p j ≥c j (4b) Among them, (4b) ensures that the pricing decisions made by the ECSP are accurate and reasonable. j It will not be lower than the server's s j Unit resource cost c j This prevents losses. (4c) and (4d) ensure server s j Its total computing resources Q j And the total resources extended to VOP Total resources F sufficient to cover drone requests j It is worth noting that since cloud server s1 has abundant computing resources, we do not impose any restrictions on its resource capacity. (3) First stage: VOP is the leader of the VOP-ECSP stage, determining the unit price of computing resources and providing computing services to ECSP. Its utility function is defined as the fee charged minus its cost: The fees charged here refer to the fees that VOPs charge to ECSPs, while the costs refer to the actual costs incurred by VOPs when providing computing services. m VOP represents the value of the m-th vehicle. m Payment. Indicates a type close to VOP θ m The number of vehicles, of which This is the total number of vehicles. Mathematically speaking, the problem of maximizing VOP profits can be expressed as: Among them, (6b) will include vehicle v m The amount of resources that can be contributed, f m Limited to Within the range, Indicates v m Total available resources. In addition, resources purchased by ECSP from VOP. The total shared capacity available to all vehicles cannot exceed the limit (6c). (6d) ensures the feasibility and universality of the decision. It is worth noting the price at which VOPs sell resources. The price must be lower than ECSP's p j Otherwise, ECSP would have no reason to purchase resources from VOP.

4. The case of extended information asymmetry as described in claim 3, characterized in that, The optimal contract model design described in step 3) is as follows: 1) Optimization problem and its simplification: Vehicle utility vehicle v m utility function U m It is expressed as the profit obtained from selling its resources minus its energy consumption costs: U m =p m -s m E m , (7) Energy consumption can be expressed as: Among them, s m f is the energy consumption discount factor. m Indicates vehicle v m The amount of computing resources contributed. and Represent the energy cost coefficient and the coefficient derived from v, respectively. m The constant is determined by the switched capacitor. Since VOPs lack accurate type information for each vehicle, we utilize the discrete uniform distribution θ of vehicle types in historical data. m and probability distribution λ m To optimize the expected utility of VOP. Specifically, based on equation (7), the vehicle v is defined. m The type is: So U m The expression can be represented as: 2) To attract temporary vehicle cooperation and ensure the authenticity of the information provided, we introduce contract theory to model the relationship between VOPs and vehicles, ensuring that all vOPs... m Accepting real feasibility θ m The contract should meet the following IR and IC conditions: IR condition: Ensure that the vehicle obtains non-negative utility when accepting any contract (f m ,p m ): IC conditions: Among them, f m′ and p m′ They represent v respectively m Dishonest CPU contributions and rewards. We note that these values ​​are reported by the vehicle model θ, representing the error type parameter. m′ Determined by VOP, This refers to a collection of idle vehicles. 3) From an incentive perspective, the reward amount obtainable by a vehicle should exhibit a monotonically increasing trend. Generally, we can assume that the more types of vehicles available, the more willing the user is to contribute resources, and therefore the corresponding reward will also increase. Meanwhile, in the context of asymmetric information contract design, VOP needs to design a set of contractual terms (f m ,p m To maximize U vop This ensures compliance with both IR and IC constraints. Therefore, the optimization problem (P3) can be expressed as a design contract satisfying the following conditions: st(6b)-(6d) Monotonic:f0≤f1<...<f m , (12d) Constraints (12b) and (12c) guarantee the feasibility of the contract. Here, f0 represents a contract of type θ0, denoted as (f0,p0) = (0,0). Furthermore, constraint (12d) guarantees that the ability or willingness to share resources varies with type θ. m It increases with the increase of. 4) In (P3'), we note that constraints (12b) and (12c) exhibit high dimensionality and coupling (with IR constraints and Constraints can complicate the solution process. To further simplify the analysis, we use the following lemma to simplify the constraints: Lemma 1: Due to the monotonicity of the IC condition in (P3'), it can be simplified to locally upward excitation compatibility (LUIC) and locally downward excitation compatibility (LDIC): Lemma 2: Assume that in (P3'), the vehicle obtains the optimal contract. The IC constraint can be replaced by: Lemma 3: Given any feasible contract (f) m ,p m The IR condition holds if and only if the utility of vehicle type 1 (θ1) is sufficiently small (close to 0). Therefore, the IR condition can be replaced by: Lemma 4: Based on Lemmas 2 and 3, we can eliminate constraints (12b) and (12c), further simplifying (P3') to obtain the optimal solution: Using the above lemma, we can eliminate the IC and IR constraints, and by substituting equation (17) into (P3'), we obtain the final simplified problem (P3”): st(6b)-(6d),(12d). Considering the different needs and interests of tasks and vehicles, we propose a multi-stage, multi-master, multi-slave nash equilibrium search algorithm. The specific steps are as follows: ① Initialization: Set the convergence tolerance ∈. ②Fixed Obtaining the optimal strategy from unmanned aerial vehicles (UAVs) ③ Fix Obtain the optimal strategy from vehicle operators (ECSPs) ④ Fix Obtain the optimal strategy from edge computing service providers (ECSPs) ⑤ Calculate U according to formulas (1), (3) and (5) respectively. i , U vop . ⑥ Repeat steps 2 through 5 until the convergence tolerance is reached. ⑦ End: Output the optimal strategy for all entities. .

5. The method for assisting UAV collaborative computing using idle vehicle resources as described in any one of claims 1-3, characterized in that, The process of optimal resource allocation and pricing based on this method is as follows: 1) Request phase: Drones with limited equipment resources submit a request to the ECSP to purchase computing services; 2) Recruitment phase: VOP observes the demand from ECSP and uses a contract-based incentive mechanism to recruit idle vehicle resources to supplement the computing power of ECSP; 3) Competition stage: Based on the willingness to purchase drones, and using the MSMLMF Stackelberg game model, find the Nash equilibrium point for resource allocation and pricing strategies; 4) Resource allocation phase: Provide computing resources to drones according to resource allocation and pricing strategies.

Images