Edge Computing Resource Optimization Method for Air Base Station-Assisted Intelligent Transportation Systems

Abstract

An edge computing resource optimization method for an intelligent transportation system assisted by an aerial base station, which consists of steps of generating tasks by modeling target vehicles, determining the transmission power and the size of offloading subtasks, determining the transmission delay of offloading subtasks, determining offloading roadside units, determining the computing delay of offloading subtasks, determining the computing delay of local subtasks, determining the total processing delay, and returning the computing results of offloading subtasks. According to the channel gain between the vehicle and the offloading object and the available computing resources of the vehicle and the offloading object, the present invention determines how the vehicle divides tasks and allocates transmission power through the dichotomy method, and adopts an approximately optimal algorithm to determine which roadside unit calculates part of the subtasks. After offloading the subtasks to the roadside unit and the aerial base station, the tasks can be processed in parallel, reducing the processing time delay of the tasks.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle networking and edge computing, and particularly to an edge computing resource optimization method for an intelligent transportation system assisted by an aerial base station. Background Art

[0002] Thanks to the rapid development of communication technologies and artificial intelligence, the scenarios applied in intelligent transportation systems have become increasingly rich, and the representative technology among them is autonomous driving. The autonomous decision-making system in autonomous vehicles can process the information generated by vehicle-mounted devices (such as cameras, radars, lidar sensors, ultrasonic sensors, global positioning systems, etc.). Due to the quantity and diversity of the information, the processing process requires relatively complex calculations. Especially for intelligent transportation systems with huge resource requirements, the design requirements for computing solutions are higher. Relying solely on the computing power of vehicles themselves is difficult to meet the low-latency and high-reliability requirements in autonomous driving. Therefore, edge computing technology can be used in intelligent transportation systems to reduce system latency and enhance reliability.

[0003] In intelligent transportation systems, roadside units are usually used as edge servers. For the tasks generated by vehicles, they can be selected to be offloaded to roadside units with stronger computing capabilities for processing to reduce the computing load of the vehicles. However, existing edge computing-based vehicle networking task offloading methods only consider processing tasks on local and roadside units connected to the vehicles, and do not consider the forwarding of tasks between roadside units. Therefore, in the case of a large number of vehicles with uneven location distributions, roadside units also face situations of insufficient computing resources or unbalanced loads. In addition, the limited communication range between edge servers and vehicles is the main factor for task offloading failures.

[0004] One technical problem that needs to be solved in current intelligent transportation systems is to consider a vehicle task offloading method that can improve resource utilization rate, minimize task processing latency, and ensure successful task offloading. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to overcome the above-mentioned technical problems, introduce an aerial base station in an intelligent transportation system, and provide an edge computing resource optimization method for an intelligent transportation system assisted by an aerial base station that maximizes resource utilization rate, can reduce task processing latency, and ensure successful task offloading.

[0006] The technical solution adopted to solve the above-mentioned technical problem consists of the following steps:

[0007] (1) Model the tasks generated by the target vehicle

[0008] Determine the tasks generated by the target vehicle according to the following formula:

[0009] Ti ={D i ,C i}

[0010] where D i is the data size of the task generated by vehicle i, i ∈ {1, 2, …, N}, N is the total number of vehicles, and N is a finite positive integer, and C i is the number of CPU cycles required for the execution of this task.

[0011] (2) Determine the transmission power and the size of the offloading subtasks

[0012] The task generated by the target vehicle is split into three subtasks {α i T i , β i T i , θ i T i} where α i T i is the roadside unit subtask, β i T i is the aerial base station subtask, and θ i T i is the local subtask. α i T i is offloaded to a certain roadside unit k, k ∈ {1, 2, …, R}, R is the total number of roadside units, and R ∈ [3, 10], β i T i is offloaded to an aerial base station, and θ i T i is processed locally by the target vehicle. The vehicle uses non - orthogonal multiple access to send α i T i and β i T i to the roadside unit j and the aerial base station closest to the target vehicle respectively. The transmission power for sending subtask α i T i is P i,r , and the transmission power for sending subtask β i T i is P i,a , P i,r , P i,a ≥0, and the total transmission power P i,r + P i,a shall not exceed the transmission power limit P max . According to the position of the current vehicle, the task generated by the target vehicle, and the computing resources, the computing resources that the aerial base station and roadside unit j can allocate to the target vehicle, use the binary search method to determine α i , β i , θ i , Pi,r ,P i,a The value of .

[0013] (3) Determine the transmission delay of the offloaded subtask

[0014] The transmission delay of the roadside unit subtask is determined as follows: and the transmission delay of the airborne base station subtask

[0015]

[0016]

[0017]

[0018]

[0019]

[0020]

[0021] Among them, α i D i is the data volume of the roadside unit subtask, β i D i is the data volume of the air base station subtask, B is the total uplink bandwidth of the system, B∈[15,25], in MHz, W is the number of subchannels, W∈[12,24], H i,1 is the channel gain from the target vehicle to the connected roadside unit j, H i,2 is the channel gain from the target vehicle to the aerial base station A, N0 is the power of Gaussian noise, N0∈[-120,-100], in dBm, g i is the Rayleigh fading component, which obeys a complex Gaussian distribution with a mean of 0 and a variance of 1. i is the straight-line distance from the target vehicle to the connected roadside unit, η is the path loss factor, g0 represents the channel gain between the target vehicle and the aerial base station when the reference distance is 1m, H is the altitude of the aerial base station, ‖y i ‖ is the horizontal projection distance from the target vehicle to the aerial base station.

[0022] (4) Determine the unloading roadside unit

[0023] The roadside unit subtask sent by the vehicle to roadside unit j is forwarded to roadside unit k through a wired link. Roadside unit k is the unloading roadside unit, and the approximate optimal method is used to determine the unloading roadside unit k.

[0024] (5) Determine the computational delay of the offloaded subtask

[0025] Determine the roadside unit subtask α according to the following formula i T i Computing delay of And the air base station subtask β i T i Computing delay of

[0026]

[0027]

[0028] f a,i =F A / N

[0029] Where α i C i Represents the number of CPU cycles required to execute the roadside unit subtask, and f k,i Is the computing power resource allocated to the target vehicle by the roadside unit k, Is the transmission rate of the wired link between the roadside unit k and the roadside unit j, and β i C i Represents the number of CPU cycles required to execute the air base station subtask, and f a,i Is the computing power resource allocated to the target vehicle by the air base station, and F A Is the total computing power resource of the air base station.

[0030] (6) Determine the computing delay of the local subtask

[0031] Determine the computing delay of the local subtask according to the following formula

[0032]

[0033] Where θ i C i Is the number of CPU cycles required to compute the local subtask, and f i Is the local computing resource.

[0034] (7) Determine the total processing delay

[0035] Determine the total processing delay T according to the following formula total :

[0036]

[0037] (8) Return the calculation result of the offloading subtask

[0038] The calculation result of the subtask is returned to the target vehicle; the edge computing resource optimization method for the intelligent transportation system assisted by the air base station is completed.

[0039] In the step (2) of determining the transmission power and the size of the unloading subtask of the present invention, the binary search method is used to determine α i , β i ,θ i , transmit power P i,r , transmit power P i,a The method is:

[0040] 1) Determine α using the following formula i , β i ,θ i :

[0041]

[0042] β i =1-α i -θ i

[0043] θ i =(H θ +L θ ) / 2

[0044] Among them, H θ , L θ are θ obtained after binary search i Upper and lower limits of values;

[0045] 2) Determine the functional relationship P between the transmission power of the vehicle unloading roadside unit subtask and the roadside unit subtask size according to the following formula: i,r (α i ) and the functional relationship between the transmission power of the vehicle unloading aerial base station subtask and the roadside unit subtask size P i,a (α i ):

[0046] P i,r (α i )=(2 X -1)N0 / H i,1

[0047] P i,a (α i )=(2 Y -1)((2 X -1)N0 / H i,1 +N0 / H i,2 )

[0048]

[0049]

[0050] In step (4) of the present invention, the approximate optimal method determines the unloading roadside unit k as:

[0051] Initialize the unloading strategy so that the roadside unit subtask α of all vehicles i T i Deliver it to the connected roadside unit for calculation, traverse all vehicles, if there is a roadside unit subtask α for a certain vehicle i T i After forwarding to other roadside units, the total time delay is reduced by a ratio ε, ε∈[0.01,0.1], then the forwarding is performed and this operation is repeated until the total time delay no longer decreases.

[0052] In step (1) of the present invention, the tasks generated by modeling the target vehicle are:

[0053] The task T generated by the modeled target vehicle is determined as follows: i :

[0054] T i ={D i ,C i}

[0055] Among them, D i The data size of the task generated for vehicle i, i∈{1,2,…,N}, N is the total number of vehicles, N∈[12,24], D i The value of C is 400~600Kbits. i The number of CPU cycles required for the task to execute, C i The value ranges from 320,000 to 720,000 cycle / s.

[0056] When determining the unloading object and the number of unloading subtasks, the present invention divides the task generated by the target vehicle into multiple subtasks based on the current position of the target vehicle, the tasks generated by the target vehicle, and the computing resources of the current target vehicle and the unloading object. Some subtasks are unloaded to roadside units and aerial base stations. Tasks can be executed in parallel, reducing the processing delay of tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 This is a flowchart of Example 1 of the present invention. DETAILED DESCRIPTION

[0058] The present invention will be further described in detail below with reference to the accompanying drawings and examples, but the present invention is not limited to the following embodiments.

[0059] Example 1

[0060] The edge computing resource optimization method for an intelligent transportation system assisted by an airborne base station in this embodiment consists of the following steps (see Figure 1 ):

[0061] (1) Modeling the tasks generated by the target vehicle

[0062] The task T generated by the modeled target vehicle is determined as follows: i :

[0063] T i ={D i ,C i}

[0064] Among them, D i The data size of the task generated for vehicle i, i∈{1,2,…,N}, N is the total number of vehicles, N∈[12,24], in this embodiment, the value of N is 18, D i The value of is 400~600Kbit. i The value of C is 500Kbit. i The number of CPU cycles required for the task to execute, C i The value of is 320000~720000cycle, in this embodiment, C i The value is 520000cycle.

[0065] (2) Determine the transmit power and offload subtask size

[0066] The task generated for the target vehicle is divided into three subtasks {α i T i ,β i T i ,θ i T i}, where α i T i is the roadside unit subtask, β i T i is the aerial base station subtask, θ i T i is the local subtask, α i T i Unload to a roadside unit k, k∈{1,2,…,R}, R is the total number of roadside units, R∈[3,10], in this embodiment, R is 6, β i T i Unload to an aerial base station,θ i T i The target vehicle processes the α locally, and the vehicle uses non-orthogonal multiple access to i T i and β i T i Send them to the roadside unit j and the aerial base station closest to the target vehicle respectively, and send subtask α i T i The transmission power is P i,r, Send subtask β i T i has a transmission power of P i,a , P i,r , P i,a ≥0, the total transmission power P i,r +P i,a shall not exceed the transmission power limit P max , according to the position of the current vehicle, the tasks generated by the target vehicle, and the computing resources, the computing resources that the aerial base station and roadside unit j can allocate to the target vehicle are determined by the binary search method to obtain α i , β i , θ i , P i,r , P i,a values.

[0067] This embodiment uses the binary search method to determine α i , β i , θ i , the transmission power P i,r , the transmission power P i,a method is as follows:

[0068] 1) Determine α i , β i , θ i :

[0069]

[0070] β i = 1 - α i - θ i

[0071] θ i = (H θ + L θ ) / 2

[0072] where H θ , L θ are respectively the upper limit and lower limit of the θ i value obtained after binary search.

[0073] 2) Determine the functional relationship P i,r (α i ) between the transmission power of the vehicle offloading the roadside unit subtask and the size of the roadside unit subtask, and the functional relationship P i,a (α i ) between the transmission power of the vehicle offloading the aerial base station subtask and the size of the roadside unit subtask:

[0074] P i,r (α i ) = (2 X-1) N0 / H i,1

[0075] P i,a (α i ) = (2 Y -1) ((2 X -1) N0 / H i,1 + N0 / H i,2 )

[0076]

[0077]

[0078] By adopting the binary search method, the optimal power allocation and sub-task size division strategy are found, making the local processing time, the offloading time of the roadside unit sub-task, and the offloading time of the aerial base station sub-task equal, achieving the goal of minimizing the processing task time and improving the efficiency of the vehicle to process tasks.

[0079] (3) Determine the transmission delay of the offloading sub-task

[0080] Determine the transmission delay of the roadside unit sub-task according to the following formula and the transmission delay of the aerial base station sub-task

[0081]

[0082]

[0083]

[0084]

[0085]

[0086]

[0087] where α i D i is the data volume of the roadside unit sub-task, β i D i is the data volume of the aerial base station sub-task, B is the total uplink bandwidth of the system, B ∈ [15, 25], the unit is MHz, and the value of B in this embodiment is 20, W is the number of sub-channels, W ∈ [12, 24], and the value of W in this embodiment is 18, H i,1 is the channel gain from the target vehicle to the connected roadside unit j, H i,2is the channel gain from the target vehicle to the aerial base station A, N0 is the power of Gaussian noise, N0 ∈ [-120, -100], the value of N0 in this embodiment is -110, with the unit of dbm, g i is the Rayleigh fading component, following a complex Gaussian distribution with a mean of 0 and a variance of 1, x i is the straight-line distance from the target vehicle to the connected roadside unit, η is the path loss factor, g0 represents the channel gain between the target vehicle and the aerial base station at unit distance, H is the altitude of the aerial base station, ‖y i ‖ is the horizontal projection distance between the target vehicle and the aerial base station;

[0088] (4) Determine the offloading roadside unit

[0089] The roadside unit subtask sent by the vehicle to roadside unit j is forwarded to roadside unit k through a wired link. Roadside unit k is the offloading roadside unit, and the offloading roadside unit k is determined by an approximate optimal method.

[0090] In this embodiment, the approximate optimal method for determining the offloading roadside unit k is as follows:

[0091] Initialize the offloading strategy to deliver the roadside unit subtasks α i T i of all vehicles to the connected roadside units for calculation. Traverse all vehicles. If there exists a roadside unit subtask α i T i of a certain vehicle, after forwarding it to other roadside units, the ratio ε of the total time delay reduction is achieved, ε ∈ [0.01, 0.1], and the value of ε in this embodiment is 0.05, then execute this forwarding, and repeat this operation until the total time delay no longer decreases.

[0092] Determining the offloading roadside unit by the approximate optimal method can balance the loads of each roadside unit compared with directly offloading the task to the roadside unit closest to the vehicle, further reduce the total system delay, and at the same time reduce the delay variance of each vehicle, ensuring the fairness of the system.

[0093] (5) Determine the calculation delay of the offloading subtask

[0094] Determine the calculation delay of the roadside unit subtask α i T i and the calculation delay of the aerial base station subtask β by the following formula i T i

[0095]

[0096] ​

[0097] f a,i = F A / N

[0098] where α i C i represents the number of CPU cycles required to execute the roadside unit subtask, and f k,i is the computing power resource allocated to the target vehicle by the roadside unit k, is the wired link transmission rate between the roadside unit k and the roadside unit j, and β i C i represents the number of CPU cycles required to execute the air base station subtask, and f a,i is the computing power resource allocated to the target vehicle by the air base station, and F A is the total computing power resource of the air base station.

[0099] (6) Determine the computing delay of the local subtask

[0100] Determine the computing delay of the local subtask according to the following formula

[0101]

[0102] where θ i C i is the number of CPU cycles required to compute the local subtask, and f i is the local computing resource.

[0103] (7) Determine the total processing delay

[0104] Determine the total processing delay T according to the following formula total :

[0105]

[0106] (8) Return the calculation result of the offloading subtask

[0107] Return the subtask calculation result to the target vehicle.

[0108] Complete the edge computing resource optimization method for the air base station-assisted intelligent transportation system.

[0109] Embodiment 2

[0110] The edge computing resource optimization method for the air base station-assisted intelligent transportation system in this embodiment consists of the following steps:

[0111] (1) Model the tasks generated by the target vehicle

[0112] Determine the task T generated by the modeled target vehicle according to the following formula i :

[0113] T i = {D i , C i}

[0114] Among them, D i is the data size of the task generated by vehicle i, i ∈ {1, 2, …, N}, N is the total number of vehicles, N ∈ [12, 24], and the value of N in this embodiment is 12. The value of D i is 400 - 600 Kbit, and the value of D i in this embodiment is 400 Kbit. C i is the number of CPU cycles required for the execution of this task. The value of C i is 320000 - 720000 cycle, and the value of C i in this embodiment is 320000 cycle.

[0115] (2) Determine the transmission power and the size of the offloading sub - task

[0116] The task generated by the target vehicle is split into three sub - tasks {α i T i , β i T i , θ i T i}. Among them, α i T i is the roadside unit sub - task, β i T i is the air - base station sub - task, θ i T i is the local sub - task. α i T i is offloaded to a certain roadside unit k, k ∈ {1, 2, …, R}, R is the total number of roadside units, R ∈ [3, 10], and the value of R in this embodiment is 3. β i T i is offloaded to an air - base station, and θ i T i is processed locally by the target vehicle. The vehicle uses non - orthogonal multiple access to send α i T i and β i T i to the roadside unit j and the air - base station closest to the target vehicle respectively. The transmission power for sending the sub - task α i T i is P i,r , and the transmission power for sending the sub - task β i T i is P i,a , P i,r , P i,a≥0, total transmit power P i,r +P i,a The transmit power limit P must not be exceeded max , according to the current vehicle position, the tasks and computing resources generated by the target vehicle, the aerial base station and the roadside unit j can allocate computing resources to the target vehicle, and the binary search method is used to determine α i ,β i ,θ i , P i,r ,P i,a The value of .

[0117] This embodiment uses a binary search method to determine α i , β i ,θ i , transmit power P i,r , transmit power P i,a The method is:

[0118] 1) Determine α using the following formula i , β i ,θ i :

[0119]

[0120] β i =1-α i -θ i

[0121] θ i =(H θ +L θ ) / 2

[0122] Among them, H θ , L θ are θ obtained after binary search i Upper and lower limits of values;

[0123] 2) Determine the functional relationship P between the transmission power of the vehicle unloading roadside unit subtask and the roadside unit subtask size according to the following formula: i,r (α i ) and the functional relationship between the transmission power of the vehicle unloading aerial base station subtask and the roadside unit subtask size P i,a (α i ):

[0124] P i,r (α i )=(2 X -1)N0 / H i,1

[0125] P i,a (α i )=(2Y -1)((2 X -1)N0 / H i,1 +N0 / H i,2 )

[0126]

[0127]

[0128] (3) Determine the transmission delay of the offloaded subtask

[0129] The transmission delay of the roadside unit subtask is determined as follows: and the transmission delay of the airborne base station subtask

[0130]

[0131]

[0132]

[0133]

[0134]

[0135]

[0136] Among them, α i D i is the data volume of the roadside unit subtask, β i D i is the data volume of the air base station subtask, B is the total uplink bandwidth of the system, B∈[15,25], in MHz, the value of B in this embodiment is 15, W is the number of subchannels, W∈[12,24], the value of W in this embodiment is 12, H i,1 is the channel gain from the target vehicle to the connected roadside unit j, H i,2 is the channel gain from the target vehicle to the aerial base station A, N0 is the power of Gaussian noise, N0∈[-120,-100], N0 in this embodiment is -120, unit is dBm, g i is the Rayleigh fading component, which obeys a complex Gaussian distribution with a mean of 0 and a variance of 1. i is the straight-line distance from the target vehicle to the connected roadside unit, η is the path loss factor, g0 represents the channel gain between the target vehicle and the aerial base station per unit distance, H is the altitude of the aerial base station, ‖y i ‖ is the horizontal projection distance between the target vehicle and the aerial base station.

[0137] (4) Determine the unloading roadside unit

[0138] The roadside unit subtask sent by the vehicle to roadside unit j is forwarded to roadside unit k through a wired link. Roadside unit k is the offloading roadside unit, and the offloading roadside unit k is determined by an approximate optimal method.

[0139] In this embodiment, the approximate optimal method is used to determine the offloading roadside unit k as follows:

[0140] Initialize the offloading policy so that the roadside unit subtask α of all vehicles i T i is delivered to the connected roadside unit for calculation. Traverse all vehicles. If there is a roadside unit subtask α of a certain vehicle i T i after being forwarded to other roadside units, the ratio ε of the total time delay reduction is achieved, where ε ∈ [0.01, 0.1]. In this embodiment, the value of ε is 0.01. Then execute this forwarding, and repeat this operation until the total time delay no longer decreases.

[0141] Other steps are the same as those in Embodiment 1.

[0142] Complete the edge computing resource optimization method for an air base station-assisted intelligent transportation system.

[0143] Embodiment 3

[0144] The edge computing resource optimization method for an air base station-assisted intelligent transportation system in this embodiment consists of the following steps:

[0145] (1) Model the tasks generated by the target vehicle <##

[0146] Determine the task T generated by the target vehicle according to the following formula i :

[0147] T i = {D i , C i}

[0148] where D i is the data size of the task generated by vehicle i, i ∈ {1, 2,..., N}, N is the total number of vehicles, N ∈ [12, 24]. In this embodiment, the value of N is 24, and the value of D i is 400 - 600 Kbit. In this embodiment, the value of D i is 600 Kbit, and C i is the number of CPU cycles required for the execution of this task. The value of C i is 320000 - 720000 cycle. In this embodiment, the value of C i is 720000 cycle.

[0149] (2) Determine the transmission power and the size of the offloading subtasks

[0150] The task generated by the target vehicle is split into three subtasks {α i T i , β i T i , θ i T i}, where α i T i is the roadside unit subtask, β i T i is the aerial base station subtask, θ i T i is the local subtask. α i T i is offloaded to a certain roadside unit k, k ∈ {1, 2, …, R}, R is the total number of roadside units, R ∈ [3, 10], and in this embodiment, R is taken as 10. β i T i is offloaded to an aerial base station, and θ i T i is processed locally by the target vehicle. The vehicle uses non-orthogonal multiple access to send α i T i and β i T i to the roadside unit j and the aerial base station closest to the target vehicle respectively. The transmission power for sending subtask α i T i is P i,r , and the transmission power for sending subtask β i T i is P i,a , P i,r , P i,a ≥ 0. The total transmission power P i,r + P i,a shall not exceed the transmission power limit P max . According to the current vehicle's position, the task generated by the target vehicle, and the computing resources, the computing resources that the aerial base station and roadside unit j can allocate to the target vehicle are used to determine the values of α i , β i , θ i , P i,r , P i,a using the binary search method.

[0151] The method for determining α i , β i , θ i , the transmission power P i,r , and the transmission power P i,a in this embodiment using the binary search method is as follows:

[0152] 1) Determine α according to the following formula i , β i , θ i :

[0153]

[0154] β i = 1 - α i - θ i

[0155] θ i = (H θ + L θ ) / 2

[0156] where H θ , L θ are respectively the upper limit and lower limit of the value of θ obtained after binary search. i

[0157] 2) Determine the functional relationship P i,r (α i ) between the transmission power of the vehicle offloading roadside unit subtask and the size of the roadside unit subtask, and the functional relationship P i,a (α i ) between the transmission power of the vehicle offloading air base station subtask and the size of the roadside unit subtask:

[0158] P i,r (α i ) = (2 X - 1)N0 / H i,1

[0159] P i,a (α i ) = (2 Y - 1)((2 X - 1)N0 / H i,1 + N0 / H i,2 )

[0160]

[0161]

[0162] (3) Determine the transmission delay of the offloading subtask

[0163] Determine the transmission delay of the roadside unit subtask according to the following formula and the transmission delay of the air base station subtask

[0164]

[0165] ​

[0166]

[0167]

[0168]

[0169]

[0170] where α i D i is the data volume of the roadside unit subtask, β i D i is the data volume of the aerial base station subtask, B is the total uplink bandwidth of the system, B ∈ [15, 25], with the unit of MHz. In this embodiment, the value of B is 25, W is the number of sub-channels, W ∈ [12, 24]. In this embodiment, the value of W is 24, H i,1 is the channel gain from the target vehicle to the connected roadside unit j, H i,2 is the channel gain from the target vehicle to the aerial base station A, N0 is the power of Gaussian noise, N0 ∈ [-120, -100]. In this embodiment, the value of N0 is -100, with the unit of dbm, g i is the Rayleigh fading component, which follows a complex Gaussian distribution with a mean of 0 and a variance of 1, x i is the straight-line distance from the target vehicle to the connected roadside unit, η is the path loss factor, g0 represents the channel gain between the target vehicle and the aerial base station at unit distance, H is the altitude of the aerial base station, ‖y i ‖ is the horizontal projection distance between the target vehicle and the aerial base station.

[0171] (4) Determine the offloading roadside unit

[0172] The roadside unit subtask sent by the vehicle to roadside unit j is forwarded to roadside unit k through a wired link. Roadside unit k is the offloading roadside unit, and the offloading roadside unit k is determined by an approximate optimal method.

[0173] In this embodiment, the approximate optimal method is used to determine the offloading roadside unit k as follows:

[0174] Initialize the offloading strategy so that the roadside unit subtasks α i T i of all vehicles are delivered to the connected roadside units for calculation. Traverse all vehicles. If there exists a roadside unit subtask α i T i of a certain vehicle can be forwarded to other roadside units to reduce the total time delay reduction ratio ε, ε ∈ [0.01, 0.1]. In this embodiment, the value of ε is 0.1, then execute this forwarding, and repeat this operation until the total time delay no longer decreases.

[0175] Other steps are the same as those in Embodiment 1.

[0176] The edge computing resource optimization method for an air-base-station-assisted intelligent transportation system is completed.

[0177] To verify the beneficial effects of the present invention, the inventor conducted comparative simulation experiments using the edge computing resource optimization method for an air-base-station-assisted intelligent transportation system in Embodiment 1 of the present invention, the method of randomly selecting offloading roadside units and optimal power and subtask allocation (Comparative Experiment 1), and the method of heuristically selecting offloading roadside units and weight-allocated power and subtasks (Comparative Experiment 2). The various experimental situations are as follows:

[0178] The evaluation indexes for the offloading situation are the average delay and the standard deviation of the delay. Among them, the average delay is the total processing delay T total divided by the number of vehicles N, and the standard deviation of the delay is the standard deviation of the processing time of each vehicle The experimental results of the offloading situation of the target vehicle are shown in Table 1.

[0179] Table 1 Comparative experimental results of offloading situation

[0180] Experimental content Average latency (ms) Standard deviation of latency (ms) Comparative experiment 1 53.3086 7.1321 Comparative experiment 2 57.1931 7.7169 The present invention 47.2349 4.5567

[0181] As can be seen from Table 1, the average delay of the present invention is lower than that of Comparative Experiment 1 and Comparative Experiment 2, and the standard deviation of the delay is lower than that of Comparative Experiment 1 and Comparative Experiment 2.

Claims

1. An edge computing resource optimization method for an intelligent transportation system assisted by an aerial base station, characterized in that It consists of the following steps: (1) Modeling the tasks generated by the target vehicle Determine the tasks generated by the target vehicle to be modeled according to the following formula: T i = {D i , C i} Among them, D i is the data size of the task generated by vehicle i, where i ∈ {1, 2,..., N}, N is the total number of vehicles, and N is a finite positive integer. C i is the number of CPU cycles required for the execution of this task; (2) Determine the transmission power and the size of the offloading subtasks The tasks generated by the target vehicle are split into three subtasks {α i T i ,β i T i ,θ ii T}, where α i T i is a roadside unit subtask, β i T i is an air base station subtask, θ i T i is a local subtask, α i T i is offloaded to a certain roadside unit k, k ∈ {1, 2,..., R}, R is the total number of roadside units, R ∈ [3, 10], β i T i is offloaded to an air base station, θ i T i is processed locally by the target vehicle. The vehicle uses non-orthogonal multiple access to send α i T i and β i T i to the roadside unit j and the air base station closest to the target vehicle respectively. The transmission power for sending subtask α i T i is P i,r , and the transmission power for sending subtask β i T i is P i,a , P i,r , P i,a ≥0. The total transmission power P i,r +P i,a shall not exceed the transmission power limit P max . According to the current vehicle position, the tasks generated by the target vehicle, and the computing resources, the computing resources that the air base station and roadside unit j can allocate to the target vehicle are determined by using the binary search method to determine the values of α i ,β i ,θ i , P i,r , P i,a ; (3) Determine the transmission delay of the offloading subtasks Determine the transmission delay of the roadside unit subtask according to the following formula and the transmission delay of the aerial base station subtask Among them, α i D i is the data volume of the roadside unit subtask, β i D i is the data volume of the aerial base station subtask, B is the total uplink bandwidth of the system, B ∈ [15, 25], the unit is MHz, W is the number of sub-channels, W ∈ [12, 24], H i,1 is the channel gain from the target vehicle to the connected roadside unit j, H i,2 is the channel gain from the target vehicle to the aerial base station A, N0 is the power of Gaussian noise, N0 ∈ [-120, -100], the unit is dBm, g i is the Rayleigh fading component, which follows a complex Gaussian distribution with a mean of 0 and a variance of 1, x i is the straight-line distance from the target vehicle to the connected roadside unit, η is the path loss factor, g0 represents the channel gain between the target vehicle and the aerial base station when the reference distance is 1m, H is the altitude of the aerial base station, ‖y i ‖ is the horizontal projection distance from the target vehicle to the aerial base station; (4) Determine the offloading roadside unit The roadside unit subtasks sent by the vehicle to roadside unit j are forwarded to roadside unit k through a wired link. Roadside unit k is the offloading roadside unit, and the approximate optimal method is used to determine the offloading roadside unit k; (5) Determine the computing delay of the offloading subtasks Determine the roadside unit subtask α according to the following formula i T i Computation delay of And the air base station subtask β i T i Computation delay of f a,i = F A / N where α i C i represents the number of CPU cycles required to execute the roadside unit subtasks, f k,i is the computing power resource allocated to the target vehicle by the roadside unit k, is the wired link transmission rate between the roadside unit k and the roadside unit j, β i C i represents the number of CPU cycles required to execute the air base station subtasks, f a,i is the computing power resource allocated to the target vehicle by the air base station, F A is the total computing power resource of the air base station; (6) Determine the computing delay of the local subtasks Determine the calculation delay of the local subtask according to the following formula where θ i C i is the number of CPU cycles required for calculating the local subtask, and f i is the local computing resource; (7) Determine the total processing delay Determine the total processing delay T according to the following formula total :[[]]END]] (8) Return the computing results of the offloading subtasks The computing results of the subtasks are returned to the target vehicle; thus completing the edge computing resource optimization method for an air base station-assisted intelligent transportation system.

2. The edge computing resource optimization method for an intelligent transportation system assisted by an aerial base station according to claim 1, characterized in that: In the step of (2) determining the transmission power and the size of the offloading sub-task, the method of using binary search to determine α i , β i , θ i , the transmission power P i,r , the transmission power P i,a is as follows: 1) Determine α according to the following formula i , β i , θ i : β i = 1 - α i -θ i θ i =(H θ +L θ ) / 2 Among them, H θ , L θ are respectively the upper limit and the lower limit of the value of θ i obtained after binary search; 2) Determine the functional relationship between the transmission power and the size of the roadside unit subtask for vehicle offloading according to the following formula P i,r (α i ) and the functional relationship between the transmission power and the size of the roadside unit subtask for vehicle offloading to the aerial base station P i,a (α i ): P i,r (α i ) = (2 X - 1)N0 / H i,1 P i,a (α i )=(2 Y -1)((2 X -1)N0 / H i,1 +N0 / H i,2 ) 3. The edge computing resource optimization method for an air-base station-assisted intelligent transportation system according to claim 1, wherein: In step (4), the approximate optimal method for determining the offloading roadside unit k is as follows: Initialize the offloading policy to make the roadside unit subtask α of all vehicles i T i Deliver it to the connected roadside unit for calculation. Traverse all vehicles. If there is a roadside unit subtask α of a certain vehicle i T i After forwarding it to other roadside units, if the ratio ε of the total time delay reduction is such that ε ∈ [0.01, 0.1], then perform this forwarding, and repeat this operation until the total time delay no longer decreases.

4. The edge computing resource optimization method for an intelligent transportation system assisted by an aerial base station according to claim 1, wherein In step (1), the tasks generated by the modeled target vehicle are as follows: Determine the task T generated by the modeled target vehicle according to the following formula i : T i = {D i , C i} Among them, D i is the data size of the task generated by vehicle i, where i ∈ {1, 2,..., N}, N is the total number of vehicles, N ∈ [12, 24], and D i ranges from 400 to 600 Kbits. C i is the number of CPU cycles required for the execution of this task. C i ranges from 320000 to 720000 cycle / s.

Images