CN111918245A - Multi-agent-based vehicle speed perception calculation task unloading and resource allocation method - Google Patents

Multi-agent-based vehicle speed perception calculation task unloading and resource allocation method Download PDF

Info

Publication number
CN111918245A
CN111918245A CN202010648403.2A CN202010648403A CN111918245A CN 111918245 A CN111918245 A CN 111918245A CN 202010648403 A CN202010648403 A CN 202010648403A CN 111918245 A CN111918245 A CN 111918245A
Authority
CN
China
Prior art keywords
vehicle
task
tasks
agent
vec
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202010648403.2A
Other languages
Chinese (zh)
Other versions
CN111918245B (en
Inventor
贺丽君
黄鑫宇
李凡
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian Jiaotong University
Original Assignee
Xian Jiaotong University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xian Jiaotong University filed Critical Xian Jiaotong University
Priority to CN202010648403.2A priority Critical patent/CN111918245B/en
Publication of CN111918245A publication Critical patent/CN111918245A/en
Application granted granted Critical
Publication of CN111918245B publication Critical patent/CN111918245B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
    • H04W4/30—Services specially adapted for particular environments, situations or purposes
    • H04W4/40—Services specially adapted for particular environments, situations or purposes for vehicles, e.g. vehicle-to-pedestrians [V2P]
    • H04W4/48—Services specially adapted for particular environments, situations or purposes for vehicles, e.g. vehicle-to-pedestrians [V2P] for in-vehicle communication
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00—Network arrangements or protocols for supporting network services or applications
    • H04L67/01—Protocols
    • H04L67/10—Protocols in which an application is distributed across nodes in the network
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00—Network arrangements or protocols for supporting network services or applications
    • H04L67/01—Protocols
    • H04L67/12—Protocols specially adapted for proprietary or special-purpose networking environments, e.g. medical networks, sensor networks, networks in vehicles or remote metering networks
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W72/00—Local resource management
    • H04W72/50—Allocation or scheduling criteria for wireless resources
    • H04W72/53—Allocation or scheduling criteria for wireless resources based on regulatory allocation policies

Abstract

The invention discloses a multi-agent-based vehicle speed perception calculation task unloading and resource allocation method, which comprises the following steps of: collecting vehicle-end computing tasks, and dividing the vehicle-end computing tasks into key tasks, high-priority tasks and low-priority tasks according to the types of the vehicle-end computing tasks; respectively calculating different computing resources and wireless resource allocation, unloading to a VEC server, executing locally and continuously waiting for corresponding time delay and energy consumption, and then forming a target function for reducing the energy consumption of the processing task at the vehicle end under the time delay threshold constraint of each task; converting the objective function into a Markov decision process; training a multi-agent reinforcement learning network; and inputting the states of the vehicle end to be distributed and the edge server into the trained multi-agent reinforcement learning network to obtain the task unloading and resource distribution results.

Description

Multi-agent-based vehicle speed perception calculation task unloading and resource allocation method
Technical Field
The invention belongs to the technical field of wireless communication, and relates to a multi-agent vehicle speed perception-based calculation task unloading and resource allocation method.
Background
With the rapid development of the internet of things, intelligent vehicle-mounted applications (including automatic driving, image-assisted navigation and multimedia entertainment) are widely applied to intelligent automobiles, and can provide more comfortable and safe environments for drivers and passengers. However, these vehicle applications require a large amount of computing resources and require extremely low processing time, and a cloud server with powerful computing and storage capabilities may be used to handle the off-load computing task of the vehicle, but since long-distance transmission may cause high latency, Vehicle Edge Computing (VEC) should be developed in order to cope with the disadvantages of the cloud server.
The VEC server is closer to a vehicle terminal, has strong computing power and is densely deployed beside a roadside unit (RSU), and processing time delay and energy consumption of vehicle-mounted application can be remarkably reduced by unloading computing consumption type tasks to the VEC server.
The development of VECs also faces many challenges. For example, the influence of vehicle speed on the delay threshold of a computing task in the VEC network, the interaction of task processing delay and energy consumption with task unloading and resource allocation strategies. Therefore, the research on the task delay threshold of vehicle speed perception and the problem of the mutual influence of the task processing delay and energy consumption and the task unloading and resource allocation strategy is very important for the overall performance of the VEC.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a computing task unloading and resource allocation method based on multi-agent vehicle speed perception, and the method can effectively improve the overall performance of a vehicle in a VEC server scene.
In order to achieve the purpose, the computing task unloading and resource allocation method based on multi-agent vehicle speed perception comprises the following steps:
1) collecting vehicle end calculation tasks, dividing the vehicle end calculation tasks into key tasks, high-priority tasks and low-priority tasks according to the types of the vehicle end calculation tasks, and determining time delay thresholds of the key tasks, the high-priority tasks and the low-priority tasks according to the vehicle speed;
2) for the key tasks, the high-priority tasks and the low-priority tasks, respectively calculating different calculation resources and wireless resource allocation, unloading to a VEC server, executing locally and continuously waiting for corresponding time delay and energy consumption, and then forming a target function for reducing the energy consumption of the processing tasks at the vehicle end under the constraint of the time delay threshold of each task;
3) converting the objective function obtained in the step 2) into a Markov decision process, and initializing a state space, an action space and reward of the Markov decision process;
4) obtaining new states, actions and rewards according to the multi-agent reinforcement learning network, and storing the new states, actions and rewards into an experience playback pool;
5) when the data in the experience playback pool reaches a threshold value, training the multi-agent reinforcement learning network until the multi-agent reinforcement learning network converges;
6) and inputting the states of the vehicle end and the edge server to be distributed into the trained multi-agent reinforcement learning network to obtain task unloading and resource distribution results, and finishing calculation task unloading and resource distribution based on the vehicle speed perception of multi-agent reinforcement learning.
According to the bandwidth and time delay requirements of the vehicle end calculation task, the vehicle end calculation task is divided into a key task phi1High priority task phi2And low priority task phi3Critical task phi1High priority task phi2And low priority task phi3The corresponding time delay thresholds are Thr respectively1、Thr2And Thr3Wherein the critical task phi1With a delay threshold of 10ms, a low priority task phi3With a delay threshold of 100ms, a high priority task phi2In relation to the current driving speed of the vehicle, wherein the mission
Figure BDA0002573996410000021
Delay threshold of
Figure BDA0002573996410000022
Comprises the following steps:
Figure BDA0002573996410000031
wherein when
Figure BDA0002573996410000032
When the temperature of the water is higher than the set temperature,
Figure BDA0002573996410000033
when in use
Figure BDA0002573996410000034
When the temperature of the water is higher than the set temperature,
Figure BDA0002573996410000035
Figure BDA0002573996410000036
speed, Thr, of vehicle k2Limiting speed v for roadmaxAnd (4) corresponding delay threshold.
Transmission rate of uplink channel n assigned to vehicle k by VEC server
Figure BDA0002573996410000037
Comprises the following steps:
Figure BDA0002573996410000038
wherein σ2Is the noise power, P is the transmission power,
Figure BDA0002573996410000039
for uplink channel interference, channel bandwidth
Figure BDA00025739964100000310
Figure BDA00025739964100000311
For the total upstream bandwidth of the VEC server,
Figure BDA00025739964100000312
for the number of upstream channels of the VEC server,
Figure BDA00025739964100000313
is an uplink channel set;
is provided with
Figure BDA00025739964100000314
Indicating whether the uplink channel n between the vehicle k and the VEC server is allocated to the vehicle k, if so, then
Figure BDA00025739964100000315
Is 1, otherwise, then
Figure BDA00025739964100000316
Is 0, the uplink transmission rate between the vehicle k and the VEC server is obtained
Figure BDA00025739964100000317
Comprises the following steps:
Figure BDA00025739964100000318
transmission rate of downlink channel n allocated to vehicle k by VEC server
Figure BDA00025739964100000319
Comprises the following steps:
Figure BDA00025739964100000320
wherein σ2Is the noise power, P is the transmission power,
Figure BDA00025739964100000321
for downlink channel interference, channel bandwidth
Figure BDA00025739964100000322
Figure BDA00025739964100000323
Downstream for VEC serversThe total bandwidth of the bandwidth is,
Figure BDA00025739964100000324
the number of the downlink channels of the VEC server,
Figure BDA00025739964100000325
is a downlink channel set;
is provided with
Figure BDA00025739964100000326
Indicating whether a downlink channel n between the vehicle k and the VEC server is allocated to the vehicle k, if so, then
Figure BDA00025739964100000327
Is 1, otherwise, then
Figure BDA00025739964100000328
Is 0, the downlink transmission rate between the vehicle k and the VEC server is obtained
Figure BDA00025739964100000329
Comprises the following steps:
Figure BDA0002573996410000041
task of vehicle k
Figure BDA0002573996410000042
Total latency offloaded to VEC server execution consumption
Figure BDA0002573996410000043
Comprises the following steps:
Figure BDA0002573996410000044
wherein the content of the first and second substances,
Figure BDA0002573996410000045
in order to be a function of rounding up,
Figure BDA0002573996410000046
for the task of vehicle k
Figure BDA0002573996410000047
The size of the file of (a) is,
Figure BDA0002573996410000048
for processing the tasks of the vehicle k
Figure BDA0002573996410000049
The required density of the calculations to be made,
Figure BDA00025739964100000410
task downloaded for vehicle k
Figure BDA00025739964100000411
The file size of the file is reduced relative to the original uploading task,
Figure BDA00025739964100000412
tasking VEC servers to vehicle k
Figure BDA00025739964100000413
Proportion of allocated computing resources, fVECIs the CPU frequency of the local VEC server,
Figure BDA00025739964100000414
for the upstream transmission rate between the vehicle and the VEC server,
Figure BDA00025739964100000415
is the downlink transmission rate.
CPU frequency f assigned according to vehicle kkTask for calculating vehicle k
Figure BDA00025739964100000416
Time consumption of local execution
Figure BDA00025739964100000417
Comprises the following steps:
Figure BDA00025739964100000418
at time t, the task of vehicle k
Figure BDA00025739964100000419
Can choose to continue waiting, offloading to local VEC server and execute locally
Figure BDA00025739964100000420
Representing the mission of a vehicle k
Figure BDA00025739964100000421
Whether to continue waiting or not, and when continuing waiting, then
Figure BDA00025739964100000422
Is 1, otherwise, then
Figure BDA00025739964100000423
Set the continuous waiting time to be Th，
Figure BDA00025739964100000424
Representing the mission of a vehicle k
Figure BDA00025739964100000425
Whether to offload to a local VEC server when
Figure BDA00025739964100000426
If 1, the task of the vehicle k is represented
Figure BDA00025739964100000427
Off-load to local VEC Server, mission for vehicle k
Figure BDA00025739964100000428
From production tasks
Figure BDA00025739964100000429
Total delay to completion of execution of an action
Figure BDA00025739964100000430
Comprises the following steps:
Figure BDA00025739964100000431
wherein the content of the first and second substances,
Figure BDA00025739964100000432
for tasks according to vehicle k
Figure BDA00025739964100000433
The time of generation.
Task of calculation when vehicle k
Figure BDA0002573996410000051
When unloading to the VEC server, the energy consumption of the unloading task comprises the energy consumed by the uploading calculation task and the energy consumed by the downloading task, and the energy consumption of the unloading to the VEC server
Figure BDA0002573996410000052
Comprises the following steps:
Figure BDA0002573996410000053
task of calculation when vehicle k
Figure BDA0002573996410000054
Upon local processing, depending on the task of processing vehicle k
Figure BDA0002573996410000055
Required energy density
Figure BDA0002573996410000056
Energy consumption resulting from local processing of tasks
Figure BDA0002573996410000057
Comprises the following steps:
Figure BDA0002573996410000058
at time t, after the vehicle executes the offload policy, the energy consumed by all vehicles within the service range of the local VEC server, e (t), is:
Figure BDA0002573996410000059
under the condition that the task delay threshold and the computing resources and the wireless resources are limited, the formed objective function for reducing the energy consumption of the processing task at the vehicle end is as follows:
Figure BDA00025739964100000510
wherein the content of the first and second substances,
Figure BDA00025739964100000511
the invention has the following beneficial effects:
the invention relates to a multi-agent-based vehicle speed perception computing task unloading and resource allocation method, which comprises the steps of firstly collecting vehicle end computing tasks, dividing the tasks into key tasks, high-priority tasks and low-priority tasks according to the types of the vehicle end computing tasks, obtaining time delay thresholds of different computing tasks by combining vehicle speed, respectively computing different computing resources and wireless resource allocation for different computing tasks, unloading to a VEC server, executing locally and continuously waiting for corresponding time delay and energy consumption, forming a target function for reducing the energy consumption of the vehicle end under the constraint of the time delay thresholds of the tasks, thereby comprehensively considering the factors of the position, the speed, the task queue, the computing resources, the wireless resources, the computing resources and the wireless resources of the VEC server end, and the like, and effectively reducing the energy consumption of the vehicle end in the time delay threshold of task processing, and finally, carrying out calculation task unloading and resource allocation by using the trained multi-agent reinforcement learning network so as to improve the overall performance of the vehicle under the scene of the VEC server.
Drawings
FIG. 1 is a schematic flow diagram of the present invention;
FIG. 2 is a distribution diagram of vehicle average task completion time delay (the number of vehicles is 7) corresponding to five algorithms;
FIG. 3 is a distribution diagram of vehicle average task completion time delay (the number of vehicles is 9) corresponding to five algorithms;
FIG. 4 is a distribution diagram of vehicle average task completion time delay (the number of vehicles is 11) corresponding to five algorithms;
FIG. 5 is a distribution diagram of vehicle average task completion time delay (the number of vehicles is 13) corresponding to five algorithms;
FIG. 6 is a distribution diagram of the average task energy consumption (number of vehicles is 7) of the vehicles corresponding to the five algorithms;
FIG. 7 is a distribution diagram of the average mission energy consumption (number of vehicles is 9) of the vehicles corresponding to the five algorithms;
FIG. 8 is a distribution diagram of the average mission energy consumption (number of vehicles 11) of the vehicles corresponding to the five algorithms;
FIG. 9 is a distribution diagram of the average mission energy consumption (13 vehicles) of the vehicles corresponding to the five algorithms;
FIG. 10 is a distribution diagram of vehicle average task completion time delay (vehicle speed range is 30-50Km/h) corresponding to five algorithms;
FIG. 11 is a distribution diagram of the average task energy consumption (vehicle speed range is 30-50Km/h) of the vehicle corresponding to the five algorithms;
FIG. 12 is a distribution diagram of vehicle average task completion time delay (vehicle speed range is 50-80Km/h) corresponding to five algorithms;
FIG. 13 is a distribution diagram of the average task energy consumption (vehicle speed range 50-80Km/h) of the vehicle corresponding to the five algorithms.
Detailed Description
The invention is described in further detail below with reference to the accompanying drawings:
assume that the set of vehicles within the service scope of the local VEC server is set to
Figure BDA0002573996410000071
The number of vehicles is K, and at the time t, the calculation task to be processed by the vehicle K is
Figure BDA0002573996410000072
The time when the task is generated is
Figure BDA0002573996410000073
The time delay consumed by the task processing in the VEC server is
Figure BDA0002573996410000074
The time delay consumed by the local execution of the task is
Figure BDA0002573996410000075
Then for the task
Figure BDA0002573996410000076
Total time delay from generation to completion of execution of action
Figure BDA0002573996410000077
Comprises the following steps:
Figure BDA0002573996410000078
wherein the content of the first and second substances,
Figure BDA0002573996410000079
representing the mission of a vehicle k
Figure BDA00025739964100000710
Whether to continue waiting or not, and if so, whether to continue waiting or not
Figure BDA00025739964100000711
Is 1, otherwise, then
Figure BDA00025739964100000712
Is 0, the continuation wait time is set to Th，
Figure BDA00025739964100000713
Representing the mission of a vehicle k
Figure BDA00025739964100000714
Whether to offload to a local VEC server or not,
Figure BDA00025739964100000715
1, then the task of vehicle k
Figure BDA00025739964100000716
Off-load to the local VEC server.
At time t, the energy consumed by the task processing at the VEC server is
Figure BDA00025739964100000717
The energy consumed by the local execution of the task is
Figure BDA0002573996410000081
After the vehicle executes the unloading strategy, the energy E (t) consumed by all vehicles within the service range of the local VEC server is as follows:
Figure BDA0002573996410000082
the invention takes the minimization of the energy consumption of a vehicle end as an optimization target, and under the condition of limited task delay threshold, calculation resources and wireless resources, the corresponding optimization problems are as follows:
Figure BDA0002573996410000083
s.t.
Figure BDA0002573996410000084
Figure BDA0002573996410000085
Figure BDA0002573996410000086
Figure BDA0002573996410000087
Figure BDA0002573996410000088
the invention relates to a multi-agent-based vehicle speed perception calculation task unloading and resource allocation method, which comprises the following steps of:
1) collecting vehicle end calculation tasks, dividing the vehicle end calculation tasks into key tasks, high-priority tasks and low-priority tasks according to the types of the vehicle end calculation tasks, and determining time delay thresholds of the key tasks, the high-priority tasks and the low-priority tasks according to the vehicle speed;
2) for the key tasks, the high-priority tasks and the low-priority tasks, respectively calculating different calculation resources and wireless resource allocation, unloading to a VEC server, executing locally and continuously waiting for corresponding time delay and energy consumption, and then forming a target function for reducing the energy consumption of the processing tasks at the vehicle end under the constraint of the time delay threshold of each task;
3) converting the objective function obtained in the step 2) into a Markov decision process, and initializing a state space, an action space and reward of the Markov decision process;
4) obtaining new states, actions and rewards according to the multi-agent reinforcement learning network, and storing the new states, actions and rewards into an experience playback pool;
5) when the data in the experience playback pool reaches a threshold value, training the multi-agent reinforcement learning network until the multi-agent reinforcement learning network converges;
6) and inputting the states of the vehicle end and the edge server to be distributed into the converged multi-agent reinforcement learning network to obtain task unloading and resource distribution results, and completing calculation task unloading and resource distribution based on the vehicle speed perception of multi-agent reinforcement learning.
The following is described in detail with reference to fig. 1:
step 11) dividing the tasks into key tasks, high-priority tasks and low-priority tasks according to the types of the tasks, and obtaining time delay thresholds of different calculation tasks by combining the vehicle speed, wherein the specific process is as follows;
according to the bandwidth and time delay requirements of the vehicle end calculation task, the vehicle end calculation task is divided into a key task phi1High priority task phi2And low priority task phi3Critical task phi1High priority task phi2And low priority task phi3The corresponding time delay thresholds are Thr respectively1、Thr2And Thr3Wherein the critical task phi1With a delay threshold of 10ms, a low priority task phi3With a delay threshold of 100ms, a high priority task phi2Related to the current running speed of the vehicle to obtain a task
Figure BDA0002573996410000091
Delay threshold of
Figure BDA0002573996410000092
Comprises the following steps:
Figure BDA0002573996410000093
wherein when
Figure BDA0002573996410000094
When the temperature of the water is higher than the set temperature,
Figure BDA0002573996410000095
when in use
Figure BDA0002573996410000096
When the temperature of the water is higher than the set temperature,
Figure BDA0002573996410000097
Figure BDA0002573996410000098
speed, Thr, of vehicle k2Limiting speed v for roadmaxAnd (4) corresponding delay threshold.
Step 12) for different computing tasks, when the tasks are unloaded to the VEC server for processing, consumed time delay
Figure BDA0002573996410000099
Comprises the following steps:
Figure BDA00025739964100000910
wherein the content of the first and second substances,
Figure BDA0002573996410000101
in order to be a function of rounding up,
Figure BDA0002573996410000102
for the task of vehicle k
Figure BDA0002573996410000103
The size of the file of (a) is,
Figure BDA0002573996410000104
for processing the tasks of the vehicle k
Figure BDA0002573996410000105
The required density of the calculations to be made,
Figure BDA0002573996410000106
task downloaded for vehicle k
Figure BDA0002573996410000107
The file size is reduced relative to the original uploading taskThe small proportion of the amount of the liquid,
Figure BDA0002573996410000108
tasking VEC servers to vehicle k
Figure BDA0002573996410000109
Proportion of allocated computing resources, fVECIs the CPU frequency of the local VEC server,
Figure BDA00025739964100001010
for the upstream transmission rate between the vehicle and the VEC server,
Figure BDA00025739964100001011
is the downlink transmission rate;
at this time, the corresponding vehicle-side energy consumption
Figure BDA00025739964100001012
Comprises the following steps:
Figure BDA00025739964100001013
wherein, P is the signal transmission power of the vehicle;
when the task is executed locally, the CPU frequency f is allocated according to the vehicle kkTo obtain the task of the vehicle k
Figure BDA00025739964100001014
Time consumption of local execution
Figure BDA00025739964100001015
Comprises the following steps:
Figure BDA00025739964100001016
at this time, the corresponding vehicle-side energy consumption
Figure BDA00025739964100001017
Comprises the following steps:
Figure BDA00025739964100001018
wherein the task of the vehicle k is processed
Figure BDA00025739964100001019
Required energy density
Figure BDA00025739964100001020
With the minimized energy consumption of the vehicle end as an optimization target, under the condition of limited task delay threshold, calculation resources and wireless resources, the corresponding optimization problem is as follows:
Figure BDA0002573996410000111
s.t.
Figure BDA0002573996410000112
Figure BDA0002573996410000113
Figure BDA0002573996410000114
Figure BDA0002573996410000115
Figure BDA0002573996410000116
calculating time delay and energy consumption of different unloading positions and resource allocation, and forming an optimization target for reducing energy consumption of a vehicle end under the time delay constraint;
step 13) initializing the state, actions and rewards of the Markov decision process, defining the state space of the vehicle k as s at the time tk(t), the state space of vehicle k includes state information of other vehicles and state information of VEC server, sk(t) is:
Figure BDA0002573996410000117
wherein v isk(t),dk(t),ck(t) represents the speed and position of the vehicle k at time t and the file size to be processed, rbVEC(t) represents the current remaining computing capacity of the VEC server at the time,
Figure BDA0002573996410000118
indicates whether the vehicle k selects the unloading position (-) at the time t, and if so, whether the unloading position (-) is selected
Figure BDA0002573996410000119
Is 1, otherwise, then
Figure BDA00025739964100001110
Is a non-volatile organic compound (I) with a value of 0,
Figure BDA00025739964100001111
indicating the proportion of computing resources that the VEC server allocates to vehicle k at time t,
Figure BDA00025739964100001112
indicating whether the upstream channel resources of the VEC server are idle at time t,
Figure BDA00025739964100001113
at time t, it is indicated whether the downlink channel resources of the VEC server are idle, and therefore, the state space of the system is defined as: st＝(s1(t),...sk(t)...,sK(t))；
For the vehicle k, the motion space is whether to continue waiting or not, whether to unload to the VEC server or not, the computing power allocated by the VEC server, and the uplink and downlink sub-channels allocated by the VEC server, that is:
Figure BDA0002573996410000121
therefore, at time t, the vehicle motion space is: a. thet＝{a1(t),...ak(t)...,aK(t)}；
Action a taken when vehicle kkWhen the state after (t) does not satisfy the conditions (c1) - (c7), the reward function is:
Figure BDA0002573996410000122
wherein, when the condition of (-) is not satisfied, Λ(·)Is-1, otherwise, Λ(·)The values are 0, l1,1,2,3 and 4 are experimental parameters.
When the post-action state of the vehicle k satisfies all of the conditions (c1) - (c4), the reward function is defined as:
rk(t)＝l2+exp(Thrk(t)-Dk(t))
wherein l2For experimental parameters, exp (-) is an exponential function, and when the post-action state of vehicle k satisfies all conditions (c1) - (c5), the reward function is:
r(t)＝l3+5·exp(Ek(t))
wherein l3,5Are experimental parameters.
Step 14) obtaining new states, actions and rewards according to the multi-agent reinforcement learning network and storing the new states, actions and rewards in an experience playback pool;
step 15) judging whether the data in the experience playback pool reaches a threshold value, and when the data in the experience playback pool reaches the threshold value, training the multi-agent reinforcement learning network until the multi-agent reinforcement learning network converges;
for the process of centralized training, the system consists of K agents, and the multi-agent reinforcementThe parameter of the learning network is theta ═ theta1,...,θKInstruction of
Figure BDA0002573996410000123
Representing the set of policies for all agents, the deterministic policy μ for agent kkThe gradient is expressed as:
Figure BDA0002573996410000131
wherein the content of the first and second substances,
Figure BDA0002573996410000132
the experience playback zone is composed of a series of states, actions and rewards, namely: (S, A, S', R),
Figure BDA0002573996410000133
the evaluation method is a centralized action-value function, the actions and some state information of all agents are input, the Q value of agent k is output, and the evaluation method is used for evaluating the network
Figure BDA0002573996410000134
Updating according to the Loss function, namely:
Figure BDA0002573996410000135
Figure BDA0002573996410000136
where γ is the discount factor and the action network is updated by minimizing the policy gradient of agent, i.e.:
Figure BDA0002573996410000137
wherein X is the size of the mini-batch, and j is the index of the sample.
And step 17) inputting the states of the vehicle end and the edge server to be distributed into the converged multi-agent reinforcement learning network to obtain task unloading and resource distribution results, and completing calculation task unloading and resource distribution based on the vehicle speed perception of multi-agent reinforcement learning.
Simulation experiment
The simulation platform is realized in a Python3.7 environment, the tensoflow version is 1.15.0, the detailed simulation parameter setting is shown in tables 1 and 2, the existing calculation unloading and resource allocation algorithms in the experimental result are AL, AV, RD and EDG algorithms, and the corresponding algorithm of the invention is JDEE-MADDPG algorithm.
TABLE 1
Figure BDA0002573996410000138
Figure BDA0002573996410000141
TABLE 2
Parameter(s) Set value Parameter(s) Set value
Number of layers 3 Layer type Fully Connected
Number of hidden units 512 Evaluating web learning rate 0.001
Optimizer Adam Action network learning rate 0.0001
Period of time 100000 Activating a function Relu
Sampling pool 128 Buffer size 20000
The average task completion time delay of the vehicles with the five algorithms is compared, and the experiment mainly evaluates the distribution condition of the average task completion time delay corresponding to each algorithm in different vehicle numbers. The experimental results are shown in fig. 2,3,4 and 5. As can be seen from fig. 2,3,4 and 5, compared with AL, AV and RD algorithms, the JDEE-madpg algorithm of the present invention can always maintain a lower task completion delay for each vehicle because the present invention can more accurately allocate computing resources and radio resources to vehicles according to task priority, task size, vehicle speed and channel status of the vehicle, and in addition, the task completion delay of some vehicles of the EDG algorithm is smaller than the JDEE-madpg algorithm of the present invention because the algorithm of the present invention sacrifices a bit of task completion delay to reduce energy consumption of the vehicle terminal without exceeding the task delay threshold.
The vehicle average task energy consumption comparison of the five algorithms is shown in fig. 6, 7, 8 and 9, and compared with other algorithms, the present invention can always maintain a lower energy consumption level, because the present invention makes an optimal unloading and resource allocation strategy according to task priority, task size, vehicle speed and channel state of the vehicle, and reduces the task energy consumption of all vehicles as much as possible.
When the vehicle speed ranges are [30,50] Km/h and [50,80] Km/h respectively, the average task completion time delay and the average task energy consumption of the vehicle with the five algorithms are compared, and the experimental results are shown in FIG. 10, FIG. 11, FIG. 12 and FIG. 13. In addition, the reason why the task completion time delay of some vehicles is higher than that of the EDG algorithm is that the JDEE-MADDPG algorithm provided by the invention distributes more wireless and computing resources of the VEC server to high-speed vehicles on the premise of not exceeding the task time delay threshold so as to reduce the overall energy consumption of the vehicle end, so that the task completion time delay of some vehicles is higher than that of the EDG algorithm.
The above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.

Claims (10)

1. A multi-agent-based vehicle speed perception computing task offloading and resource allocation method is characterized by comprising the following steps:
1) collecting vehicle end calculation tasks, dividing the vehicle end calculation tasks into key tasks, high-priority tasks and low-priority tasks according to the types of the vehicle end calculation tasks, and determining time delay thresholds of the key tasks, the high-priority tasks and the low-priority tasks according to the vehicle speed;
2) for the key tasks, the high-priority tasks and the low-priority tasks, respectively calculating different calculation resources and wireless resource allocation, unloading to a VEC server, executing locally and continuously waiting for corresponding time delay and energy consumption, and then forming a target function for reducing the energy consumption of the processing tasks at the vehicle end under the constraint of the time delay threshold of each task;
3) converting the objective function obtained in the step 2) into a Markov decision process, and initializing a state space, an action space and reward of the Markov decision process;
4) obtaining new states, actions and rewards according to the multi-agent reinforcement learning network, and storing the new states, actions and rewards into an experience playback pool;
5) when the data in the experience playback pool reaches a threshold value, training the multi-agent reinforcement learning network until the multi-agent reinforcement learning network converges;
6) and inputting the states of the vehicle end and the edge server to be distributed into the trained multi-agent reinforcement learning network to obtain task unloading and resource distribution results, and finishing calculation task unloading and resource distribution based on the vehicle speed perception of multi-agent reinforcement learning.
2. The multi-agent-based vehicle speed perception computation task offloading and resource allocation method of claim 1, wherein the vehicle-side computation task is divided into a critical task phi according to bandwidth and latency requirements of the vehicle-side computation task1High priority task phi2And low priority task phi3Critical task phi1High priority task phi2And low priority task phi3The corresponding time delay thresholds are Thr respectively1、Thr2And Thr3Wherein the critical task phi1With a delay threshold of 10ms, a low priority task phi3With a delay threshold of 100ms, a high priority task phi2In relation to the current driving speed of the vehicle, wherein the mission
Figure FDA0002573996400000011
Delay threshold of
Figure FDA0002573996400000012
Comprises the following steps:
Figure FDA0002573996400000021
wherein when
Figure FDA0002573996400000022
When the temperature of the water is higher than the set temperature,
Figure FDA0002573996400000023
when in use
Figure FDA0002573996400000024
When the temperature of the water is higher than the set temperature,
Figure FDA0002573996400000025
Figure FDA0002573996400000026
speed, Thr, of vehicle k2Limiting speed v for roadmaxAnd (4) corresponding delay threshold.
3. The multi-agent vehicle speed perception based computational task offloading and resource allocation method of claim 1, wherein the VEC server allocates the transmission rate of the uplink channel n of vehicle k to
Figure FDA0002573996400000027
Comprises the following steps:
Figure FDA0002573996400000028
wherein σ2As noise power, P is transmissionThe power is transmitted to the power transmission device,
Figure FDA0002573996400000029
for uplink channel interference, channel bandwidth
Figure FDA00025739964000000210
Figure FDA00025739964000000211
For the total upstream bandwidth of the VEC server,
Figure FDA00025739964000000212
for the number of upstream channels of the VEC server,
Figure FDA00025739964000000213
is an uplink channel set;
is provided with
Figure FDA00025739964000000214
Indicating whether the uplink channel n between the vehicle k and the VEC server is allocated to the vehicle k, if so, then
Figure FDA00025739964000000215
Is 1, otherwise, then
Figure FDA00025739964000000216
Is 0, the uplink transmission rate between the vehicle k and the VEC server is obtained
Figure FDA00025739964000000217
Comprises the following steps:
Figure FDA00025739964000000218
4. multi-agent vehicle speed perception based computational task offloading and resources as recited in claim 1Method for allocating data, characterized in that the VEC server allocates the transmission rate of the downlink channel n of the vehicle k
Figure FDA00025739964000000219
Comprises the following steps:
Figure FDA00025739964000000220
wherein σ2Is the noise power, P is the transmission power,
Figure FDA00025739964000000221
for downlink channel interference, channel bandwidth
Figure FDA00025739964000000222
Figure FDA00025739964000000223
For the total bandwidth downstream of the VEC server,
Figure FDA00025739964000000224
the number of the downlink channels of the VEC server,
Figure FDA00025739964000000225
is a downlink channel set;
is provided with
Figure FDA0002573996400000031
Indicating whether a downlink channel n between the vehicle k and the VEC server is allocated to the vehicle k, if so, then
Figure FDA0002573996400000032
Is 1, otherwise, then
Figure FDA0002573996400000033
Is 0, the downlink transmission rate between the vehicle k and the VEC server is obtained
Figure FDA0002573996400000034
Comprises the following steps:
Figure FDA0002573996400000035
5. multi-agent vehicle speed perception based computational task offloading and resource allocation method according to claim 1, characterized in that task of vehicle k
Figure FDA0002573996400000036
Total latency offloaded to VEC server execution consumption
Figure FDA0002573996400000037
Comprises the following steps:
Figure FDA0002573996400000038
wherein the content of the first and second substances,
Figure FDA0002573996400000039
in order to be a function of rounding up,
Figure FDA00025739964000000310
for the task of vehicle k
Figure FDA00025739964000000311
The size of the file of (a) is,
Figure FDA00025739964000000312
for processing the tasks of the vehicle k
Figure FDA00025739964000000313
The required density of the calculations to be made,
Figure FDA00025739964000000314
task downloaded for vehicle k
Figure FDA00025739964000000315
The file size of the file is reduced relative to the original uploading task,
Figure FDA00025739964000000316
tasking VEC servers to vehicle k
Figure FDA00025739964000000317
Proportion of allocated computing resources, fVECIs the CPU frequency of the local VEC server,
Figure FDA00025739964000000318
for the upstream transmission rate between the vehicle and the VEC server,
Figure FDA00025739964000000319
is the downlink transmission rate.
6. The multi-agent vehicle speed perception based computational task offloading and resource allocation method of claim 1, wherein the CPU frequency f allocated according to vehicle k iskTask for calculating vehicle k
Figure FDA00025739964000000320
Time consumption of local execution
Figure FDA00025739964000000321
Comprises the following steps:
Figure FDA00025739964000000322
7. multi-agent based multi-agent system according to claim 1Method for offloading and allocating vehicle speed-aware computing tasks, characterized in that at time t, the tasks of vehicle k
Figure FDA00025739964000000323
Can choose to continue waiting, offloading to local VEC server and execute locally
Figure FDA0002573996400000041
Representing the mission of a vehicle k
Figure FDA0002573996400000042
Whether to continue waiting or not, and when continuing waiting, then
Figure FDA0002573996400000043
Is 1, otherwise, then
Figure FDA0002573996400000044
Set the continuous waiting time to be Th，
Figure FDA0002573996400000045
Representing the mission of a vehicle k
Figure FDA0002573996400000046
Whether to offload to a local VEC server when
Figure FDA0002573996400000047
If 1, the task of the vehicle k is represented
Figure FDA0002573996400000048
Off-load to local VEC Server, mission for vehicle k
Figure FDA0002573996400000049
From production tasks
Figure FDA00025739964000000410
Total delay to completion of execution of an action
Figure FDA00025739964000000411
Comprises the following steps:
Figure FDA00025739964000000412
wherein the content of the first and second substances,
Figure FDA00025739964000000413
for tasks according to vehicle k
Figure FDA00025739964000000414
The time of generation.
8. The multi-agent vehicle speed perception based computational task offloading and resource allocation method of claim 1, wherein the computational tasks for vehicle k are when
Figure FDA00025739964000000415
When unloading to the VEC server, the energy consumption of the unloading task comprises the energy consumed by the uploading calculation task and the energy consumed by the downloading task, and the energy consumption of the unloading to the VEC server
Figure FDA00025739964000000416
Comprises the following steps:
Figure FDA00025739964000000417
9. the multi-agent vehicle speed perception based computational task offloading and resource allocation method of claim 1, wherein the computational tasks for vehicle k are when
Figure FDA00025739964000000418
Upon local processing, depending on the task of processing vehicle k
Figure FDA00025739964000000419
Required energy density
Figure FDA00025739964000000420
Energy consumption resulting from local processing of tasks
Figure FDA00025739964000000421
Comprises the following steps:
Figure FDA00025739964000000422
at time t, after the vehicle executes the offload policy, the energy consumed by all vehicles within the service range of the local VEC server, e (t), is:
Figure FDA00025739964000000423
10. the multi-agent-based vehicle speed perception computational task offloading and resource allocation method of claim 1, wherein under the condition of limited task delay thresholds, computational resources and radio resources, the objective function for reducing the energy consumption of the processing task at the vehicle end is formed as follows:
Figure FDA0002573996400000051
wherein the content of the first and second substances,
Figure FDA0002573996400000052
CN202010648403.2A 2020-07-07 2020-07-07 Multi-agent-based vehicle speed perception calculation task unloading and resource allocation method Active CN111918245B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010648403.2A CN111918245B (en) 2020-07-07 2020-07-07 Multi-agent-based vehicle speed perception calculation task unloading and resource allocation method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010648403.2A CN111918245B (en) 2020-07-07 2020-07-07 Multi-agent-based vehicle speed perception calculation task unloading and resource allocation method

Publications (2)

Publication Number Publication Date
CN111918245A true CN111918245A (en) 2020-11-10
CN111918245B CN111918245B (en) 2021-11-19

Family

ID=73227589

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010648403.2A Active CN111918245B (en) 2020-07-07 2020-07-07 Multi-agent-based vehicle speed perception calculation task unloading and resource allocation method

Country Status (1)

Country Link
CN (1) CN111918245B (en)

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112616131A (en) * 2020-12-11 2021-04-06 北京邮电大学 Internet of vehicles resource allocation method based on video content priority
CN112689296A (en) * 2020-12-14 2021-04-20 山东师范大学 Edge calculation and cache method and system in heterogeneous IoT network
CN112685186A (en) * 2021-01-08 2021-04-20 北京信息科技大学 Method and device for unloading computing tasks, electronic equipment and storage medium
CN112929412A (en) * 2021-01-22 2021-06-08 山东大学 Method, device and storage medium for joint bandwidth allocation and data stream unloading based on MEC single-vehicle single-cell
CN113452625A (en) * 2021-06-28 2021-09-28 重庆大学 Deep reinforcement learning-based unloading scheduling and resource allocation method
CN113542376A (en) * 2021-07-01 2021-10-22 广东工业大学 Task unloading method based on energy consumption and time delay weighting
CN113726858A (en) * 2021-08-12 2021-11-30 西安交通大学 Self-adaptive AR task unloading and resource allocation method based on reinforcement learning
CN113918240A (en) * 2021-10-15 2022-01-11 全球能源互联网研究院有限公司 Task unloading method and device
CN114285847A (en) * 2021-12-17 2022-04-05 中国电信股份有限公司 Data processing method and device, model training method and device, electronic equipment and storage medium
CN114637608A (en) * 2022-05-17 2022-06-17 之江实验室 Calculation task allocation and updating method, terminal and network equipment
CN115022322A (en) * 2022-06-02 2022-09-06 湖南商务职业技术学院 Edge cloud cooperation task unloading method based on crowd evolution in Internet of vehicles
CN115237506A (en) * 2022-09-21 2022-10-25 暨南大学 Reliability-driven hierarchical task unloading method and device in CPS (control performance Standard) system
CN116541106A (en) * 2023-07-06 2023-08-04 闽南理工学院 Computing task unloading method, computing device and storage medium

Citations (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2012056437A1 (en) * 2010-10-29 2012-05-03 École Polytechnique Fédérale De Lausanne (Epfl) Omnidirectional sensor array system
US20160066360A1 (en) * 2014-09-02 2016-03-03 Apple Inc. Api for web access directly or through intermediary
CN106134263A (en) * 2014-01-29 2016-11-16 交互数字专利控股公司 Ul transmissions in radio communication
US20180109428A1 (en) * 2016-10-19 2018-04-19 Tata Consultancy Services Limited Optimal deployment of fog computations in iot environments
CN108541027A (en) * 2018-04-24 2018-09-14 南京邮电大学 A kind of communication computing resource method of replacing based on edge cloud network
CN109257432A (en) * 2018-10-12 2019-01-22 桂林电子科技大学 A kind of target switching method, computer installation and readable storage medium storing program for executing
US20190042513A1 (en) * 2018-06-30 2019-02-07 Kermin E. Fleming, JR. Apparatuses, methods, and systems for operations in a configurable spatial accelerator
CN109478057A (en) * 2016-05-09 2019-03-15 强力物联网投资组合2016有限公司 Method and system for industrial Internet of Things
CN110035410A (en) * 2019-03-07 2019-07-19 中南大学 Federated resource distribution and the method and system of unloading are calculated in a kind of vehicle-mounted edge network of software definition
CN110650457A (en) * 2019-10-14 2020-01-03 河海大学 Joint optimization method for task unloading calculation cost and time delay in Internet of vehicles
CN110868700A (en) * 2019-10-16 2020-03-06 深圳大学 Cooperative computing unloading method based on splittable tasks in vehicle-mounted edge computing environment
CN110933157A (en) * 2019-11-26 2020-03-27 重庆邮电大学 Industrial Internet of things-oriented edge computing task unloading method
CN111124647A (en) * 2019-12-25 2020-05-08 大连理工大学 Intelligent edge calculation method in Internet of vehicles
CN111142883A (en) * 2019-12-03 2020-05-12 沈阳航空航天大学 Vehicle computing task unloading method based on SDN framework
CN111262940A (en) * 2020-01-17 2020-06-09 中南大学 Vehicle-mounted edge computing application caching method, device and system

Patent Citations (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2012056437A1 (en) * 2010-10-29 2012-05-03 École Polytechnique Fédérale De Lausanne (Epfl) Omnidirectional sensor array system
CN106134263A (en) * 2014-01-29 2016-11-16 交互数字专利控股公司 Ul transmissions in radio communication
US20160066360A1 (en) * 2014-09-02 2016-03-03 Apple Inc. Api for web access directly or through intermediary
CN109478057A (en) * 2016-05-09 2019-03-15 强力物联网投资组合2016有限公司 Method and system for industrial Internet of Things
US20180109428A1 (en) * 2016-10-19 2018-04-19 Tata Consultancy Services Limited Optimal deployment of fog computations in iot environments
CN108541027A (en) * 2018-04-24 2018-09-14 南京邮电大学 A kind of communication computing resource method of replacing based on edge cloud network
US20190042513A1 (en) * 2018-06-30 2019-02-07 Kermin E. Fleming, JR. Apparatuses, methods, and systems for operations in a configurable spatial accelerator
CN109257432A (en) * 2018-10-12 2019-01-22 桂林电子科技大学 A kind of target switching method, computer installation and readable storage medium storing program for executing
CN110035410A (en) * 2019-03-07 2019-07-19 中南大学 Federated resource distribution and the method and system of unloading are calculated in a kind of vehicle-mounted edge network of software definition
CN110650457A (en) * 2019-10-14 2020-01-03 河海大学 Joint optimization method for task unloading calculation cost and time delay in Internet of vehicles
CN110868700A (en) * 2019-10-16 2020-03-06 深圳大学 Cooperative computing unloading method based on splittable tasks in vehicle-mounted edge computing environment
CN110933157A (en) * 2019-11-26 2020-03-27 重庆邮电大学 Industrial Internet of things-oriented edge computing task unloading method
CN111142883A (en) * 2019-12-03 2020-05-12 沈阳航空航天大学 Vehicle computing task unloading method based on SDN framework
CN111124647A (en) * 2019-12-25 2020-05-08 大连理工大学 Intelligent edge calculation method in Internet of vehicles
CN111262940A (en) * 2020-01-17 2020-06-09 中南大学 Vehicle-mounted edge computing application caching method, device and system

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
JUN LIU: "A Task Oriented Computation Offloading Algorithm for Intelligent Vehicle Network With Mobile Edge Computing", 《IEEE ACCESS》 *
WENHAN ZHAN: "Deep-Reinforcement-Learning-Based Offloading Scheduling for Vehicular Edge Computing", 《 IEEE INTERNET OF THINGS JOURNAL》 *
基于马尔科夫决策过程的车载边缘计算切换策略: "李波", 《计算机工程与科学》 *
谷晓会: "移动边缘计算在车载网中的应用综述", 《计算机应用研究》 *

Cited By (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112616131A (en) * 2020-12-11 2021-04-06 北京邮电大学 Internet of vehicles resource allocation method based on video content priority
CN112616131B (en) * 2020-12-11 2022-05-13 北京邮电大学 Internet of vehicles resource allocation method based on video content priority
CN112689296A (en) * 2020-12-14 2021-04-20 山东师范大学 Edge calculation and cache method and system in heterogeneous IoT network
CN112689296B (en) * 2020-12-14 2022-06-24 山东师范大学 Edge calculation and cache method and system in heterogeneous IoT network
CN112685186A (en) * 2021-01-08 2021-04-20 北京信息科技大学 Method and device for unloading computing tasks, electronic equipment and storage medium
CN112685186B (en) * 2021-01-08 2023-04-28 北京信息科技大学 Method and device for unloading computing task, electronic equipment and storage medium
CN112929412B (en) * 2021-01-22 2022-03-11 山东大学 Method, device and storage medium for joint bandwidth allocation and data stream unloading based on MEC single-vehicle single-cell
CN112929412A (en) * 2021-01-22 2021-06-08 山东大学 Method, device and storage medium for joint bandwidth allocation and data stream unloading based on MEC single-vehicle single-cell
CN113452625A (en) * 2021-06-28 2021-09-28 重庆大学 Deep reinforcement learning-based unloading scheduling and resource allocation method
CN113542376A (en) * 2021-07-01 2021-10-22 广东工业大学 Task unloading method based on energy consumption and time delay weighting
CN113726858A (en) * 2021-08-12 2021-11-30 西安交通大学 Self-adaptive AR task unloading and resource allocation method based on reinforcement learning
CN113918240A (en) * 2021-10-15 2022-01-11 全球能源互联网研究院有限公司 Task unloading method and device
CN114285847A (en) * 2021-12-17 2022-04-05 中国电信股份有限公司 Data processing method and device, model training method and device, electronic equipment and storage medium
CN114637608A (en) * 2022-05-17 2022-06-17 之江实验室 Calculation task allocation and updating method, terminal and network equipment
CN115022322A (en) * 2022-06-02 2022-09-06 湖南商务职业技术学院 Edge cloud cooperation task unloading method based on crowd evolution in Internet of vehicles
CN115022322B (en) * 2022-06-02 2024-02-02 湖南第一师范学院 Edge cloud cooperation task unloading method based on crowd-sourced evolution in Internet of vehicles
CN115237506A (en) * 2022-09-21 2022-10-25 暨南大学 Reliability-driven hierarchical task unloading method and device in CPS (control performance Standard) system
CN116541106A (en) * 2023-07-06 2023-08-04 闽南理工学院 Computing task unloading method, computing device and storage medium
CN116541106B (en) * 2023-07-06 2023-08-29 闽南理工学院 Computing task unloading method, computing device and storage medium

Also Published As

Publication number Publication date
CN111918245B (en) 2021-11-19

Similar Documents

Publication Publication Date Title
CN111918245B (en) Multi-agent-based vehicle speed perception calculation task unloading and resource allocation method
CN112601197B (en) Resource optimization method in train-connected network based on non-orthogonal multiple access
CN110035410B (en) Method for joint resource allocation and computational offloading in software-defined vehicle-mounted edge network
CN111124647B (en) Intelligent edge calculation method in Internet of vehicles
CN111711666B (en) Internet of vehicles cloud computing resource optimization method based on reinforcement learning
CN111132074B (en) Multi-access edge computing unloading and frame time slot resource allocation method in Internet of vehicles environment
Huang et al. Vehicle speed aware computing task offloading and resource allocation based on multi-agent reinforcement learning in a vehicular edge computing network
CN113568675A (en) Internet of vehicles edge calculation task unloading method based on layered reinforcement learning
CN114143346B (en) Joint optimization method and system for task unloading and service caching of Internet of vehicles
CN111240821B (en) Collaborative cloud computing migration method based on Internet of vehicles application security grading
CN113641417B (en) Vehicle security task unloading method based on branch-and-bound method
CN113645273B (en) Internet of vehicles task unloading method based on service priority
CN114268923A (en) Internet of vehicles task unloading scheduling method and system
CN115277845A (en) Multi-agent near-end strategy-based distributed edge cache decision method for Internet of vehicles
CN115835294A (en) RAN slice and task unloading joint optimization method assisted by deep reinforcement learning in Internet of vehicles
CN114374949A (en) Power control mechanism based on information freshness optimization in Internet of vehicles
CN112888021B (en) Task unloading method for avoiding interruption in Internet of vehicles
Li et al. Collaborative optimization of edge-cloud computation offloading in internet of vehicles
Wang et al. Research on V2I/V2V hybrid multi-hop edge computing offloading algorithm in IoV environment
CN112835637A (en) Task unloading method for vehicle user mobile edge calculation
CN114786156A (en) Internet of vehicles optimization task edge unloading method based on deep learning
CN116367231A (en) Edge computing Internet of vehicles resource management joint optimization method based on DDPG algorithm
CN115865914A (en) Task unloading method based on federal deep reinforcement learning in vehicle edge calculation
CN113573280B (en) Vehicle edge calculation cost-effective optimization method, system, equipment and terminal
Lyu et al. Service-driven resource management in vehicular networks based on deep reinforcement learning

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant