Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

TD3 algorithm-based calculation unloading distribution method and device

A technology for computing unloading and distribution methods, which is applied in computing, computing models, energy-saving computing, etc., and can solve problems such as unreasonable unloading ratio and transmit power

Active Publication Date: 2021-04-02
NANJING UNIV OF POSTS & TELECOMM
View PDF4 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] Purpose of the present invention: the present invention provides a calculation offload distribution method and device based on the TD3 algorithm, which solves the unreasonable problems of the unloading ratio and the unreasonable transmission power for performing offloading in the existing calculation offloading

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • TD3 algorithm-based calculation unloading distribution method and device
  • TD3 algorithm-based calculation unloading distribution method and device
  • TD3 algorithm-based calculation unloading distribution method and device

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0097] The TD3 network includes 1 pair of actor networks and 2 pairs of critic networks. A pair of actor networks includes 1 actor current network and 1 actor target network. A pair of critic networks contains 1 critic current network and 1 critic target network. The actor target network has the same structure as the actor current network, and the critic target network has the same structure as the critic current network. First set the total number of training times to zero and the number of round updates to zero. Then initialize the state of the computing offload system, set the current number of network updates of the critic to zero, and construct training samples. Input the system state into the actor's current network, and the actor's current network obtains the current strategy through the environment, and obtains the current action after adding noise. After the action is applied to the interactive environment, the calculation and offloading system state and reward val...

Embodiment 2

[0169] A computing offloading distribution device based on the TD3 algorithm, comprising:

[0170] Computing offload TD3 network building blocks for pre-training compute offloading TD3 networks;

[0171] The distribution calculation offloading module is used to obtain the calculation offloading system status, and input the calculation offloading system status into the pre-trained actor current network of the calculation offloading TD3 network to obtain the calculation offload ratio and the calculation offload transmission power; the calculation offload system status includes: IoT device to edge server channel gain, IoT device battery power and harvestable renewable energy.

[0172] Further, at time slot t+1, the battery power b of the IoT device t+1 Expressed as:

[0173]

[0174] The energy consumption required to perform computing tasks for local IoT devices is:

[0175]

[0176] E. t 1The power required to send the offloading task from the local IoT device to t...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a TD3 algorithm-based calculation unloading distribution method and device, and the method comprises the steps: obtaining the state of a calculation unloading system, inputtingthe state of the calculation unloading system into a pre-trained actor current network of a calculation unloading TD3 network, and obtaining a calculation unloading proportion and calculation unloading transmitting power, wherein the computing unloading system state comprises a channel gain from the Internet-of-Things device to the edge server, a battery capacity of the Internet-of-Things deviceand collectable renewable energy. Through a deep reinforcement learning method, the optimal unloading strategy of the equipment is determined, and the energy consumption of the intelligent equipment is reduced by utilizing energy collection.

Description

technical field [0001] The invention belongs to the field of mobile communication and deep reinforcement learning, and in particular relates to a calculation offloading distribution method and device based on TD3 algorithm. Background technique [0002] In the past few years, the emergence of emerging technologies such as computer vision, natural language processing, and augmented reality all require a large number of computing tasks and the lowest possible latency, which often exceeds the capabilities of existing IoT (Internet of Things) devices. Calculate ability. Computational offloading is considered to be one of the effective methods to solve this problem. Computing offloading refers to the method of offloading computing tasks to high-performance servers in the edge network for processing, and then retrieving the processing results from the proxy server, thereby reducing computing delays. [0003] Secondly, for the power supply problem of IoT devices, especially in re...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): H04L29/08G06N20/00G06N3/08G16Y20/30
CPCH04L67/10H04L67/101H04L67/1023G06N20/00G06N3/084G16Y20/30Y02D10/00
Inventor 赵莎莎秦立丹张登银孙晨辉周晓宇朱子洁何培源
Owner NANJING UNIV OF POSTS & TELECOMM
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products