hd818 / soCoM

A semi-online Computational Offloading Model, which utilizes users’ behavior prediction method to optimize task offloading in edge computing environments.

soCoM

Python code to reproduce our soCoM model, which utilizes users’ behavior prediction methods to optimize task offloading in edge computing environments. This is the code of paper in title 'Semi-online Computational Offloading by Dueling Deep-Q Network for User Behavior Prediction'. DOI: 10.1109/ACCESS.2020.3004861

It includes:

• soCoM.py: The system model for soCoM, including definition of the task, user, MEC server, communication model, computing model, and energy consumption model.

• OFFLOAD.py: RL offloading training process.

• RLbrain*.py: RL algorithm of DQN, Dueling DQN, Double DQN, Prioritized replay.

• Simulation.py: run this file for soCoM, creating a simulated environment.

• soCoMM.py, OFFLOADM.py, Simulation-multi.py: Multiple servers senario.

Required packages

• SimPy: https://simpy.readthedocs.io/en/latest/
• Tensorflow 1.0

How the code works

• For the soCoM simulation, run the file Simulation.py.

• For changing the numbers of user equipment, change the global variable 'UN' in the file soCoM.py.

• For changing the DQN algorithms, change the import of package in the file OFFLOAD.py.

About authors

• Shinan Song, songshinan AT 163.com