DynaMAC

---

ECE257B Project - Reinforcement Learning based MAC selection engine - Ankit Agarwal, Rohit Kumar and Yeswanth Guntupalli

Description

This repo contains the simulation code used for simulating and evaluating DynaMAC - an RL based MAC Selection Engine.

Requirements

Python with numpy installed

Code

The main functions involved in simulating DynaMAC are:

• generate_events
• csma_simulator
• tdma_simulator
• qlearning_egreedy
• qlearning_boltzmann
• decision_final
• Dynamac_SOMAC_test

Steps to run the code (demo.ipynb)

• A sample version of the above test and results are also included in a python notebook demo.ipynb and inluded in the repository

This notebook allows you to run and test the following:

• Run Standalone CSMA simulator for inreasing number of nodes and different network conditions over configurable monte-carlo simulations
• Run Standalone TDMA simulator for inreasing number of nodes and different network conditions over configurable monte-carlo simulations
• Run DynaMAC comparison for a given network condition and get a comparison of performance across TDMA only, CSMA only , DynaMAC with eplison greedy approach and DynaMAC with softmax approach

DynaMAC Comparison Test

• This simulation gives the user control over the number of nodes, number of packets per node per second, duration of high traffic, duration of low traffic etc
• Based on the above parameters, the code calculates the throughput and latency for the network for the 4 different approaches mentioned above