Finding and Locating Byzantine Robots in a Swarm using Machine Learning and Blockchains.
Updated: Feb 1
This project was a part of my Artificial Intelligence Course in Spring 2019. I did this in collaboration with one other student who did most of the blockchain work whereas I worked on the problem formulation and solution through machine learning.
This project attempts to deal with Byzantine or faulty nodes in multi-agent decision making process for swarm robotics. It's based on a recent paper that considers a problem of finding the dominant color in a two color grid using swarm robots and attempts to make the right decision despite the presence of Byzantine robots, using blockchain technology. We found the approach to be highly complex and also computationally expensive. We instead modeled the same problem with a little bit of modification and tried to learn the presence of Byzantine bots in said task using Stacked Autoencoders with a softmax layer. Report for the project is provided at this link.