This repository contains the code for the paper:

V.Strobel, E. Castello Ferrer, M. Dorigo. 2018. Managing Byzantine Robots via Blockchain Technology in a Swarm Robotics Collective Decision Making Scenario.

To appear in Proceedings of the 17th Conference on Autonomous Agents and MultiAgent Systems. International Foundation for Autonomous Agents and Multiagent Systems (AAMAS 2018).

A pre-print is available at:
http://iridia.ulb.ac.be/IridiaTrSeries/

There is a .ova virtual machine file that contains a runnable version of this repository. The link to the VM image is https://drive.google.com/open?id=1aDJcWafd-GziCL12MOfWdgTizpACPAb4 (the files are in ~/blockchain-swarm-robotics/).

User: Iridia
pw: Iridia

Execute Experiment 1 with:

bash start_experiment1.sh 0 0 1

Please note that we are currently working on a cleaned-up and more versatile version of this code. Below are some hints for the installation (more will follow soon). It's maybe also a good idea to check out the virtual machine file (see above) to see how the dependencies are set up.

• ARGoS 3
• ARGoS-epuck
• git
• cmake
• golang (tested with version 1.7.3)
• optional: sendmail

The easiest way to install ARGoS and the epuck plugin is via the script install_argos.sh (install the dependencies mentioned at https://github.com/ilpincy/argos3 first). install_argos.sh creates a setup script that should be executed everytime you want to run ARGoS. You can, for example, add the script to your .bashrc: source "~/argos3-dist/bin/setup_argos3"

Then, execute create_geths.sh and add the created geths (default is just geth0) to your PATH (incl. export).

Install the solidity compiler solc.

Edit the file start_experiment1.sh (change the variables BASE, mailto, etc.).

Compile the code (mkdir build; cd build; cmake ..; make).

Run an experiment: bash start_experiment1.sh 0 0 1

There is a video tutorial on how to install the framework at https://www.youtube.com/watch?v=0hEZGHlqdJ8

Using the robot swarm simulator ARGoS 3, we study a collective decision scenario, in which robots sense which of two features in an environment is the most frequent one---a best-of-2 problem. Our approach is based on the collective decision scenario of Valentini et al. Via blockchain-based smart contracts using the Ethereum protocol, we add a security layer on top of the classical approach that allows to take care of the presence of Byzantine robots. Our blockchain approach also allows to log events in a tamper-proof way: these logs can then be used to analyze, if necessary, the behavior of the robots in the swarm without incurring the risk that some malicious agent has modified them. In addition, it provides a new way to understand how we debug and do data forensics on decentralized systems such as robot

The goal of the robot swarm is to make a collective decision and to reach consensus on the most frequent tile color of a black/white grid. Each robot has a current opinion about the correct color, and via dissemination/decision-making strategies, they influence their peers. At the end of a successful run, all robots have the opinion of the majority color (in our experiments it is always the white).