Simulation and design of a glider swarm robotics platform

Abstract

Tezin ana konusu, Hücresel Otomasyon Glider Patern formasyonunun kullanarak Game of Life algoritmasını gerçekleştirecek donanımsal sürü robotların ve gerçek zamanlı simulasyon arayüzünün dizaynı, üretilmesi ve testlerini içerir.Hücresel otomasyon robotların yol bulmasında oldukça verimli bir metod olarak kullanılmaktadır. Bununla birlikte sürü robotların kontrolünde Glider modeli uygulaması ve Game of Life algoritması hedeflenmiştir.Her robot kendi yol ve harekete başlama ve bitirme sinyallerini PC üzerindeki simülasyondan alır. Robotların tekil olarak hareketlerini tamamlamasından sonra, PC tarafındaki simülasyona gelen sinyal ile ekrandaki dizilim şekli yenilenir.Robotların elektronik temelinde Arduino nano board, NRF24L01 iletişim modülü, 28BJY-48 adım motor, ULN2803 motor sürücü entegre ve 4xAA pil kullanılmıştır.Robotlarda sensör kullanılmadığı için mekanik, elektronik ve yazılım yönüyle üretimleri ucuz ve kolay seviyede tutulmuştur.Platform üzerinde yapılabilecek geliştirmeler arasında sensörlerin eklenmesi ile konum bilgisi, PC ihtiyacı olmadan robotların kendi aralarında haberleşmenin sağlanması gibi ilerlemeler kaydedilebilir.

The main purpose of my thesis is to develop a Swarm Robotics Platform which will use Cellular Automata Glider model and generate itself with `Game of Life` rules. The method was experimentally tested with autonomous mobile robots and real-time PC based simulation software, in all cases very good paths were obtained with negligible processing effort, and low cost production. Presented results indicate that the Cellular Automata approach is a very promising method for real time path planning and Glider like robotic swarms can be used for self-replicating and moving swarm robots.We designed five identical mobile robots that interact with each other and the simulation software at PC through RF connection. PC runs a Glider simulation simultaneously while robots play `Game of Life` on the grid. Presented model is a self-organized and a self-driven mechanism.The base platform used is a lattice of squared cells, but the shape of cells can be hexagonal and other shapes as well. Each cell can exist in 2 or more different states (not simultaneoulsy). Most basically ON/OFF states of bright LEDs at the top of each robot is controlled as an indicator to show dead/alive modes of the robots. We expect our robots to move on the lattice base as in Glider form and keep their formation patterns after each step.Mobile robots are based on Arduino Nano boards. NRF24L01 RF modules used for communication. ULN2803 IC is used to drive two stepper motors (28BJY-48). Each robot is powered with a pack of 4 AA batteries.Each member robot will keep its track and location information and inform the main PC simulation software. After each robot completes its action, the simulation software moves to the next pattern of Glider and the LEDs of robots will be turned ON (Live Mode).Atmel328P based board with a NRF24L01 RF module establishes the real time communication between Glider robots and the PC. Main module transfers required pattern data to each individual Glider robot and receives a confirmation of correct data transmission from each robot. After each robot gets its required data, then the main module updates the information of simulation software that runs on the PC. It is possible to observe the pattern evolution of Glider robots on the simulation software. Processing programming language is used to create the simulation software.With all these simple and commonly found parts, each robot was produced with quite a cheap and simple way. Thus, the total number of robots can be increased to more than 5 and different CA models can be realized with this platform.Some improvements should be done on our system such as including another NRF24L01 module for each 6 mobile robots due to available channel number restriction of RF module used. Atmel2560 based board can provide much more communication capability for crowder Swarms with additional RF modules.Another future development of the system should be to include a path finding algorithm into simulation and create `maze solving` or `target searching` Swarm group with much better performance. This method will allow to conclude results in much shorter times and with much less effort.