Behaviour-based control of autonomous robots

Abstract

Ciel'om bakalárskej práce bola implementácia frameworku v jazyku C++, ktorý bude slúžit' na kompaktnejší vývoj umelej inteligencie (UI) autonómnych robotov založenej na chovaní (behaviour-based robotics). Táto UI je založená na pravidlách, parametroch a roznych "chovaniach" robota, v našom prípade vopred definovaných heuristicky. Platforma umožňuje posunút' t'ažisko vývoja takejto UI na definovanie týchto chovaní a pravidiel namiesto programovania. Súčast'ou platformy sú taktiež genetické algoritmy, ktoré umožňujú optimalizáciu parametrov heuristického riešenia. Funkčnost' konceptu bola overená na konkrétnom simulovanom pokuse s viacerými robotmi, ktorým bola zadaná jednoduchá konštrukčná úloha. Všetky simulácie boli uskutočnené pomocou simulačného softvéru Player/Stage.

The aim of this bachelor's thesis was an implementation of a library, which should enable to develop controllers for autonomous mobile robots. The controllers are based on ideas of Behaviour-based robotics. This means that such a robot has a set of elementary behaviours and its control system consists of a set of parametrized rules, which decide according to the state of its sensors which one of the pre-defined elementary behaviours the robot has to use. This program shifts the substantial part of a robot control development from programming to defining such elementary behaviours and sets of rules. It includes a genetic algorithm that can help to optimize parameters of the rules used and/or the behaviours used in a controller. As a proof-of-concept a simulated experiment with multiple agents was conducted, in which the robots were assigned a simple construction task. All simulations were done in the Player/Stage simulator.