Lokalizacija mobilnega robota v znanem okolju z uporabo merilnikov razdalj

GERČAR, MARJAN (author), Klančar, Gregor (mentor)

Enega temeljnih problemov v mobilni robotiki predstavlja lokalizacija mobilnega robota, saj je uspeh, da se le-ta dobro lokalizira, povezan z različnimi dejavniki. Osnovni pojem pri obravnavani tematiki, predstavlja verjetnost, na temelju česar je tudi nastala verjetnostna robotika, katere verjetnostne teorije so popularne in izredno uspešno uporabljene v praksi. Njihova prednost je predvsem ta, da upoštevajo negotovosti, tako pri izračunih kot tudi pri načrtovanju sistema ter tudi, da z njihovo pomočjo lahko z verjetnostno porazdelitvijo predstavimo stanje nekega sistema. V diplomskem delu sem se osredotočil na splošen problem globalne lokalizacije, katera predstavlja problem znanega zemljevida ter neznane začetne lokacije robota. Za reševanje omenjenega problema je pomembno, da poznamo lokalizacijske algoritme, ter njihove osnove. Algoritmi izračunavajo verjetnostno porazdelitev, tako da je lokalizacija mobilnega robota uspešna. Za namene diplomskega dela sem izvedel dva projekta, pri katerih sem uporabil dva verjetnostna algoritma, histogram filter ter filter delcev. Oba izhajata iz Bayesovega filtra, ki obenem predstavlja izhodišče za vse verjetnostne algoritme. Prvi za predstavitev verjetnostne porazdelitve uporablja diskretno mrežno lokalizacijo (v definirani mreži celic), drugi pa uporablja zvezno Monte Carlo metodo lokalizacije.

• Language: Slovenian
• Keywords: Lokalizacija mobilnega robota, globalna lokalizacija, algoritmi za verjetnostno porazdelitev, histogram filter, filter delcev, Bayesov filter, mrežna lokalizacija, Monte Carlo lokalizacija.
• Work type: Undergraduate thesis
• Organization: FE - Faculty of Electrical Engineering
• Year: 2016
• PID: 20.500.12556/RUL-82775
• Publication date in RUL: 23.05.2016

Secondary language

• Language: English
• Title: Localisation of mobile robot in known environment using distance sensors

Abstract

One of the core problems in mobile robotics is the localisation of the mobile robot, as the success rate of achieving localisation depends on many different factors. The core idea with my chosen topic is represented by probability, which advanced into its own field of robotics – probabilistic robotics, of which theories are incredibly widespread and popular in practice. Their advantage lies mainly in acknowledging uncertainties; in calculations, system planning and in presenting the state of a certain system through probabilistic distribution The focus of my thesis is in the general problem of global localisation, which represents the problem of the known map and unknown starting location of the robot. In order to solve the aforementioned problem it is important to have knowledge of the localisation algorithms, which calculate the probabilistic distribution in a manner that leads to the successful localisation of a robot. In support of my thesis I engineered two projects, which used two probabilistic algorithms: a histogram filter and a particle filter. Both descend from the Bayes filter, which represents the foundation for all probabilistic algorithms. The former algorithm uses, for the purposes of probabilistic distribution, a discreet grid localisation (in a defined grid cell) while the latter uses the Monte Carlo localisation method.

• Keywords: Mobile robot localisation, global localisation, probabilistic distribution algorithms, histogram filter, particle filter, Bayes filter, grid localisation, Monte Carlo localisation.