Unapređivanje učinkovitosti višerobotskih mobilnih sustava optimizacijom individualnih upravljačkih pravila

master's thesis

• Title (english): Improving multi-robot system performance by optimizing individual control policies
• Author: Mirko Kokot
• Mentor: Zdenko Kovačić (mentor)
• Committee member: Zdenko Kovačić (predsjednik povjerenstva)
• Committee member: Stjepan Bogdan (član povjerenstva)
• Committee member: Matko Orsag (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2017-07-17, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Electrical Engineering

Abstract

U ovome radu predložena je metoda distribuiranog upravljanja višerobotskim sustavom temeljena na individualnim pravilima. U slučaju konflikta dvaju vozila, svako vozilo donosi zasebnu odluku na temelju razlike njihovih vektora značajki. Kako bi se došlo do odluke, predložena je nova metoda odlučivanja u N-dimenzionalnom prostoru značajki gdje je svaki od ishoda predstavljen N-sferom s pripadajućim parametrima koji omogućuju optimiranje metode. Predložena metoda je detaljno opisana i analizirana na temelju primjera. Proučen je učinak promjene pojedinih parametara odlučivanja. Kako bi se sustav mogao optimirati predložen je hibridni postupak optimizacije korištenjem genetskih algoritama i simuliranog kaljenja. Za provjeru ispravnosti metode izrađen je upravljački model vozila koji implementira predloženu metodu u okviru ROS meta-operacijskog sustava. Uz njega je razvijen i simulacijski čvor za simuliranje u ubrzanom vremenu kako bi se moglo optimirati parametre modela. Na kraju rada prikazani su rezultati simulacija, istaknute važne poteškoće u razvoju te prijedlozi za daljnji razvoj.

Abstract (english)

This work proposes a method of distributed multi-agent system control based on individual coordination rules. In the case of conflict between two vehicles, each vehicle makes a separate decision based on the difference between their feature vectors. In order to make a decision, a new decision-making method is proposed in the N-dimensional feature space where each of the outcomes is presented with a N-sphere with corresponding parameters that allow the method to be optimized. The proposed method is described in detail and analyzed on the basis of examples. The effect of changing individual decision parameters is studied. In order to optimize the system, a hybrid optimization process is proposed using genetic algorithms and simulated annealing methods. To test the correctness of the method, a vehicle controller that uses the proposed method is developed for the ROS meta-operating system. In addition, a simulation node for faster than real-time simulation has been developed for optimizing model parameters. At the end of the work, the results of the simulations, important development challenges and suggestions for further development are discussed.

• Keywords: višerobotski sustav; distribuirani sustav upravljanja; odlučivanje N-sferama
• Keywords (english): multi-agent system; distributed system control; N-sphere decision making
• Language: croatian
• URN:NBN: urn:nbn:hr:168:475524
• Study programme: Title: Electrical Engineering and Information Technology; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka elektrotehnike i informacijske tehnologije (magistar/magistra inženjer/inženjerka elektrotehnike i informacijske tehnologije)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2019-03-13 21:38:03