Συντάχθηκε 27-09-2013 18:22
από Balasi Panagiota
Email συντάκτη:
Ενημερώθηκε:
-
Ιδιότητα: -.
ΠΟΛΥΤΕΧΝΕΙΟ ΚΡΗΤΗΣ
Σχολή Ηλεκτρονικών Μηχανικών και Μηχανικών Υπολογιστών
Πρόγραμμα Προπτυχιακών Σπουδών
ΠΑΡΟΥΣΙΑΣΗ ΔΙΠΛΩΜΑΤΙΚΗΣ ΕΡΓΑΣΙΑΣ
ΕΥΑΓΓΕΛΟΥ ΜΙΧΕΛΙΟΥΔΑΚΗ
με θέμα
Δυναμικός Πολυρομποτικός Συντονισμός
για το Πρωτάθλημα Standard Platform του RoboCup
Dynamic Multi-Robot Coordination
for the RoboCup Standard Platform League
Δευτέρα 30 Σεπτεμβρίου 2013, 2:30μμ
Αίθουσα Εργαστηρίου Intelligence, Κτίριο Επιστημών, Πολυτεχνειούπολη
Εξεταστική Επιτροπή
Αναπληρωτής Καθηγητής Μιχαήλ Γ. Λαγουδάκης (επιβλέπων)
Επίκουρος Καθηγητής Γεώργιος Χαλκιαδάκης
Επίκουρος Καθηγητής Άγγελος Μπλέτσας
Abstract
Coordination among robots is a popular research topic in robotics, artificial intelligence, and multi-agent systems. Robotic soccer, known as RoboCup, represents a complex, stochastic, real-time, multi-agent competitive domain. In such domains, cooperative strategy is required for the successful of the team's mission, thus the ability of team coordination is crucial. Team coordination in RoboCup offer mechanisms for player positioning and role assignment. This thesis studies the problem of coordination for a robotic soccer team and offers a real-time module that achieves dynamic positioning and role assignment using the basic principles of utility theory. Specifically, a formation generator component is responsible for generating a set of candidate positions in the field, based on the global estimated ball position as provided by the shared estimated global game state. The formation type (offensive or defensive) is determined dynamically, depending on which half of the field the ball lies in and the number of candidate positions represent certain roles (supporter, attacker, defender, etc.). Then, a role assignment component uses a team utility function to evaluate mappings of players (robots) to positions (roles) and decides which mapping is the best for the current game situation. The utility function combine a variety of features to characterize good aspect of teamwork. Since the candidate positions/roles may be more than the robots, the number of possible mappings can be extremely large. Two role assignment algorithms are proposed; one based on exhaustive search that guarantees optimal solution with high computational cost and another based on stochastic local search (Particle Sward Optimization), which cannot guarantee an optimal solution, but has very low computational cost. The selected mapping is adopted by all robots and each robot executes the behavior that implements the assigned role. Our coordination modules has been integrated into the software architecture of our RoboCup team “Kouretes” and is used for coordinating the team robots during RoboCup games.
Συντάχθηκε 30-09-2013 10:18
από Michail Lagoudakis
Email συντάκτη: lagoudakis<στο>tuc.gr
Ενημερώθηκε:
-
Ιδιότητα: ΔΕΠ ΗΜΜΥ.
Προσοχή! Αλλαγή αίθουσας!
Η παρουσίαση θα πραγματοποιηθεί στην Αίθουσα Συνεδριάσεων ΗΜΜΥ, Κτίριο Επιστημών, Πολυτεχνειούπολη.
