|Tipo di tesi||Tesi di dottorato di ricerca|
|Titolo||Coordinazione del Traffico per sistemi AGV: un approccio completo per la modellazione|
|Titolo in inglese||Traffic Coordination for AGV Systems: an Ensemble Modeling Approach|
|Settore scientifico disciplinare||ING-INF/04 - AUTOMATICA|
|Corso di studi||Scuola di D.R. in INGEGNERIA DELL'INNOVAZIONE INDUSTRIALE|
|Data inizio appello||2016-03-21|
|Disponibilità||Accessibile via web (tutti i file della tesi sono accessibili)|
Il tema della tesi consiste nello sviluppo di strategie globali per l'ottimizzazione delle performance in sistemi multi-robot. In particolare, si sono analizzati e sviluppati metodi per la coordinazione, l'ottimizzazione e la gestione del traffico di una flotta di automated guided vehicles (AGV) operanti in ambienti industriali.
This thesis deals with an ensemble strategy for optimizing the overall performance of multi-robot systems. Specifically, methodologies are presented for the coordination, optimization and traffic management of a fleet of automated guided vehicles (AGV) operating in industrial environments. This dissertation presents first an algorithm for the automatic creation of a roadmap for AGV systems which maximizes the connectivity, the redundancy and the coverage. The roadmap is built in such a way that the environment is filled by as roads as possible whose directions are assigned by maximizing the connectivity of the associated graph. A traffic coordinator is then developed in such a way that the optimal performance is guaranteed when it is used with a roadmap generated with the previous method. The coordination is performed by means of both decentralized and centralized control policies. The former is based on a priority scheme for the resource allocation, the latter is performed by modeling the coordination problem as a quadratic programming problem (QP) by aiming at minimizing the total time required for the fleet in order to accomplish its tasks. Along with the traffic coordinator, a hierarchical 2-layers control architecture is developed. The architecture exploits two layers to manage the problem. A layer is a topological graph representing the roadmap where each node is an area of the roadmap called sector. Then the other layer represents the actual roadmap within each sector. In this way, the overall scenario is modeled as a lumped parameter model where the traffic congestions and then the complexity are bounded in specific areas. Furthermore a probabilistic dynamic model of the traffic is introduced. The model is used to predict the evolution of the traffic in a future horizon. This information is then exploited by a traffic-based planner which coordinated the vehicles in order to minimize the total arrival time. The methodologies are supported by simulations and experiments in real world scenario, specifically in real automatic warehouses.