Naslov
Decentralizirano upravljanje automatski vođenim vozilima sa slobodnim odabirom putanja u industrijskim okruženjima
(Decentralized control of free ranging automated guided vehicles in industrial environments)

Autori
Draganjac, Ivica

Vrsta, podvrsta i kategorija rada
Ocjenski radovi, doktorska disertacija

Fakultet
Fakultet elektrotehnike i računarstva

Mjesto
Zagreb

Datum
02.02

Godina
2018

Stranica
140

Mentor
Zdenko Kovacic

Ključne riječi
automated guided vehicles, decentralized control, multi-robot systems, algorithm design and analysis, path planning, motion coordination, trac management, con ict resolution, distributed resource allocation, material handling systems, automated warehousing

Sažetak
Systems with automated guided vehicles (AGVs) play an essential role in material handling processes within various manufacturing facilities, automatic warehouses and distribution centers. However, despite their long presence on the factory oor and signi cant advances they experienced over the last decades, the current industrial state of the art still relies on a centralized controller, which dispatches transport missions to vehicles along a predetermined network of paths. While this approach makes it easier to ensure correct system operation, it su ers from several drawbacks: the paths have to be laid out in advance, which is time-consuming and can be suboptimal ; centralized planning does not scale well as the number of vehicles increases ; and it introduces a single point of failure into the system. In order to eliminate the drawbacks of the centralized control architecture, this PhD thesis introduces a decentralized trac management algorithm that ensures high scalability and exibility of AGV systems by providing vehicles with full autonomy in the mission execution process. By executing on each vehicle in the system, the proposed control algorithm allows vehicles to autonomously plan their own paths to complete assigned missions and negotiate with other vehicles in the local area for right of way in order to avoid collisions and deadlocks. The algorithm relies on a two-layer control approach with topological representation of the working environment on the top layer and statelattice representation on the bottom layer. The path planning part of the algorithm implements a free-ranging motion scheme by determining the shortest feasible paths through di erent working space sectors, considering non-holonomic vehicle constraints. The motion coordination part of the algorithm ensures safe vehicle motions by reliable detection and resolution of di erent con ict situations with other vehicles in the shared workspace. Con ict resolution is based on a vehicle priority scheme and results in temporary stopping or removal of the lower priority vehicles taking part in the con ict. Removal actions are always performed within the vehicle's private zone, i.e. the dynamically allocated local region of the workspace surrounding the vehicle. By encoding information on the vehicle size and its kinematic constraints, the introduced private zone mechanism provides the necessary physical space required for successful execution of every removal action. Algorithm correctness, and especially its deadlock-free and livelock-free properties, has been formally analyzed and proved, while its performance has been validated in simulation with up to fty virtual vehicles, and also experimentally on two di erent setups - a laboratory setup comprising six Pioneer 3DX vehicles and by two state of the art autonomous forklifts in industrial-like operating conditions. Validation results con rm correct system operation and successful implementation of several advancements over the current state of the art, primarily in terms of high system scalability, increased exibility to changing requirements, and high tolerance to fault conditions.

Izvorni jezik
Engleski

Znanstvena područja
Elektrotehnika, Temeljne tehničke znanosti

POVEZANOST RADA