Robust system design for consensus control in dynamically positioned vessel fleet.

• Ano de defesa: 2019
• Autor(a) principal: Ianagui, André Seiji Sandes
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Controle cooperativo; Controle robusto não-linear; Cooperative control; Dynamic positioning systems; Embarcações; Robust nonlinear control; Sistema de posicionamento dinâmico
• Link de acesso: http://www.teses.usp.br/teses/disponiveis/3/3152/tde-07012020-155948/

As offshore operations grow larger and more complex, the requirements for a larger number of agents - from vessels to equipment - working simultaneously and cooperatively become evident. In this scenario, the ability to perform such tasks safely coordinating all these elements will eventually reach human limits, bounding also the complexity that can be achieved. Enhanced levels of autonomy emerge as a response to these requirements, giving margin to safer, larger and possibly more cost-effective operations. This work presents the development of cooperative control algorithms applied in Dynamic Positioned (DP) vessels. In this approach, DP vessels are treated as \"Drone ships\", which can not only perform station keeping but also trajectory tracking tasks collectively. The proposed methods encompass the use of robust nonlinear techniques designed in the form of cooperative protocols, such as Sliding Mode Control (SMC) and Super Twisting Algorithm (STA), to solve a multi-agent consensus problem. Such methods need to account for effects from geometric and dynamic nonlinearities and for environmental disturbances that affect DP systems. By considering a fixed network topology, but prone to communication breakdowns, the use of a parallel Kalman Filter for neighbor agents\' states estimation is proposed to enhance the system robustness. Finally, a consensus filter is considered for the improvement of measurements and estimations of environmental data, benefiting from the multiple vessels in the network. All methods were validated through a set of study cases, tested in simulation and in small scale models.