Learning Multirobot Hose Transportation and Deployment by Distributed Round-Robin Q-Learning

Multi-Agent Reinforcement Learning (MARL) algorithms face two main difficulties: the curse of dimensionality, and environment non-stationarity due to the independent learning processes carried out by the agents concurrently. In this paper we formalize and prove the convergence of a Distributed Round Robin Q-learning (D-RR-QL) algorithm for cooperative systems. The computational complexity of this algorithm increases linearly with the number of agents. Moreover, it eliminates environment non sta tionarity by carrying a round-robin scheduling of the action selection and execution. That this learning scheme allows the implementation of Modular State-Action Vetoes (MSAV) in cooperative multi-agent systems, which speeds up learning convergence in over-constrained systems by vetoing state-action pairs which lead to undesired termination states (UTS) in the relevant state-action subspace. Each agent's local state-action value function learning is an independent process, including the MSAV policies. Coordination of locally optimal policies to obtain the global optimal joint policy is achieved by a greedy selection procedure using message passing. We show that D-RR-QL improves over state-of-the-art approaches, such as Distributed Q-Learning, Team Q-Learning and Coordinated Reinforcement Learning in a paradigmatic Linked Multi-Component Robotic System (L-MCRS) control problem: the hose transportation task. L-MCRS are over-constrained systems with many UTS induced by the interaction of the passive linking element and the active mobile robots.

Performance of an Ocean Energy Conversion System with DFIG Sensorless Control

The 2009/28/EC Directive requires Member States of the European Union to adopt a National Action Plan for Renewable Energy. In this context, the Basque Energy Board, EVE, is committed to research activities such as the Mutriku Oscillating Water Column plant, OWC. This is an experimental facility whose concept consists of a turbine located in a pneumatic energy collection chamber and a doubly fed induction generator that converts energy extracted by the turbine into a form that can be returned to the network. The turbo-generator control requires a precise knowledge of system parameters and of the rotor angular velocity in particular. Thus, to remove the rotor speed sensor implies a simplification of the hardware that is always convenient in rough working conditions. In this particular case, a Luenberger based observer is considered and the effectiveness of the proposed control is shown by numerical simulations. Comparing these results with those obtained using a traditional speed sensor, it is shown that the proposed solution provides better performance since it increases power extraction in the sense that it allows a more reliable and robust performance of the plant, which is even more relevant in a hostile environment as the ocean.

An adaptive sliding mode control law for the power maximization of the wind turbine system

POWERENG 2011

A Real-Time Sliding Mode Control for a Wind Energy System Based on a Doubly Fed Induction Generator

In this paper, a real time sliding mode control scheme for a variable speed wind turbine that incorporates a doubly feed induction generator is described. In this design, the so-called vector control theory is applied, in order to simplify the system electrical equations. The proposed control scheme involves a low computational cost and therefore can be implemented in real-time applications using a low cost Digital Signal Processor (DSP). The stability analysis of the proposed sliding mode controller under disturbances and parameter uncertainties is provided using the Lyapunov stability theory. A new experimental platform has been designed and constructed in order to analyze the real-time performance of the proposed controller in a real system. Finally, the experimental validation carried out in the experimental platform shows; on the one hand that the proposed controller provides high-performance dynamic characteristics, and on the other hand that this scheme is robust with respect to the uncertainties that usually appear in the real systems.