Experience generalization for multi-agent reinforcement learning

On-line learning methods have been applied successfully in multi-agent systems to achieve coordination among agents. Learning in multi-agent systems implies in a non-stationary scenario perceived by the agents, since the behavior of other agents may change as they simultaneously learn how to improve their actions. Non-stationary scenarios can be modeled as Markov Games, which can be solved using the Minimax-Q algorithm a combination of Q-learning (a Reinforcement Learning (RL) algorithm which directly learns an optimal control policy) and the Minimax algorithm. However, finding optimal control policies using any RL algorithm (Q-learning and Minimax-Q included) can be very time consuming. Trying to improve the learning time of Q-learning, we considered the QS-algorithm. in which a single experience can update more than a single action value by using a spreading function. In this paper, we contribute a Minimax-QS algorithm which combines the Minimax-Q algorithm and the QS-algorithm. We conduct a series of empirical evaluation of the algorithm in a simplified simulator of the soccer domain. We show that even using a very simple domain-dependent spreading function, the performance of the learning algorithm can be improved.

Multi-Agent Framework for Spatial Load Forecasting

A multi-agent framework for spatial electric load forecasting, especially suited to simulate the different dynamics involved on distribution systems, is presented. The service zone is divided into several sub-zones, each subzone is considered as an independent agent identified with a corresponding load level, and their relationships with the neighbor zones are represented as development probabilities. With this setting, different kind of agents can be developed to simulate the growth pattern of the loads in distribution systems. This paper presents two different kinds of agents to simulate different situations, presenting some promissory results.

A multi-agent system with a percolation approach to simulate the driving pattern of Plug-In ELectric Vehicles

A multi-agent system with a percolation approach to simulate the driving pattern of Plug-In Electric Vehicle (PEV), especially suited to simulate the PEVs behavior on any distribution systems, is presented. This tool intends to complement information about the driving patterns database on systems where that kind of information is not available. So, this paper aims to provide a framework that is able to work with any kind of technology and load generated of PEVs. The service zone is divided into several sub-zones, each subzone is considered as an independent agent identified with corresponding load level, and their relationships with the neighboring zones are represented as network probabilities. A percolation approach is used to characterize the autonomy of the battery of the PVEs to move through the city. The methodology is tested with data from a mid-size city real distribution system. The result shows the sub-area where the battery of PEVs will need to be recharge and gives the planners of distribution systems the necessary input for a medium to long term network planning in a smart grid environment. © 2012 IEEE.

Planejamento e despacho ótimo de reativos (VAr) em sistemas de energia elétrica multi-áreas

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)

Rank Aggregation for Pattern Classifier Selection in Remote Sensing Images

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)

Teaching real-time with a scheduler simulator

In this paper we describe a scheduler simulator for real-time tasks, RTsim, that can be used as a tool to teach real-time scheduling algorithms. It simulates a variety of preprogrammed scheduling policies for single and multi-processor systems and simple algorithm variants introduced by its user. Using RTsim students can conduct experiments that will allow them to understand the effects of each policy given different load conditions and learn which policy is better for different workloads. We show how to use RTsim as a learning tool and the results achieved with its application on the Real-Time Systems course taught at the B.Sc. on Computer Science at Paulista State University - Unesp - at Rio Preto.

Spatial electric load forecasting using a local movement approach

An agent based model for spatial electric load forecasting using a local movement approach for the spatiotemporal allocation of the new loads in the service zone is presented. The density of electrical load for each of the major consumer classes in each sub-zone is used as the current state of the agents. The spatial growth is simulated with a walking agent who starts his path in one of the activity centers of the city and goes to the limits of the city following a radial path depending on the different load levels. A series of update rules are established to simulate the S growth behavior and the complementarity between classes. The results are presented in future load density maps. The tests in a real system from a mid-size city show a high rate of success when compared with other techniques. The most important features of this methodology are the need for few data and the simplicity of the algorithm, allowing for future scalability. © 2009 IEEE.

Spatial load forecasting using a demand propagation approach

A method for spatial electric load forecasting using multi-agent systems, especially suited to simulate the local effect of special loads in distribution systems is presented. The method based on multi-agent systems uses two kinds of agents: reactive and proactive. The reactive agents represent each sub-zone in the service zone, characterizing each one with their corresponding load level, represented in a real number, and their relationships with other sub-zones represented in development probabilities. The proactive agent carry the new load expected to be allocated because of the new special load, this agent distribute the new load in a propagation pattern. The results are presented with maps of future expected load levels in the service zone. The method is tested with data from a mid-size city real distribution system, simulating the effect of a load with attraction and repulsion attributes. The method presents good results and performance. © 2011 IEEE.

A Markov random field model for combining optimum-path forest classifiers using decision graphs and game strategy approach

The research on multiple classifiers systems includes the creation of an ensemble of classifiers and the proper combination of the decisions. In order to combine the decisions given by classifiers, methods related to fixed rules and decision templates are often used. Therefore, the influence and relationship between classifier decisions are often not considered in the combination schemes. In this paper we propose a framework to combine classifiers using a decision graph under a random field model and a game strategy approach to obtain the final decision. The results of combining Optimum-Path Forest (OPF) classifiers using the proposed model are reported, obtaining good performance in experiments using simulated and real data sets. The results encourage the combination of OPF ensembles and the framework to design multiple classifier systems. © 2011 Springer-Verlag.

Transaction scheduling using dynamic conflict avoidance

Software transaction memory (STM) systems have been used as an approach to improve performance, by allowing the concurrent execution of atomic blocks. However, under high-contention workloads, STM-based systems can considerably degrade performance, as transaction conflict rate increases. Contention management policies have been used as a way to select which transaction to abort when a conflict occurs. In general, contention managers are not capable of avoiding conflicts, as they can only select which transaction to abort and the moment it should restart. Since contention managers act only after a conflict is detected, it becomes harder to effectively increase transaction throughput. More proactive approaches have emerged, aiming at predicting when a transaction is likely to abort, postponing its execution. Nevertheless, most of the proposed proactive techniques are limited, as they do not replace the doomed transaction by another or, when they do, they rely on the operating system for that, having little or no control on which transaction to run. This article proposes LUTS, a lightweight user-level transaction scheduler. Unlike other techniques, LUTS provides the means for selecting another transaction to run in parallel, thus improving system throughput. We discuss LUTS design and propose a dynamic conflict-avoidance heuristic built around its scheduling capabilities. Experimental results, conducted with the STAMP and STMBench7 benchmark suites, running on TinySTM and SwissTM, show how our conflict-avoidance heuristic can effectively improve STM performance on high contention applications. © 2012 Springer Science+Business Media, LLC.

An approach to implement data fusion techniques in wireless sensor networks using genetic machine learning algorithms

Wireless Sensor Networks (WSNs) can be used to monitor hazardous and inaccessible areas. In these situations, the power supply (e.g. battery) of each node cannot be easily replaced. One solution to deal with the limited capacity of current power supplies is to deploy a large number of sensor nodes, since the lifetime and dependability of the network will increase through cooperation among nodes. Applications on WSN may also have other concerns, such as meeting temporal deadlines on message transmissions and maximizing the quality of information. Data fusion is a well-known technique that can be useful for the enhancement of data quality and for the maximization of WSN lifetime. In this paper, we propose an approach that allows the implementation of parallel data fusion techniques in IEEE 802.15.4 networks. One of the main advantages of the proposed approach is that it enables a trade-off between different user-defined metrics through the use of a genetic machine learning algorithm. Simulations and field experiments performed in different communication scenarios highlight significant improvements when compared with, for instance, the Gur Game approach or the implementation of conventional periodic communication techniques over IEEE 802.15.4 networks. © 2013 Elsevier B.V. All rights reserved.

Aplicações de sistemas multiagentes na previsão espacial de demanda elétrica em sistemas de distribuição

Pós-graduação em Engenharia Elétrica - FEIS

Uma arquitetura multiagentes para sistema educacional baseado na web semântica

Pós-graduação em Ciência da Computação - IBILCE

An evolutionary approach to improve connectivity prediction in mobile wireless sensor networks

Connectivity is the basic factor for the proper operation of any wireless network. In a mobile wireless sensor network it is a challenge for applications and protocols to deal with connectivity problems, as links might get up and down frequently. In these scenarios, having knowledge of the node remaining connectivity time could both improve the performance of the protocols (e.g. handoff mechanisms) and save possible scarce nodes resources (CPU, bandwidth, and energy) by preventing unfruitful transmissions. The current paper provides a solution called Genetic Machine Learning Algorithm (GMLA) to forecast the remainder connectivity time in mobile environments. It consists in combining Classifier Systems with a Markov chain model of the RF link quality. The main advantage of using an evolutionary approach is that the Markov model parameters can be discovered on-the-fly, making it possible to cope with unknown environments and mobility patterns. Simulation results show that the proposal is a very suitable solution, as it overcomes the performance obtained by similar approaches.

Arquitetura de um sistema multiagente autônomo para supervisão e controle

This paper presents a multi-agent architecture that was designed to develop processes supervision and control systems, with the main objective to automate tasks that are repetitive and stressful, and error prone when performed by humans. A set of agents were identified, based on the study of a number of applications found in the literature, that use the approach of multi-agent systems for data integration and process monitoring to faults detection and diagnosis, these agents are used as basis of the proposed multi-agent architecture. A prototype system for the analysis of abnormalities during oil wells drilling was developed.