Decision-theoretic Planning for Active Perception and Smart Energy
Grids
na próxima terça-feira o professor Matthijs Spaan, da Delft University, referência na área de aprendizado por máquina, irá proferir uma palestra que promete ser bem interessante na interseção entre redes de energia e ciência da computação. Trata-se do uso de inteligência artificial para determinar como realizar balanceamento de carga. A palestra será acessível a quem não tem conhecimentos específicos em qualquer uma das duas áreas abordadas.
Após a palestra, o professor Matthijs estará disponível para trocar ideias.
Título:
Decision-theoretic Planning for Active Perception and Smart Energy
Grids
Data e horário: 11 de agosto de 2015, 1.30 pm
Local:
sala DLC/corredor do Departamento de Ciência da Computação/CCMN/UFRJ
(prédio do NCE, virar a esquerda sem passar as catracas)
Abstract:
Decision making is an important skill of autonomous agents, but, in
many real-world systems, this task is complicated by uncertainty about
the effects of actions and limited sensing capabilities. In
particular, we will be concerned with planning problems that optimize
how an agent should act given a model of its environment and its
task. In this talk, I will give an overview of our recent work on
decision-theoretic planning under uncertainty, formalized as Partially
Observable Markov Decision Processes (POMDPs).
First, I will introduce the POMDP framework and related algorithms.
Next, I will present a modeling framework for active perception based
on POMDPs. A key feature is that we remain in the standard POMDP
setting while balancing information-gain and task-specific objectives.
Finally, I will present recent results on load scheduling and
congestion management in smart energy grids. Here we exploit a
scenario representation to account for hard-to-model external factors
such as renewable energy supply.
Bio:
Dr. Matthijs Spaan is an Assistant Professor at the Algorithmics
group, Delft University of Technology, Delft, The Netherlands. He has
received a PhD degree in CS (2006) and an MSc degree in AI (2002) from
the University of Amsterdam, The Netherlands. His scientific
interests are in planning under uncertainty, sequential decision
making, cooperative multiagent/multi-robot systems, smart energy
grids, (decentralized) partially observable Markov decision processes,
reinforcement learning, machine learning and artificial intelligence
in general.
www.st.ewi.tudelft.nl/~mtjspaan