2012年5月29日（火） 18:00-20:00 吹田キャンパス 工学研究科 M4-201にて第47回創成塾を開催致します．
18:00-18:30 Stefano Ghidoni (on Skype)
18:50-19:10 Eduardo Castello
Talk 1: Stefano Ghidoni （University of Padova, Italy）
“Intelligent environments with distributed, multimodal sensing”
In recent years, the rise of camera networks have substantially improved video surveillance systems, increasing their performance and letting them perform higher-level tasks. This is achieved thanks to the high number of sensors available in such infrastructures, to the availability of cost-effective embedded hardware that enables local pre-processing, and to the large bandwidth of the connections among the nodes. In this talk, some applications developed at the Intelligent Autonomous Systems (IAS) lab of the University of Padova, Italy, will be described, which address fields like ambient intelligence for home automation and security applications. Such systems are capable of exploiting not only a number of cameras, but also audio sensors, in order to perform event classification based on multi-sensorial data. Results obtained with a small camera and microphone network installed inside the laboratory will be presented, and some insights about future research directions, exploiting camera networks for 3D reconstruction, will be provided.
Talk 2: Eduardo Castello (Engineering Science Faculty, Osaka University）
“Dynamic Task Allocation for a Multi-Robot System based on the Attractor Selection Model.”
Biological systems are often composed of many well-organized elements, examples including swarm behavior or cell differentiation in the developing process of animals. However, developing a multi-robot system with similar functions, that could be flexible and adapt to environmental changes, is increasingly complex. Recently, it has been suggested that the fluctuations within a biological system play an important role in adapting to change. It is the intention of this research to propose a new control method for the dynamic task allocation of multi-robot cooperation as inspired by the adaptability of a biological system. Also, this research aims to apply the proposed method in a cleaning task simulation within a dynamic environment. Based on results of experiments, we conclude that the proposed method is able to efficiently adapt to changing environments and robot failures.