Multiagent robotic systems

点击翻译按钮获取中文摘要

The planets beckon.  Robotics play and will continue to play a central role in planetary exploration.  They will precede us to Mars and will, in a very literal sense, pave the way for our return to the Moon.  Robots are currently our surrogate explorers but they will eventually become our companions as we begin to take steps farther into the Solar System.  Robot colonies will be required for resource utilization in situ and to construct habitats on alien soil.  But the robots must be robust, autonomous and "intelligent."  Herein lies the great challenge not just for space robotics but robotics in general.     The proposed research program is specifically aimed at multiagent robotic systems.  The work is very multidisciplinary, drawing on engineering, computer science, physics, biology and philosophy.  The goal of this project is to advance the understanding of the dynamics of multiagent systems and to develop biologically based (biomimetic) methods of control and self-organization in robotics.     In particular, the development of what we call an "artificial neural tissue" (ANT) is proposed.  We see this work as establishing a new paradigm for robotic control.  The objectives for the ANT architecture is to enable spatial differentiation observed in biological brains that accounts for specialization, to demonstrate the emergence of intelligence from interaction with the real world and to embody the concept and evaluate its performance on real robots.  Artificial evolutionary approaches and artificial developmental biology will characterize the tack taken in this work.  Cellular automata will also be used to study multiagent interactions and self-organization.  While the proposed research will target planetary-exploration applications, broader applications closer to Earth will also be considered.  Moreover, it is hoped that this work will also shed light on the biology of motor control processes in general that might in turn help us understand neuromotor diseases.  Indeed, in a related work, the modelling of cancer metastasis using cellular automata is proposed.