Neural-Symbolic Methods for Learning and Transforming Embodied Knowledge in Heterogeneous Robotics

• 作 者: Min Jiang
• 机 构: XiamenUniversity,China
• 出 处: 第三届先进计算智能与智能信息学国际研讨会(The Third International Workshop on Adv
• 发表日期: 2013年10期

　　Effective transfer of embodied knowledge among heterogeneous robots would have significant theoretical value and potentially dramatic practical value as well. It could enable robots to rapidly master skills and adapt to the new surroundings with relatively low computational cost. However, the path to achieving this goal is obstructed by a number of apparent difficulties, related to computational resource limitations, the uncertainty of information acquisition, and the omnipresence of ambient noise. Our research sees to surmount these obstacles via a unique approach to neural-symbolic integration. Our research will closely link deep learning and transfer learning and will utilize the symbolic reasoning power and nonlinear dynamical attention allocation of the OpenCog integrative Artificial General Intelligence architecture. While the formalism, ideas and software developed will largely be generally applicable, the initial target application area will be hand-eye coordination, a topic of significant importance unto itself. The results will be applied to heterogeneous robots carrying out handeye coordination in the real world, giving them unprecedented capability for knowledge transformation and associated rapid learning.