Realization of Self-Organizing Brain Device and its Application to Autonomous Mobile Robot

为了实现自主机器人的人工脑功能，我们开发了自组织神经网络、SOM的SOM、粗糙赢者通吃神经网络和自适应自组织映射的FPGA实现。我们提出了改进的算法来实现这种有效的硬件实现。将自适应自组织地图硬件应用于自主机器人的视觉处理，实现了比微处理器实现快1万倍的速度。此外，我们还开发了一个基于硬件/软件复杂系统的自组织神经网络动态可重构平台，以方便地使用所开发的设备。

In order to realize an artificial brain function of autonomous robots, we developed FPGA implementations of Self-Organizing Neural Networks, SOM of SOMs, Rough-Winner-Take-All Neural Networks and Adaptive Self-Organizing Maps. We proposed modified algorithms for this effective hardware implementation. The Adaptive Self-organizing Maps hardware was applied to a visual processing of autonomous robot, and achieved 10,000 times faster than a micro-processor implementation. Furthermore, we developed a dynamically reconfigurable platform for self-organizing neural networks based on hardware/software complex system to use the developed device conveniently.