Studies on nonlinear dynamics hidden in an optokinetic reflex and brain inspired robot vision development

自2011年3月日本发生自然和人为灾害以来，在人类无法安全操作的危险情况下，机器人是有用的助手。在辅助机器人的构建中，开发具有与人脑视觉能力相当的机器人视觉（称为类人视觉）是迫切的研究任务。在类人机器人视觉的发展中，我们使用神经模型进行视动反射和机器人运动控制，以便对每个神经模型的发射率斜率进行分析。我们揭示了非线性动力学机制，如分叉过渡，这与可能显着相关的眼动控制的放电率的变化。此外，我们开发了一个双目机器人视觉设备，在其中将实现这样的神经模型。

Robots, which are helpful assistants in dangerous situations where humans cannot safely operate, have been required since natural and man-made disasters happened in Japan on March 2011. In the assistant robot construction, developments of the robot vision with the visual ability comparable to visual perceptions in the human brain (called humanoid vision) are the urgent research tasks. In the developments of human-like robot vision, we used neural models for an optokinetic reflex as well as robotic motion controls, in order to propose analysis on slopes of the firing rate for each neural model. We revealed nonlinear dynamical mechanisms such as bifurcation transitions, which were related to changes of the firing rate that might significantly be related to eye movement controls. In addition, we developed a binocular robot vision device in which such neural models will be implemented.