Modeling of function of biological internal fluctuation and its application to adaptive system

本研究提出的“人工生物粒子”自适应方案示意图如图1所示。这里假设粒子有一个概率的内部波动源，该波动源通过执行器控制粒子的运动。该粒子的行为类似于受内部涨落运动的布朗粒子。将外场的检测信息作为自发波动源的输入，以产生运动的控制信号。即使输入是固定的，由于波动源的概率性质，输出也会波动。通过波动源的学习能力，逐步建立给定环境场的期望输入输出概率关系；1)评估控制器所指示的运动的适应性，并反馈给波动源；Ii)如果控制产生对外部场的自适应运动，则识别波动源，从而更频繁地产生控制。波动模式越丰富，粒子的适应能力越强。

The adaption scheme of " artificial biological particle" we proposed in this research is schematically shown in Fig.1. Here the particle is assumed to have a probabilistic internal source of fluctuation which controls the motion of the particle through actuator. The particle behaves like a Brownian particle moved by internal fluctuation. The sensored information of external field is used as an input to the source of spontaneous fluctuation in order to generate the control signal for the motion. Even if the input is fixed the output fluctuates because of the probabilistic property of the fluctuation source. The desired input-output probabilistic relation for given environmental field is gradually built by trial and error through the learning ability of the fluctuation source;i) the adaptability of the motion indicated by a control is evaluated and feedback to the fluctuation source; ii) if a control gives rise to an adapted motion to external field, then the fluctuation source is recognized so as to generate the control more frequently. The more abundant the patterns of fluctuation are, the more flexible the adaptation of the particle are.