Foraging and reproduction based on memory and learning in flys : foraging trajectory and chaotic attractor shift

最优觅食理论以生物理性行为为前提，完整地估计当前和未来的收益，并应用与现实相去甚远的最优算法来计算觅食策略。因此，行为生态学家想要重建觅食理论，考虑到现实生物对记忆和学习能力的现实限制。在本研究项目中，我们对果蝇的行走轨迹进行了行为学实验。我们希望通过分析黑腹蛇觅食轨迹的模式来揭示觅食效率的终极因素(例如，生物适合性)。首先，我们发现竞走遵循混合模型的估计，介于基于Levy-Flight行走的“功率分布”和由随机行走得出的“指数分布”之间，并且步行呈现出方向偏向的趋势。这一结果可以解释为自然选择促进了列维-飞行行走，提高了觅食效率的组合行为，以及对现实有机体的体现约束。此外，基于局部平稳自回归模型的时间序列分析表明，在保持自回归模型不变的情况下，移动速度在一段时间内保持不变，而折算转折度不保持，这表明移动速度是自然选择。因此，通过物理和统计方法排序的本项目分析了决策过程，并显示了超出当前行为生态学的独特结果。

Optimal foraging theory has a premise that organisms behave rationally estimating current and future profits completely and calculates the foraging strategy applying the optimal algorism that are far from the realism. Therefore, behavioral ecologists would like to re-build the foraging theory, taking into account the realistic constrain on memory and learning ability of actual organisms. In the present research project, we conducted the behavior experiment on walking trajectory in the fruit fly, Drosophila melanogasuter. We expected to reveal the ultimate factor (e.g., biological fitness) in the foraging efficiency, analyzing the pattern in the diet-searching trajectory of D. melanogaster. Firstly, we discovered that a walking bout follows estimation of the mixed model that are in between "power distribution" based on Levy-flight walking and "exponential distribution" inferred from random walking, and that the walking showed a directionally-biased tendency. This results can be explained as combined behaviors by natural selection which promoted Levy-flight walking increasing the foraging efficiency and by embodiment constraint on a realistic organism. Furthermore, conducting the time-series analysis based on locally stationary auto-regression model, we showed that the movement speed was kept for a certain period with maintaining the same auto-regression model and that the conversion turn degree was not, suggesting a natural selection on the movement speed. Thus, the present project that was ordered by physical and statistical methods analyzed decision-making processes, and it showed unique results beyond the current behavioral ecology.