Learning Strategy in Time-to-Contact Estimation of Falling Objects
Learning Strategy in Time-to-Contact Estimation of Falling Objects
复制标题
DOI：
10.20965/jaciii.2011.p0972
复制
发表时间：
2011-10-01
影响因子：
0.7
通讯作者：
Koike, Yasuharu
中科院分区：
文献类型：
--
作者：
Kambara, Hiroyuki;Ohishi, Keiichi;Koike, Yasuharu
在日常生活中，估计物体接触前的剩余时间（接触时间或TTC）的能力至关重要。在本文中，我们研究了中枢神经系统（CNS）如何能够估计一个球在不同的加速度下降的TTC。根据对人类接球能力的实验，我们假设CNS可以容纳多个TTC估计器，每个TTC估计器针对不同的加速度进行训练，并且其中一个用于接球试验中的TTC估计。在这里，我们做了一个假设，当存在估计误差时，如何训练每个TTC估计器。(1)如果估计误差较小，则重新校准试验中采用的TTC估计器。(2)如果估计误差较大，则创建新的TTC估计器。为了验证这一假设，我们在虚拟环境中进行了两种类型的接球实验，在每个实验中，虚拟球的加速度逐渐或突然改变。两个实验中捕捉表现的差异支持了我们的假设。
The ability to estimate the time that remains before contact (Time-To-Contact or TTC) of a falling object is critical in daily life. In this paper, we investigated how the Central Nervous System (CNS) becomes able to estimate the TTC of a ball falling at various accelerations. According to experiments on the human ability to catch a ball falling at various accelerations, we assumed that the CNS can hold multiple TTC estimators each of which is trained for a different acceleration, and one of them is adopted for TTC estimation in a ball-catching trial. Here we made a hypothesis about how each TTC estimator is trained when there is an estimation error. (1) If the estimation error is small, the TTC estimator adopted in the trial is recalibrated. (2) If the estimation error is large, a new TTC estimator is created. To test this hypothesis, we conducted two types of ball-catching experiments in a virtual environment where the acceleration of a virtual ball is changed gradually or suddenly in each experiment. The difference in catching performances in the two experiments supported our hypothesis.