Statistical Characteristics of Biomimetic Image-Based Inverted Pendulum Control Systems Using Just-In-Time Method

摘要

In attempting to statistically answer the question, "In what ways should robots attempt or imitate humanlike behavior?" We focus on a human stick-balancing task and a biomimetic visual servoing system for the stabilization of an inverted pendulum. In particular, we applied proportional and differential (PD) and Just-In-Time (JIT) control, which is a memory-based scheme. This method can stabilize the inverted pendulum for a certain period of time, similar to the case of humans. We analyzed the behavior of such mechanical systems statistically to determine the on-off intermittency and the power law in the human stick-balancing task. Statistically analyzing pendulum behavior may enable us to interpret human motor control.