The effect of behavioral preferences on skill acquisition in determining unspecified, suitable action patterns to control humanoid robots

This research investigated the effect of behavioral preferences on learning efficiency when attempting to determine unspecified, but suitable action sequences for unfamiliar tasks. The goal of this research was to develop a skill acquisition support system for the elderly to aid them in using unfamiliar IT products, particularly those of welfare systems. Here, behavioral preference is defined as the type of action sequences that people would prefer to adopt for completing unfamiliar tasks. To achieve this goal, this research investigated the action sequences of participants when they attempt to control the posture of an unfamiliar humanoid robot with an unfamiliar controller. The participants were assigned the task of making the humanoid stand on one foot. Machine-learning-based methods were presented for analyzing the behavioral preferences. The analysis results indicate that participants having behavioral preferences of adopting random action sequences can complete the task in a much shorter time, compared to participants having a behavioral preference of adopting action sequences similar to those of previous actions.