Abstract
At present state, almost the robot tasks are performed by the teaching playback method, in which the robot repeats the motion taught in advance using a teaching pendant, etc. This method is based on the comparatively high repeatability of the robot arm, thus the environment is required to be strictly constant for all the trials. On the other hand, a human can skillfully achieve the task such as assembly with adapting to the environmental change. In this process, he uses the sensor feedback, i.e., he adjusts the motion based on the sensor information. In this research, the pin insertion task to a hole without chamfering and with small clearance is focused on. This task is difficult for a robot by only applying the teaching playback method, so the force sensor feedback is required. The problem here is that this task is difficult even by employing a force sensor, since the way of insertion based on the force information is not established. In this research, the pin insertion task is aimed to be achieved by imitating a human motion, extracting and defining the rule using membership functions, and executing the task based on fuzzy control.
