Abstract
In this paper, robustness condition for robot trajectory against environmental variation and effectiveness of direct teaching are discussed based on the parameter space expression, through a case study on a peg-in-hole task. In the peg-in-hole experiment, hole position is artificially varied and robustness of two different assembly strategies are compared based on the degree of achievement and success rate. One strategy is generated by direct teaching and is similar to that human usually adopts, while the other is generated by programming, which is similar to conventional methods such as using teaching pendant. Experimental results show that the human-like trajectory is more robust against variation of position where appropriate manipulator compliance effectively absorbs errors. This indicates the effectiveness of direct teaching because such human-like trajectory is difficult to be taught through conventional methods.
