Many robots are used in industrial field. In many control algorithms for these robot, PID control algorithm is most widely used, but PID control algorithm has some difficulties for determining its gains properly. And the algorithms based on model has difficulties in obtaining precise model and accurate values of its parameters. There are other algorithms which do not require a model of the system. If a task is periodic, then control input is also periodic. So control input can be approximated by linear combinations of periodic functions. The periodic functions which are fit for approximating a control input should be selected. Generally the Fourier functions are used most widely. However they are continuous, it is hard to approximate a control input which should be changed abruptly to overcome the friction force which dominates in the control input. To overcome this difficulty, the Walsh function is used in this thesis. The Walsh function has a good property in approximating the system's friction by its discontinuous nature. By simulation and implementation to a robot that has much friction, it is shown that the proposed method is better than Fourier function method.