There are many research works on robotic devices to assist in movement training following neurologic injuries such as stroke with effects on upper limbs. Conventional neurorehabilitation appears to have little impact on spontaneous biological recovery, to this end robotic neurorehabilitation has the potential for a greater impact. Clinical evidence regarding the relative effectiveness of different types of robotic therapy controllers is limited, but there is initial evidence that some control strategies are more effective than others. This paper consider the contribution on a haptic training method based on kinesthetic guidance scheme with a non linear control law (proxy-based second order sliding mode control) with the human in the loop, and with purpose to guide a human user's movement to move a tool (pen in this case) along a predetermined smooth trajectory with finite time tracking, the task is a real maze. The path planning can compensate for the inertial dynamics of changes in direction, minimizing the consumed energy and increasing the manipulability of the haptic device with the human in the loop. The Phantom haptic device is used as experimental platform, and the experimental results demonstrate the effectiveness of this application.