These days, many rehabilitation robots that support patients physically or mentally are being proposed. The robots that offer physical support are used as treatment to improve the physical condition such as weakness or lack of muscles by estimating and changing dynamic control depending on the muscle status. These robots are successfully supporting the patient. However, control of the robot based on the patient's emotion such as depression and stress is more difficult than to support the patient physically. One of the problems comes from the lack of the clear relation between accurate detection of emotion such as depression and stress, and successful control of the robot. To solve this problem, we firstly try to measure the relation between the control of the rehabilitation robot and the dynamic change in the patient's emotion. We propose a method to control the rehabilitation robot based on the user's mental status. We evaluated the emotions of the person operating the rehabilitation robot using an index to estimate their mental state from heart rate variability (HRV). HRV is used to measure the valence, an index of pleasure. We conducted an experiment in which the participants were directed to control the rehabilitation robot with three different levels of pushing force required to operate the robot. From the results, we confirmed that the emotion of the participants had significantly changed depending on the load applied to the robot.