Studying human motor control requires innovative engineering solutions including robots. For example, the motor control task(s) must be reliable before using system identification (SYSID) tools to estimate any motor control descriptors, e.g. neurophysiological parameters. Robots can improve the experiment reliability in contrast to passive devices, e.g. seated balance on a hemisphere. First, we present some novel physical human-robot interaction (pHRI) tasks in the biomechanics field. Then, we show how one pHRI can achieve excellent reliability measures. Moreover, robots can allow the use a wide variety of input/perturbation signals that may be designed specifically for better SYSID results, such as better estimation error variance. With standard signals, however, further analysis is required to improve the SYSID outcomes. Therefore, we devised an analysis method based on Fisher information to reduce the estimation error variance.