The safety and dynamic stability of the two-wheeled self-balancing personal mobility (SBPM) are critical for their operation, and hence these issues need to be evaluated objectively. But due to the nature of the SBPM, they must be operated by the human rider even for the evaluation tests, so the objectivity and repeatability of the test cannot be ensured, and there even could be the safety issue for the test rider. To solve this problem, we propose an anthropomorphic test dummy robot (ATDR) that reproduces the human rider’s dynamic behavior using the control moment gyroscope (CMG) so that it could drive the SBPM instead of a human rider for the tests. As the human rider controls the SBPM by the disturbance torque made by the posture (i.e. the center of gravity) change, CMG-based ATDR reproduces the dynamic behavior of the SBPM by creating the same disturbance torque using the CMG. Design and manufacturing details of the CMG-based ATDR are introduced, and the control algorithm ensuring that CMG-based ATDR reproduces the same behavior as the human rider is explained. For the validation of the proposed CMG-based ATDR, manufactured CMG-based ATDR was controlled to drive the SBPM in the most common driving scenarios such as acceleration/deceleration ride and turning motion, and its dynamic response data was compared with that obtained from the human rider case. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —This paper was motivated by the objectivity issue of the safety certification tests for the SBPM. Existing approaches to the safety certification process requires tests run by human operators, which essentially leads to the question of objectivity in the test results. This paper suggests a new approach using an ATDR that mimics the human operator’s motion intent and eliminates the need of the human operator. In this paper, we present the working principle how ATDR regenerates the human operator’s motion and the detailed design of the ATDR to implement this goal. We then show the control algorithm designed to drive the SBPM as if it was driven by a human operator using the ATDR. Experiments carried out for the major hazardous driving scenarios required for the safety certification suggest that the ATDR could be used as a viable device for the objective safety certification process of the SBPM. In future research, we will expand the application of the ATDR for different types of SBPMs.