The need for a common taxonomy and benchmarks to achieve “ human-like ” performance in bipedal robot locomotion

A common taxonomy for the term “human-like locomotion” is essential to enhance the progress in the field of bipedal robot locomotion. In literature widespread use of this term can be found implying that human motion is optimal and worthy of imitation. However, a common basic understanding of the fundamental principles and characteristics of human locomotion is yet to be completed. In this talk we review various interpretations of this term in the literature and elaborate briefly on the most relevant characteristics of human motion trajectories. Further, essential methods from modeling and simulation to locomotion performance evaluation are discussed. We present a possible definition for “human-like locomotion” and a general concept for a better comparability of human and robot locomotion performance. The expressed ideas are supported by interim results obtained within the BioBiped project. I. THE IMPORTANCE OF A COMMON TAXONOMY FOR THE TERM “HUMAN-LIKE LOCOMOTION” Achieving bipedal robot locomotion performance that comes close to that of the biological counterpart is still a major and challenging research topic in the field of humanoid robotics which is not only due to the complexity of the problem requiring the know-how from various disciplines. Rather, a common taxonomy for the term “human-like locomotion” is missing. Today we have a wealth of different concepts for the actuation and motion control of humanoid robots and different views for the locomotion performance evaluation without a generally accepted understanding of the neuromechanics and core functionality underlying human locomotion. The overwhelming majority of authors take a general but not rigorously specified understanding of “human-like locomotion” for granted, but in fact it is associated with very individual interpretations and definitions. A common taxonomy is expected to enormously enhance the progress in the field of bipedal robot locomotion and allow a better and valid comparability of results among the different research groups. II. LITERATURE REVIEW The overwhelming majority of authors take a general understanding of “human-like locomotion” for granted and do not even attempt to explain the use of this term in their papers. Some exemplary uses are as follows. In [1] snapshots of the robot motion sequence are presented and the robot This work was supported by the German Research Foundation (DFG) under grant no. STR 533/7-1. The authors are with the Simulation, Systems Optimization and Robotics Group, Technische Universität Darmstadt, 64289 Darmstadt, Germany. E-mail: {radkhah,stryk}@sim.tu-darmstadt.de model’s and author’s step frequency are compared to assess the human-likeness. Visual comparison of snapshots is a popular tool also in [2], accompanied by statements such as: “The periodic running motion looks very natural, (compare the corresponding animation at our website)”. Also, biomechanical gait characteristics such as duty factor, step length, ground reaction forces (GRF) and the vertical center of mass (COM) trajectories are computed and compared to those known from a human. In addition to the GRF and COM trajectories, sagittal joint angle histories obtained from the simulation and real robot model are compared to human joint kinematics in [3]. Selfstabilization of a gait is seen as a further important property of human locomotion. Also, walking with stretched knees and heel-contact and toe-off are considered as important characteristics of human walking [4]. As proof of concept it is shown that the GRF of the robot agree very well in pattern and peaks with those collected from a human subject. In [5] the well known cost of transport is introduced as a measure to rate the degree of human-likeness. A robot walking at low cost of transport, like the Cornell biped, is regarded as quite “human-like”. For the authors of [6], a characteristic of “natural humanlike motion” in general is the exploitation of the passive body dynamics. These various interpretations suggest that a generally accepted catalog of all relevant characteristics of human locomotion and their measurement tools is an essential requirement for progress in the field of bipedal robot lo-