International Journal of Humanoid Robotics

Falls are undesirable in humanoid robots, but also inevitable, especially as robots get deployed in physically interactive human environments. We consider the problem of fall prediction: to predict if the balance controller of a robot can prevent a fall from the robot's current state. A trigger from the fall predictor is used to switch the robot from a balance maintenance mode to a fall control mode. It is desirable for the fall predictor to signal imminent falls with sufficient lead time before the actual fall, while minimizing false alarms. Analytical techniques and intuitive rules fail to satisfy these competing objectives on a large robot that is subjected to strong disturbances and exhibits complex dynamics. We contribute a novel approach to engineer fall data such that existing supervised learning methods can be exploited to achieve reliable prediction. Our method provides parameters to control the tradeoff between the false positive rate and the lead time. Several combinations of parameters yield solutions that improve both the false positive rate and the lead time of hand-coded solutions. Learned solutions are decision lists with typical depths of 5–10, in a 16-dimensional feature space. Experiments are carried out in simulation on an ASIMO-like robot.

This paper, describes the development of a motion capture system with novel features for biped robots. In general, motion capture is effectively utilized in the field of computer animation. In the field of humanoid robotics, the number of studies attempting to design human-like gaits by using expensive optical motion capture systems is increasing. The optical motion capture systems used in these studies have involved a large number of cameras because such systems use small-sized ball markers; hence the position accuracy of the markers and the system calibration are very significant. However, since the human walking gait is a simple periodic motion rather than a complex motion, we have developed a specialized motion capture system for this study using dual video cameras and large band-type markers without high-level system calibration in order to capture the human walking gait. In addition to its lower complexity, the proposed capture method requires only a low-cost system and has high space efficiency. An image processing algorithm is also proposed for deriving the human gait data. Finally, we verify the reliability and accuracy of our system by comparing a zero moment point (ZMP) trajectory calculated by the motion captured data with a ZMP trajectory measured by foot force sensors.

This paper documents the mechanical system of the electric cable differential (ECD) leg, and its incorporation into a monopod hopping robot named "Thumper" and a bipedal robot named "Mabel." The ECD leg is designed with physical springs and other passive dynamics to match a mathematically simple, bioinspired mass-spring model, which can exhibit robust and economic walking and running gaits. With this design approach, existing spring-mass theory-based controllers can be used to control the robot. The scientific goals of this work focus on finding an energetically optimal leg stiffness for running, and results from experimentation on Thumper and on a simulation of Thumper are presented.

In this paper, we propose a universal biped walking generator that can plan smooth and flexible walking motions in complex environments, including stairs, slopes, and obstacles. In addition to generating collision-free patterns while keeping the balance of the robot, this generator also checks whether the patterns are achievable for the robot. Aiming at this goal, we introduce a simplified walking method to specify the complete biped motions with the unit of a walking step, instead of a time step of the robot's working time used in usual methods. The zero moment point sampling search is also proposed as an efficient method to determine the walking solution for a feasible footstep plan while simultaneously considering dynamic balance and feasibility. Furthermore, another sampling-based algorithm is proposed to revise unfeasible footstep plans to feasible ones. The feasibility and the efficiency of our proposed approach are demonstrated by experiments on simulated and real Nao robots.

Research on human–human communication has identified that people apply different constraints in communication with each other. Application of such constraints as social concerns over feeling, imposition, and disapproval and task concerns over clarity and effectiveness has also been found to be influenced by people's self-construal, being independent or interdependent. Do these constraints and individual difference in self-construal matter in communication with humanoid robots? This study uses the theoretical framework of communication constraints to compare whether or not people of different self-construals apply social-oriented and task-oriented constraints differently to humanoid social robot targets. A total of 161 students from the University of Hawaii at Manoa participated in the study. The participants completed a questionnaire that determined their concern for the five communication constraints (feelings, nonimposition, disapproval, clarity, and effectiveness) in situations involving robots, as well as scales measuring self-construal. The results show interdependent self-construal related significantly with the concerns over avoiding hurting the humanoid's feelings, avoiding inconveniencing the humanoid robot, and avoiding being disliked by the humanoid robot. On the other hand, independent self-construal related significantly with the concern over clarity in communicating with the humanoid robot. However, self-construal did not influence one's concern of effectiveness (a task-oriented constraint) in interaction with humanoid robots. The results of the research offer new insight into the linkage between self-construal, a cultural concept at the individual level, and how human–robot communication is psychologically structured and constrained.

The paper presents a novel method for solving the kinematics problems in real time for fast-moving robot manipulators, animation characters, and hexapod robots. The method uses certain properties of the kinematics map and is based on spatial decomposition, classification with fuzzy logic, and neural network representation of data that are performed during an off-line process. As a result of the preprocessing, the online time for computing the kinematics is extremely small making it possible to perform real-time operations. Examples are provided to demonstrate the performance of the method.