Ken Tomiyama

Sensor-based biped gait generation scheme for humanoid - implementation and evaluation

A new gait generation method, called sensor-based biped gait generation, for humanoids is proposed here. The proposed method chooses a gait generating method from a set of prepared ones according to sensor measurements of robots environment and to its behavior state. A gait generator that is based on the proposed method is designed and implemented onto an original humanoid to demonstrate its effectiveness.

Surprise generator for virtual KANSEI based on human surprise characteristics

This paper proposes a technique to generate emotion state of surprise and to generate motions to express this emotion for robots. Surprise emotion is known to have unique characteristics compared to other emotions such as anger and grief. It is often caused by discrepancies between prediction and actual event and is also an instant emotion. A conceptual structure of a surprise generator to generate emotion state of surprise for robots is developed based on these human surprise characteristics. The generator contains a predictor that is used to fined discrepancies between the real events and expectations.

An Adaptive Biped Gait Generation Scheme Utilizing Characteristics of Various Gaits

The purpose of this study is to develop a biped locomotion movement generator, named the Sensor-Based Gait Generation system, for humanoid robots that puts biped locomotion to practical use in real environment. The proposed method, in short a gait generator, enables humanoids to use various gaits according to walking surface condition. Policies and structure of the Sensor-Based Gait Generation system are described. A gait generator that is based on the proposed method is designed and implemented onto an original humanoid to demonstrate effectiveness of the proposed method. Human beings clearly differentiate their walking movements on the downward slope, on the slippery surface and on the flat ground. In other words, humans use a gait that is appropriate to the condition of the walking surface. At present, however, most humanoid robots use only a single gait with possibly adjustable parameters and therefore have clear disadvantage compared with humans. The proposed method imitates gait selection strategy of human and thus eliminates this deficiency. Many gait generation methods have been proposed already. A gait generated by any one of those methods has good characteristics and shortcomings and therefore has advantages and disadvantages against a given walking surface condition. When humanoids adopt the proposed gait generator, they will be able not only to walk on every road conditions but also to take advantages of good characteristics of each gait. Especially, we focus on policies of the Sensor-Based Gait Generation system in this paper. One of the two major topics is the explanation of its main components and the other is the configuration of criteria for gait evaluation. In addition, structure of the Sensor-Based Gait Generation system based on this methodology is discussed. After explaining the developed and implemented system that realizes the proposed method, details of environment for experiments are described. Experimental results clearly exhibit practical advantages of the proposed method. Capabilities of the implemental system shown by experimental results are summarized. Conclusions and items for further study are listed at the end.

Building Human Motion Map with human states estimation in indoor dynamic environments

In recent years, robots working in human living space with human-robot interaction are actively researched. To these robots, it is important to perform environmental cognition including to estimate human states around the robots and to have environment map for autonomous motion of the robots. In this research, we focus on building maps of dynamic environment. In this paper, we proposed Human Motion Map, a high-dimensional map, which incorporating human states information into results of conventional SLAM. The Human Motion Map is with high potential on further human estimation and motion planning for robots by considering the uncertainties of environment and human states.

Hands-on Education of Robotics Department for Four Years of College

In this paper, we introduce engineering education at the Department of Advanced Robotics, Chiba Institute of Technology. At the department, we try to teach useful knowledge and provide laboratory work leading to useful experience. One purpose of the curriculum is to enable students to design a system with a mechanism, control circuit, and computer programming. We then provide many lectures related to system design – control engineering, mechanics, mechanical dynamics, electronic circuits, information engineering, mechanical drawing, and so on – and provide laboratory work on related theory in the lectures. Laboratory work helps students understand abstract theories that are difficult to understand based on desk study alone. This laboratory work continues from the first to fourth years. In addition, we provide many project studies. Some students try to develop their own systems through extracurricular studies. Through the project, students obtain much knowledge and experience. After introducing our curriculum, we discuss the results of this curriculum.

Stabilization of Biped Locomotion Utilizing Compensation in Double-Supporting Phase

A gait generation method for biped robots that realizes stabilization of the single-supporting phase by taking advantage of robustness of the double-supporting phase is developed. Considering these points, the proposed method utilizes the double-supporting phase to eliminate disturbances that are caused by ground condition and/or other factors affecting stability of biped locomotion. In our method, ample stability margin in the double-supporting phase is also exploited to realize the angular momentum that is required by the next single-supporting phase. The proposed method can achieve more stable continuous walk due to these features. In this paper, the proposed gait generation method is explained in detail and some results of experiments using originally developed humanoid robot are reported.

Study of kawaii-ness in motion: physical properties of kawaii motion of roomba

In this paper, as the second report of the study on Kawaii-ness in motion, we investigate the relationship between physical properties of motion and Kawaii-ness using Roomba. Kawaii is one of the representative concepts of Japan-original Kansei. First, we computed parameters of seven physical properties (position, velocity, acceleration, angle, angular velocity, angular acceleration, and time) from three types of motions of Roomba used in the first study. Second, we composed 24 types of robot motions and asked the subjects to evaluate their impressions. We asked the subjects to answer the questionnaire consisting of the 20 pairing adjectives prepared according to the SD method. The extracted physical features in seven physical parameters in composed motions are correlated with the Kawaii-ness based on the result of the questionnaire. We report our findings in detail in this paper.

Attitude determination framework by globally and asymptotically stable bias error estimation with disturbance attenuation and rejection

This paper presents a new methodology of attitude determination for cost-effective and small-size inertial measurement units, consisting of tri-axial gyroscope sensors, accelerometers and geo-magnetometers. Introduced for the algorithm development is a quaternion feedback structure, where bias terms in the gyroscope rate information, appearing in the phase-locked loop, are identified with the theoretical guarantee of the global and asymptotical stability. The bias-term identification and modification process is performed continuously without the influence of the disturbance by combining the proposed disturbance evaluation scheme and the conventional vector matching method. Practical validity of the presented framework is fully evaluated by experiments. The evaluation result shows that the approach in this paper can offer drift-free performance with property of disturbance attenuation and rejection under arbitrary high-frequency attitude test motion of spatial rotation and translation.

A remote control system with combination of manual and automatic control using gripping force

In this paper, we propose a new method to combine manual and automatic controls for remote controlled robots using gripping force. An important issue in the remote control system is how to combine the manual and automatic controls, namely, how to determine the mixing ratio of the two controls. We propose to adopt the gripping force of a joystick for this purpose because humans seem to grip the joystick tightly when they want to operate the robot aggressively and in emergency situations. First, we confirmed this tendency and identified relationship between the gripping force and the urgency of operation. An experimental system to simulate a remote controlled robot, consisting of a force feedback joystick with a pressure sensor and an image of working field on a display is developed. We compared several remote control methods on the developed simulator and found that the proposed method has superiority against other methods in many aspects.