Koki Kameta

A Walking Pattern Generator around Singularity

Humans utilize the extended-knee configuration to achieve efficient walk. A humanoid robot, however, cannot do so, since this posture is a singular one, and hence, an extremely large joint velocity would be generated, provided a conventional control method is used. Some other methods exist, though, that are able to handle motion control at such postures. One example is the singularity consistent (SC) approach, which we used in a previous work to generate a static walk for a humanoid robot. In this paper, we propose a new method to generate a dynamic walking pattern through the singularity neighborhood, by using proper ankle control and pattern generation based on a spherical inverted pendulum model. Experimental results show the effectiveness of this approach

Biped walk based on vertical pivot motion of linear inverted pendulum

Typical human walking patterns include extended-knee postures. Humanoid robots, however, walk with bended knees in order to avoid controlling motion in the presence of kinematic singularities, appearing at the extended-knee postures. Hence, the walking motion patterns of humanoid robots are not as natural and as efficient as those of humans. We have addressed the problem in previous works and shown that, for a static walking pattern, it is possible to include the extended-knee postures, provided control is based on the singularity-consistent (SC) approach, which can ensure stable control in the neighborhood of kinematic singularities. In this paper, we propose a method to generate a dynamic walking pattern including up-and-down waist motion, such that the support leg enters the singularity neighborhood in the middle of each support phase. The basic motion pattern is generated with the help of a linear inverted pendulum model, whose pivot is controlled in the vertical direction to ensure the up-and-down waist motion, independently of the horizontal motion of the pendulum. Experimental results show the effectiveness of our approach.

Walking control using the SC approach for humanoid robot

Conventional humanoid robots cannot achieve human like walking patterns. The reason is that human walking patterns include singularity configurations, i.e. configurations with the extended knees. Such kinematic singularities cause spurious joint motions. In other words, a humanoid robot cannot handle the singularity problem. On the other hand, humans make use of singularity configurations, because at such configurations large forces in certain directions can be generated. Therefore, singularities play an important roll for creating efficient walking patterns. In our previous work, we proposed the singularity consistent (SC) approach that can handle the singularity problem without instabilities. Until now, the approach has been applied to various manipulators. In this paper, we implement the SC approach into a walking pattern generator for a humanoid. Some issues and resolutions addressed. Experimental results show the effectiveness of our approach

Walking control around singularity using a Spherical Inverted Pendulum with an Underfloor Pivot

Humans utilize the configuration with extended knees to achieve efficient walk. A humanoid robot, however, cannot do so, since this posture is a singular configuration. In order to tackle this problem, in our previous work, a walking pattern generator based on the spherical inverted pendulum model has been proposed. It enables to generate up-and-down waist motions and utilizes the singularity neighborhood in the middle of each support phase. However, the spherical inverted pendulum model cannot move the ZMP during each single support phase. In this paper, we propose a new approach to generate a walking pattern based on a Spherical Inverted Pendulum model with an Underfloor Pivot (the SIPUP method). With this method, the ZMP gains mobility during each singleleg support phase, and walking with a smaller acceleration is achieved. In addition, it becomes easy to utilize the singularity neighborhood. As a result, the knee joint can be almost fully extended (to less than 0.01 rad). Experimental results show the effectiveness of our approach.

A Walking Pattern Generator Using a Spherical Inverted Pendulum with an Underfloor Pivot(Mechanical Systems)

Humans utilize the configuration with extended knees to achieve efficient walk. A humanoid robot, however, cannot do so, since this posture is a singular configuration. In order to tackle this problem, in our previous work, a walking pattern generator based on the spherical inverted pendulum model has been proposed. It enables to generate up-and-down waist motions and utilizes the singularity neighborhood in the middle of each support phase. However, the spherical inverted pendulum model cannot move the ZMP during each single support phase. In this paper, we propose a new approach to generate a walking pattern based on a Spherical Inverted Pendulum model with an Underfloor Pivot (the SIPUP method). With this method, the ZMP gains mobility during each single-leg support phase, and walking with a smaller acceleration is achieved. In addition, it becomes easy to utilize the singularity neighborhood. As a result, the knee joint can be almost fully extended (less than 0.01 rad). Experimental results show the effectiveness of our approach.

Walking control based on the SIPUP model

In our previous work, a walking pattern generator based on the spherical inverted pendulum model has been proposed. It enables to generate up-and-down waist motions and utilizes the singularity neighborhood in the middle of each support phase. However, the spherical inverted pendulum model cannot move the ZMP during each single support phase. In this paper, we propose a new approach to generate a walking pattern based on a spherical inverted pendulum model with an underfloor pivot (the SIPUP method). With this method, the ZMP gains mobility during each single-leg support phase, and walking with a smaller acceleration is achieved. In addition, it becomes easy to utilize singularity neighborhood. As a result, the knee joint can be almost fully extended (less than 0.01 rad). Experimental results show the effectiveness of our approach.