in Japanese

Research Interests

Mechanical Analysis of Contact Forces in Manipulation

In Robotic manipulation, appropriate contact forces are applied to target objects by robots and environments, and help the robots accomplish stable grasping and manipulation. Hence analysis of the forces is a key issue for performance of dexterous manipulation.
Rigid-body models, which is with less parameters of the model facilitate the analysis. We study constraint conditions of frictional forces occurred in rigid-body models and analyze contact forces in manipulation for further dexterity. An important idea in the analysis is to approximate physical phenomena by rigid-body models without any suspicion, that is, the approximation corresponds exactly to the principle.
Proposed analysis of contact forces can be applied to both static grasping and quasi-static graspless manipulation.

Grasping
<Grasping>

Graspless manipulation
<Graspless manipulation>

Moreover as applications of the analysis, we evaluate robustness of manipulation: whether the planned manipulation can be accomplished with no influences by disturbances, and study joint torques optimization in manipulation.

Roustness evaluation
<Robustness evaluation for pushing manipulation>

Joint torque optimization for graspless manipulation
<Joint torque optimization for graspless manipulation>

3D Multifingered Caging

Caging is a geometrical method to confine an object by robots surrounding it.

3D Multifingered Caging
<3D Multifingered Caging>

Grasping as a typical object constraint can be performed by applying contact forces from robots. Then sensing of the contact forces and force control are often required for appropriate grasp forces.
On the other hand, caging does not need them and is achieved even by position-controlled robots to capture objects. In addition, only geometrical features of objects are required for planning of caging grasps.
We propose 3D multifingered caging that is caging by multifingered hands in three-dimensional space and study sufficient conditions of the caging, object recognition to retrieve geometrical features of objects for caging and motion planning.

Classification of 3D Multifingered Caging

Three types of 3D multifingered caging is classified as strategy.

Envelope-type Caging

Envelope-type caging can be achieved by robots surrounding a target object, and then it cannot escape from the constraint called as cage. Three fingers at least are required for the caging.


Ring-type Caging

Ring-type caging is a useful strategy to capture objects with holes. A robot finger is inserted to the hole and approaches closely to another finger to form Hopf Link with the ring-shaped object.


Waist-type Caging

Waist-type caging focuses on constricted parts of objects. When robot fingers winds about the constricted region, the ring-formed fingers collide with other thick parts, and then the object cannot escape from the hand.

Object Recognition for Ring-type Caging

For Ring-type caging, we have to obtain ring-shaped features of objects. We deal with scissors and mugs by using webcams and RGB-D sensors (e.g. Kinect).

Caging grasps by prosthetic hands

Prosthetic hands are used for amputees as substitution of hands and arms. We consider the difficulty of flexible force control of prosthetic hands in grasping and manipulation, and apply caging to them because caging can be accomplished without any conditions of contact forces.

caging_mug mug_exp caging_papercup papercup_exp
<Multifingered caging by a prosthetic hand>

If a robot hand is asymmetry as human hands, it is difficult to confine objects completely. Then we propose Partial caging, where the robot hand surrounds the object partially and allows it to escape from the hand. Nevertheless the object can be prevented from escaping such as by gravitational forces (See the above figures).

Analysis of Elastic and Dynamic Properties of Fingers Attributed to Muscle-Tendon Complex

Muscle-tendon complex, which is combination of muscles and tendons, is a key issue of human motions in biomechanics. A tendon makes connection between a muscle and a bone, and it is little lengthened by usual muscular tension. Thus the tendon helps the bone be in motion caused by shortening of muscles. The tendon is, however, stretched when some external forces are applied to it and then it behaves as springs, which can exert tension and elastic energy. For example, we can jump higher after squatting than that from standing. It is because elastic energy in tendons is translate to kinetic energy, which contributes to high speed of body motion. Consequently appropriate uses of elasticity of tendons achieve high performance that cannot be impossible only by muscles.
We focus on muscle-tendon complexes of human hands and analyze their mechanical properties and dynamics. They assume to contribute performances in sports such as volleyball and baseball pitching. Then mechanism of dynamic manipulation by human hands can be unveiled from the analysis of dynamics.

hand-mtc
<Muscle-tendon complex of hand>

Manipulation of Small Object Utilizing Force Control and Vision-based Position Control

Master's thesis (Satoshi Makita)
Dept. of Mechanical Engineering, Div. of Systems Integration, Graduate School of Engineering, Yokohama National University
(Apr 2005 - Mar 2007)
Micro Work Cell
(Product of AJI Co., Ltd.)

Collaborators