This paper describes the design and development of softening deformable structure which enables robots to have high shape adaptability. This structure is made of a low-melting-point alloy and can be made soft and deformable through a phase change induced by heating. In the soft deformable state, it can be changed their shape by pressing themselves against objects and by using their own weight. On the other hand, in the hard state, it is enough stiff to be a part of a robot. Using this structure, we have developed two types of legged robots which have softening deformable legs. The first robot can reconstruct its original shape through reheating. The second robot has higher deformation ability and realize grasping objects, walking on a pillar and going up a ladder, which are difficult to do with its original shape, by deforming the shapes of the legs.
This paper proposes a method of relocating storage devices attached to environment for sharing knowledge efficiently among multiple robots. If a robot stores acquired knowledge into the storage device attached to the environment, other robots can achieve the task by using the knowledge. In order to use the storage device efficiently, it is required to store more useful knowledge for achieving tasks. Therefore, we propose a method of evaluating usefulness of the knowledge with respect to the reducible cost and storing more useful knowledge in the storage device. A memory unit containing with useless knowledge is removed from the storage device and added to the device which more useful knowledge is required. Then, the robots can store useful knowledge for whole environment and the memory units are flexibly relocated according to the change of a series of the tasks. The simulation results of multiple robots navigation for repetitive transportation tasks show the effectiveness of our method.
Serial link robots with unactuated (passive) joints are attracting research interest. If two of such robots can couple with each other, they can reconfigure to a parallel robot, which can make the number of actuators equal to the number of degrees of freedom. If they can couple with each other at different portions of them, they can constitute a reconfigurable parallel robot. This paper proposes this concept of a robot, presenting that two two-link robots, the first joints of which are unactuated, can reconfigure to a 5-link planar parallel configuration and a 4-link planar parallel configuration plus one actuated link. This reconfigurable parallel robot has only two actuators but can have multiple functions by reconfigurations. Due to the unactuated joints, whether or not the two-link robots can couple with each other is a non-trivial problem. We propose coupling sequences for forming the 4-link and 5-link configurations and verify those experimentally.
In this study, a scheme using the Finite Element Method (FEM) for calculating inverse dynamics is proposed and applied to open- and closed-loop link mechanisms. In this scheme, the entire system is subdivided into discrete elements and evaluated as a continuum. A single-link structure of a pin joint and a rigid bar is expressed by using the Shifted Integration (SI) technique. The proposed scheme calculates nodal forces by evaluating equations of motion in a matrix form, and thus information from the entire system can be handled in parallel. The obtained nodal forces are then used to calculate the joint torque in the system. Simple numerical tests on open-and closed-loop link mechanisms are carried out, and it is verified that the scheme can be used as a unified numerical scheme independent of the system configuration.
Manipulator dynamics on the velocity-dependent nonlinear joint torques due to the Coriolis and centrifugal forcesis investigated. The nonlinear mapping from the velocity space to the joint torque space is characterized by geomet-rically describing a joint torque set as an image of a velocity set with a constant norm. The geometry represents the characteristics of the nonlinear joint torques independent of velocity directions. The characteristics are investigated in two cases: 2DOF and more than 2DOF. 2DOF manipulators have the characteristic that the image of the isotropic velocity set has the shape of an ellipse in the joint torque space. In the case of more than 2DOF, the image is included in an ellipsoid. These characteristics are verified with numerical manipulator models.