Abstract
Dexterity is an important issue for design, installation planning, trajectory planning, etc. of manipulators.
Many evaluating methods have been proposed such as Yoshikawa's manipulability. They have, however, some problems, one of which is that they cannot consider characteristics of manipulator motion. It has been impossible to evaluate how efficient a manipulator mechanism is from a viewpoint of characteristics of a required task. A manipulator should be evaluated to be dexterous when it can efficiently perform its required task even if it cannot accomplish other kind of tasks very well.
In this paper, a new dexterity measure stochastic manipulability measure is presented. It can evaluate kinematic dexterity considering deviation of direction of required manipulator motion, which is one of the characteristics of a task. Stochastic manipulability is based on expected value of mean of necessary joint velocity under a stochastic interpretation of required motion. The deviation is specified using a probability density function.
Applications to 2-, 3- and 6-DOF manipulators revealed the following:
1. stochastic manipulability can evaluate dexterity considering deviation of motion,
2. it can be used for comparison of manipulators with different DOFs,
3. it is important to consider the deviation of motion in evaluation,
4. singular points can substantially lose their singularity, and
5. optimal trajectory of redundant manipulators can be obtained using maps in null space of Jacobian matrix.
