A fundamental
and ever present problem in robotics is the inverse
kinematics problem. The standard approach to this problem
is through the Jacobian approach that solves the problem
at the velocity level and then uses integration to find
the joint angles. This approach is prone to a number
of difficulties including singularities.
To
resolve these issues, we formulate and solve the inverse
kinematic problem at the position level using a genetic
algorithm approach. In addition, to achieve real-time
characteristics, we have developed a special procedure
that optimizes the number of generations in the evolutionary
process.
The
method has been applied to both non-redundant (Puma
60) and redundant arms (Robotics Research K1207 shown
right) with good tracking accuracies. |