Notes:
[addendum, Feb-05-2002]
Since writing this, I have come to understand this CCD method as
simply an instance of Jacobi (or Gauss-Seidel) iteration to satisfy
constraints. I suspect that Wang and Chen were also not aware of
this, hence their coining of a new monicker. This being an iterative
relaxation scheme, one would expect that the convergence of this
method could be improved upon by using Successive Over-Relaxation
(SOR). SOR is basically intentionally overshooting your 1DOF
solutions by some factor. While not at all about manipulator
kinematics, the first few chapters of William Briggs' A Multigrid
Tutorial (publisher SIAM, 2000) give an excellent introduction to
iterative relaxation schemes (Jacobi,Gauss-Seidel,SOR,etc) and their
convergence properties. See also the talk given by Thomas Jakobsen on
"Advanced Character Physics" which he delivered at the 2001
Game Developer's Conference. The two key components of his approach
are a Verlet integration scheme and the use of Jacobi iteration to
handle constraints.