I can't give you all the links, but some of the names are:-
Melt-grow algorithm - one unit moves at a time, and the algorithm uses
an intermediate shape:-
http://www.mit.edu/~vona/publications/Rus_Vona__1999__Self-reconfiguration_Planning_with_Compressible_Unit_Modules.pdf
Pacman algorithm, improves melt-grow with parrallel movements, but can
get stuck in minima. No intermediate shape.
Hormone control for CONRO
http://ieeexplore.ieee.org/Xplore/login.jsp?url=/iel5/9887/31421/01461243.pdf.
This is parrallel but I think also suffers from local minima
Then for metamorphic an early work uses a simulated annealing with a
heuristic. I think the paper is called "useful similarity metrics for
a class of self-configuring robots" or something.
The M-TRAN robot, which is one of the best reconfiguring systems, has
difficulty planning. They have a software suite which allows them to
combine search with manual planning.
There is alot more to read out there but that should give you a start