Project Abstract:
Randomized motion planning techniques have proven to be useful
in solving motion planning problems in environments with many
degrees of freedom. However, it is well know that their performace
degrades considerably in environments where there are narrow
passages. In this project we try to investigate and propose some
alternatives to alleviate this drawback. In particular, we have been
working in Adaptive Hybrid Sampling. The main idea in that case is
to combine automatically different component samplers which are
chosen according to their observed performance. The results are
encouraging. We are also working on a general representation of
the geometry of the robot such that it could not only be used for
collision checking purposes, but also to provide information for
the planning algorithm. |