This paper tries to quantify the quality of a motion database
since to their knowledge it hasn't been done before.
By doing this, they try to make it easier on how to improve
existing databases or use it in other methods.
Also they try to solve the problem on moving through
environments with obstacles using an embedded graph.
To quantify how well a motion graph enables a character to
navigate through an environment, they suggest a method that
divides the environment into uniform grids an constrain
ever motion edge to start in one center of the grid and end
in one center of the grid by using motion editing while
posing some thresholds on how much can be edited.
During creating an embedded graph which is done by trying
every possible senario in a 4D space, edges that are obscured
by obstacles are removed. This technique isn't really scaleable
since it requires 4D space, time, and by creating an invironment
that has multiple stories, can easily require another dimension.