this is for holding javascript data
Sven Schmit edited problems.tex
over 9 years ago
Commit id: 9d319b5c07cb0b8e6a197d0c0c43c93ee42b96b3
deletions | additions
diff --git a/problems.tex b/problems.tex
index 1b88337..57cebcd 100644
--- a/problems.tex
+++ b/problems.tex
...
\section{Problems}
Using Given the
above setup, one can come up with setup as outlined above, there are several interesting
problems.
Some are very easy, others might be very complicated.
We list a few that problems one can investigate.
In our case, we are
interesting interested in
exploring, where we want comparing two approaches to
start simple and build more complicated algorithms as we go. learning to herd:
\begin{enumerate}
\item
A dog has Dogs act independently, while they can observe each other, they cannot explicitly coordinate their movements.
\item There is a shepherd AI that coordinates the movements of the dogs. This is a two=tiered approach, where dogs are trained to
move perform specific smaller tasks (move to a target
\item A dog has location, circle the flock clockwise, etc.) and a central AI that gives the instructions to
move one sheep the dogs.
\end{enumerate}
To investigate these two approaches, we propose to
start with some simple tasks, and expand upon them.
Some examples of simple tasks include assigning a target
\item One for the dog
has to move
multiple sheep to a target
\item One too, and one dog
must move multiple sheep, each to a different target
\item Multiple dogs have having to move
multiple one sheep to a
target.
\item Learn dogs the target location.
More complicated tasks can be based upon these basic
herding commands\footnote{see \url{http://en.wikipedia.org/wiki/Herding_dog} under `Basic herding dog commands' for an overview} and then have a seperate AI give commands to the dogs.
\end{enumerate} tasks.