We use the functions $agent(X, S_i), crate(cratename, X, S_i)$ and
$target(cratename, X)$ to easily represent the information.
\item{Q3}\\
+ $\begin{array}{lll}
+ connected(loc11, loc21, east)\wedge &
+ connected(loc11, loc12, north)\wedge &
+ connected(loc12, loc22, east)\wedge\\
+ connected(loc12, loc13, north)\wedge &
+ connected(loc13, loc23, east)\wedge &
+ connected(loc13, loc14, north)\wedge\\
+ connected(loc14, loc24, east)\wedge &
+ connected(loc21, loc31, east)\wedge &
+ connected(loc21, loc22, north)\wedge\\
+ connected(loc22, loc32, east)\wedge &
+ connected(loc22, loc23, north)\wedge &
+ connected(loc23, loc33, east)\wedge\\
+ connected(loc23, loc24, north)\wedge &
+ connected(loc31, loc32, north)\wedge &
+ connected(loc32, loc33, north)\wedge\\
+ connected(loc21, loc11, west)\wedge &
+ connected(loc12, loc11, south)\wedge &
+ connected(loc22, loc12, west)\wedge\\
+ connected(loc13, loc12, south)\wedge &
+ connected(loc23, loc13, west)\wedge &
+ connected(loc14, loc13, south)\wedge\\
+ connected(loc24, loc14, west)\wedge &
+ connected(loc31, loc21, west)\wedge &
+ connected(loc22, loc21, south)\wedge\\
+ connected(loc32, loc22, west)\wedge &
+ connected(loc23, loc22, south)\wedge &
+ connected(loc33, loc23, west)\wedge\\
+ connected(loc24, loc23, south)\wedge &
+ connected(loc32, loc31, south)\wedge &
+ connected(loc33, loc32, south)\wedge\\
+ crate(cratec, loc21, s0)\wedge &
+ crate(crateb, loc22, s0)\wedge &
+ crate(cratea, loc23, s0)\wedge\\
+ target(cratea, loc12)\wedge &
+ target(crateb, loc13)\wedge &
+ target(cratec, loc11)\wedge\\
+ agent(loc32, s0)\wedge\\
+ \end{array}$
\item{Q4}\\
+ $goal(s)\rightarrow\\
+ crate(cratea, loc12, s) \wedge
+ crate(crateb, loc13, s) \wedge
+ crate(cratec, loc11, s)$
+
\end{itemize}
\subsubsection{Task 2: Actions}
\begin{itemize}
\item{Q5}\\
- \begin{equation}
- \begin{split}
- Poss(move(x, y), s) \equiv &\\
- & (\exists z: connected(x, y, z)) \wedge\\
- & \neg(crate(x, y, s)) \wedge
- \end{split}
- \end{equation}
-
+ $Poss(move(x, y), s) \equiv \\
+ (\exists z: connected(x, y, z)) \wedge\\
+ \neg(crate(x, y, s))$
\item{Q6}\\
$Poss(push(x, y), s) \equiv\\
agent(x, s) \wedge\\
\wedge\\
(\exists \alpha:
connected(z, \alpha, y) \wedge
- \neg (\exists \beta: crate(\beta, \alpha, s))))$
+ (\nexists \beta: crate(\beta, \alpha, s))))$
\item{Q7}\\
$agent(x, result(z, s)) \rightarrow\\
(\exists y: z = move(y, x)) \vee\\
z = push(\beta, \alpha) \wedge
connected(\beta, x, \alpha)) \vee\\
(\exists \epsilon, \gamma:
- not(z = move(x, \epsilon)) \wedge
- not(z = push(x, \gamma)) \wedge
+ z \neq move(x, \epsilon) \wedge
+ z \neq push(x, \gamma) \wedge
agent(x, s))\\
crate(x, y, result(A, s)) \rightarrow\\
- (\exists z,alpha:
+ (\exists z,\alpha:
A = push(z, \alpha),
(\exists \beta:
connected(z, \beta, \alpha),
)
) \vee\\
(\exists z,\alpha:
- not(A = push(z, \alpha)),
- connected(z, y, \alpha),
+ A \neq push(z, \alpha)\wedge
+ connected(z, y, \alpha)\wedge
crate(x, y, s)
)
$
\subsection{Part 3: Extending the domain}
-\subsection{Evaluation}
-\begin{itemize}
- \item{How much time did it take?}\\
- p1t1: 45m
- \item{What would you like to see changed?}\\
- p1t1: There is an ambiguity in p1t1Q3, it's not clear if the starred
- locations should be included in the initial state(therefore not only in
- the goal state).\\
-\end{itemize}
+\subsection{Part 4: General questions}
+\subsubsection{Task 10: Sitcalc expressivity}
+Situation calculus(sitcalc from now on) is very expressive because you can
+express yourself very detailed without encountering the frame problem. When the
+problem space expands the computational strength needed explodes. Sitcalc is
+therefore not very usefull when you want to plan far behind. For comparison,
+calculating a sokoban path 10 steps in the future already takes hours on a
+normal computer.
+
+The model is easy to extend to bigger and more complex problems, it doesn't
+scale that well however...
+
+\subsubsection{Task 11: Related work}
+Zhou, N. (2013). A Tabled Prolog Program for Solving Sokoban, 124, 561–575. doi:10.3233/FI-2013-849