6. LM-Cut heuristic

Exercise 1: Consider the following STRIPS task $\Pi=\langle F,A,s_0,g\rangle$ where

  • $F=\{a,b,c,d,e\}$,
  • $s_0=\{a\}$,
  • $g=\{c,d,e\}$,
  • $A=\{o_1,o_2,o_3,o_4\}$
action pre add cost
$o_1$ $\{a\}$ $\{b,c\}$ $3$
$o_2$ $\{a\}$ $\{d\}$ $5$
$o_3$ $\{b\}$ $\{c,d\}$ $1$
$o_4$ $\{a,b\}$ $\{e\}$ $4$

In the following exercises, we will compute LM-Cut at $s_0$. As the first step, construct the STRIPS task $\Pi_{s_0}$ extended by new facts $\bot,\top$ so that $\{\bot\}$ will be the new initial state and $\{\top\}$ the new goal state.

Solution

Exercise 2: Compute the fixed point of $\Gamma_{max}$ for the new initial state $\{\bot\}$ and construct the justification graph for the precondition choice function that selects for an action $a$ its most expensive fact $p\in\mathsf{pre}_a$ based on the fixed point.

Solution

Exercise 3: Find the $\bot$-$\top$-cut in the justification graph, its action landmark, its contribution to the LM-Cut heuristic, and the remaining cost function after subtracting the saturated cost for the landmark.

Solution

Exercise 4: Do the next iteration of LM-Cut computation.

Solution

Exercise 5: Do the final iteration of the LM-Cut heuristic and compute its final value.

Solution

courses/pui/tutorials/tutorial06.txt · Last modified: 2024/03/19 10:32 by xhorcik