Response lecture for planning block

1.3.

The "\" symbol means: except

pre(a1,a2) $\rightarrow$ [ a1;a2 ] $\rightarrow$ add(a1,a2) \ del(a1,a2)

Where:

add(a1;a2) = add(a1) \ del(a2) union add(a2)

del(a1;a2) = del(a1) \ add(a2) union del(a2)

and pre(a1,a2) = pre(a1) $\cup$ (pre(a2) \ add(a1) )

So:

pre(puc,mc))= ({Rloc=cs,$\neg$ Rhc} $\cup$ {$\emptyset$ Rhc}= {Rloc=cs,$\neg$ Rhc}

add(puc;mc) = {Rhc} \ {Rloc=cs} $\cup$ {Rloc=off}={Rhc, Rloc=off}

del(puc;mc) = ( $\emptyset$ \ {Rloc=off}) $\cup$ {Rloc=cs} = {Rloc=cs}

1.8

My confusion for this exercise is that the $\rightarrow$ is not a logical implication. What is required is that a rule displays all possible options. Teacher has no clue what is going on as well.

Defintion of accuracy: precise and minimal.

Minimal: left hand side of arrow is required by preconditions. For prop. 1 (c=X) is not specified in any precondition, thus it is not minimal.

Precise: covers all relevant cases

2.2

Unpack the boxes before the states: contains boolean expressions for preconditions. E.g. for Garb~1~ = Garb~0~ $\lor$ ( Carry~0~ \/ Dolly~0~)

The boxes before the actions contain constraints on the actions at a particular k. E.g. Carry~0~ and Cook~0~ exclude each other. A box connects the mutual exclusive actions.

A partial order plan is very brief for the example, see slides.

3.6

My h~add~ was correct (17).

My h~max~ was also correct (6)

For my Relaxed Planning Graph, I miss the last connection from h to f. Also I have connected to G at step 4. The example answer draws horizontal lines from f to f and j to j from step 4 to 5. That way all paths end up in the same goal node. For the final path this does not make a difference. In the path extraction, we move from the goal at t=5 directly to j and f at t=3, because there they are supported for the first time.

Supporter: earliest state that is defined. (Time proceeds from left to right).

Correction: h~ff~ = 16.

Optimal, go over h->f, with cost 16.

Eyeball the graph without the use of any algorithm: h^+^(G) = 12. Feedforward search heuristic are greedy and do not provide optimal solutions.