Due [Mar 27, 2003]

Qn 1. (Empirical Homework): See if CHAFF solver changed the dynamics
between Blakbox and GP-CSP since 2000. Basically, compare GP-CSP and
Blackbox (with CHAFF solver) on the problems  listed in the GP-CSP
paper as the ones that GP-CSP does better over Blackbox
(e.g. rocket-ext-b and hsp-bw-4). See if the relative performance
changes. Look at GP-CSP AIJ paper to find out what were the best
settings for it, and use those settings. For Blackbox, use CHAFF
solver. In both cases, use just the single solver (no shifting between 
solvers).


Qn 2. Consider the planning graph from Qn III.a (the PG with static
interferences alone) in Homework 3. Convert it into CSP encoding.
Compare it to the encoding you got in Qn V of Homework 3 (which is a
CSP encoding with propagated mutexes). Show that doing 3-consistency
enforcement will bring in the additional constraints present in the
second encoding (it is enough if you show the derivation of one of the 
additional constraints)

Qn 3. Consider our old problem (Problem III from previous homework)
again

There are two actions: A1 and A2

A1: prec: p eff: q

A2 prec: r eff: ~q,w

We start with init state where p and r are true. 

**and our goals are q and w. 

3.a. Write the 2-step SAT encoding based on progression proof (write
all the constraints)

3.b. Write the 2-step SAT encoding based on the regression proof. 

3.c. Write the 2-step SAT encoding based on the causal proof.