In the past six months we have developed a domain independent, logic
programming-based planner for action domain descriptions written in a
language designed under the basis of 
K . The planner,
called COPLAS, is able to generate conditional plans from domains with
sensing actions. The domain can also describe actions with
non-deterministic effects. An action can cause the lost of knowledge
if its effect is non-deterministic. Take for example the action of
tossing a coin. Assume that in the initial situation we knew that the
coin was showing heads. If we toss the coin we know it will land with
either heads showing or with tails showing, but exactly which cannot
be predicted. Non-deterministic actions and sensing actions have
opposite effects on an agent's knowledge. The planer can also handle
a limited class of ramification constrains in the form of axioms and
actions with universally quantified effects. For example. moving a
table will cause all the objects on the table to move along with
the table.