Planning with Sensing Actions and Incomplete Information Using Logic Programming

 

 Boolean Version

 Multivalued Version

                 

We present a logic programming based conditional planner that is capable of generating conditional plans and conformant plans in the presence of sensing actions and incomplete information. A planning problem P = (D,I,G), where (D,I) is an action theory written in either A_k or AM_k and G is a conjunction of literals, is represented as a Sictus programs with the keywords 'causes', 'caused', 'executable', and 'determines'. It will be translated into a smodels program, called p(P), using a translator. We use the smodels system to compute answer sets of p(P).

To date, the planner has been tested with a number of domains from the literature and the output result is compared with that of the SGP system for the same domains.

We offer two versions for the planner: one for domains with boolean fluents only and the other for domains with multivalued fluents.

Download all experiments data and scripts used in the paper here: tar gzip file.

Please contact me if you have any questions or comments on this stuff.