Clone wiki

Planner for Relevant Policies / Compare


Want to compare with PRP or use PRP in an experiment? Excellent! If you don't mind me ducking out as a potential reviewer (and I certainly don't), then please contact me to make sure everything runs smoothly. This especially holds true if you ever have an issue running PRP out-of-the-box -- disabling certain features for whatever reason can be quite detrimental to the performance. If there is a concern about keeping the research agenda private, I will always openly answer any questions that surround what I'm up to with the software (I dislike re-inventing or unknowingly co-inventing things just as much as the next researcher out there).

Comparing Against PRP

I love finding the holes in PRP for the problems out there. If you have a new technique / insight / planner that puts PRP to shame then let me know! I'll do my best to help you tease apart why PRP fails. I view such work as opportunity to advance the field, and I am more than happy to help support such endeavours. That said, once the work is recognized publicly I'll do my best plug any hole people point out (note -- I improve the software quite frequently, and the best way to make sure I don't invalidate the results with the next iteration is to contact me and make sure you have something reasonable to compare against). Also, see the script usage for details on comparing versions of PRP.

Comparing With PRP

If you want to use PRP in your research, let me know if you would like help gearing things up specifically for your use. Yes, PRP can be used as a black-box FOND solver, but if you are using it for application Foo, then you may have a much easier time if you tweak the heuristics...or you change how the policy is scored...or whatever. Until I can properly document things, a quick conversation with me could save heaps of headaches when trying to modify the solver for your needs.

Existing Comparisons

This section will eventually house the comparisons out there that use PRP (either to compare against or with). It will also try to shed some insight on the results, and any followup work that has occurred as a result.