Two best planners from this planning competition are not based on ASP https://ipc2018-cla <https://ipc2018-classical.bitbucket.io/scores.html>s <https://ipc2018-classical.bitbucket.io/scores.html> sical.bitbucket.io/scores.html <https://ipc2018-classical.bitbucket.io/scores.html>, first one is based on deep-learning and second one on a heuristic with a pattern database. In 2014 half of the planners were based on IBM CPLEX solver.. i guess the reason is that one can't have a universally optimal solver for np-hard problems, but one can find some shortcuts for a reasonably small subset of them.
ср, 28 апр. 2021 г. в 22:25, Linas Vepstas <[email protected]>: > Ben, > > On Wed, Apr 28, 2021 at 1:59 PM Ben Goertzel <[email protected]> wrote: > >> >>> >>> Now, what I say above is "easy to say" but is "hard to do" -- >>> implementing what I suggest is a large project. But then, in software, >>> nothing is free. facebook and google and amazon employ thousands of >>> engineers because writing good software is hard. Imagining that you can >>> create a new planner out of thin air in a few months is not a realistic >>> dream. Don't repeat history; learn from it. >>> >>> >> As I understand what is being proposed here is a student research project >> not a large-scale engineering project... >> > > Anatoly implied that he wants to create a new planner in hyperion. > Creating hyperion, and a new planner for it, using "more natural" > expressiveness, is .. well, sure you can treat hyperon as a student project > consisting of throw-away code. But the way you talk about hyperon is as if > it is going to be a replacement for the atomspace, and not as some > experimental throw away code. > > Anyway, I'm not really talking about throw-away code; I am just saying > that comp sci journals, such as the "Journal of Optimization" have > thousands of articles on planning algorithms, developed over many decades; > Ignoring these, and walking in with a clean-slate hyperon design ... well, > don't be naive. > > >> >> Exploring constraint-satisfaction-based planning makes sense, but for >> some planning domains this approach may not be best. E.g. if you're >> planning in a highly dynamic environment (as faced say by robots moving >> around in a house or on the street) then I'm not sure the available >> constraint satisfaction algos can deal well w/ the needed real-time plan >> updating...? >> > > (1) ASP will solve most planning and constraint satisfaction problems in > milliseconds. So you can do a 60-frames-per-second update rate if you > wish. > > (2) It's not like people haven't done this before. I went to some talk, I > think in HKSTP, where someone talked about using SLAM (simultaneous > localization and mapping) from ultrasound sensors on some robot, and then > hooking that into ASP (specifically, the Potsdam solver was mentioned) to > perform basic reasoning tasks ("if the door is closed, try to push the > door", "if the door won't push try to pull the door" "else reach for the > doorknob") -- and this was not the first time I've heard of ASP being > integrated in with probabilistic systems in robotics -- talk titles tend to > be along the lines of "how to integrate probabilistic and crisp logic > reasoning in robotics" or something similar. > > >> My 2 tokens worth... >> > > Are these one of those bitcoin tokens that everyone is talking about these > days? > > --linas > > -- > You received this message because you are subscribed to the Google Groups > "opencog" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To view this discussion on the web visit > https://groups.google.com/d/msgid/opencog/CAHrUA36MR7FJyUyeTjQtgnUufb%3DA7Kyd%3DTfJ3%3Dtp1k-JxVbr-Q%40mail.gmail.com > <https://groups.google.com/d/msgid/opencog/CAHrUA36MR7FJyUyeTjQtgnUufb%3DA7Kyd%3DTfJ3%3Dtp1k-JxVbr-Q%40mail.gmail.com?utm_medium=email&utm_source=footer> > . > -- You received this message because you are subscribed to the Google Groups "opencog" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/opencog/CAFj%2Bw-vzWbSXnLMmGm_VqZwt3Lk6MBBHCF1spqs%3D_D4QMFun%3Dw%40mail.gmail.com.
