Shop OBEX P1 Docs P2 Docs Learn Events
AI robotics — Parallax Forums

AI robotics

WaldoDTDWaldoDTD Posts: 142
edited 2006-01-13 20:29 in Robotics
Hey,
Recently I have just finished my first project in AI robotics and I used some parallax components in the system. I would like to know other people's opinons of the direction AI and robotics is heading since both are inextricably linked. I would like to hear if anyone else has thought about this. I mean we have come a long way since David Silver and Bill Gosper from MIT. I was amazed to find out that there are a few hackers from the Homebrewed computer club over in California which saw the Altair and more importantly the Apple. Just wanting to get a new view from other people on the topic.-Signol01

Comments

  • Bruce BatesBruce Bates Posts: 3,045
    edited 2006-01-13 03:54
    Signol01 -

    You say "saw the Altair and more importantly the Apple" like that was some kind of feat. Many of us owned them, and some of us ran Computer Bulletin Board Systems (BBS) using them. I happened to be one of the early TRS-80 users, and ran a 24 x 7 x 365 BBS from my home, from the fall of 1968 to the summer of 1986 (when I unfortunately got married <sigh>), beginning on a 16 K TRS-80 Model I with a 4 digit serial number, and ending with a 64K TR-80 Model III.

    Some of the folks on this forum wrote their own BBS software to run on these early machines. Yup, some of us dinosaurs are still alive and some of us STILL own those very machines of which I spoke.

    Regards,

    Bruce Bates
    Formerly of
    MCTBBS
  • Robert KubichekRobert Kubichek Posts: 343
    edited 2006-01-13 04:14
    Bruce Bates said...
    Signol01 -

    You say "saw the Altair and more importantly the Apple" like that was some kind of feat. Many of us owned them, and some of us ran Computer Bulletin Board Systems (BBS) using them. I happened to be one of the early TRS-80 users, and ran a 24 x 7 x 365 BBS from my home, from the fall of 1968 to the summer of 1986 (when I unfortunately got married <sigh>), beginning on a 16 K TRS-80 Model I with a 4 digit serial number, and ending with a 64K TR-80 Model III.

    Some of the folks on this forum wrote their own BBS software to run on these early machines. Yup, some of us dinosaurs are still alive and some of us STILL own those very machines of which I spoke.

    Regards,

    Bruce Bates
    Formerly of
    MCTBBS

    Hmmm, I started out with a TRS80 Model 1-2, and then went to a C64 with 3 floppies.turn.gif
    Then to an early Tandy 1k with a 27meg hd....I was "Mystery Manor BBS" back then , and I also did some trouble shooting for Exec-PC BBS "The largest call-in BBS in the world"....tongue.gif

    Those were the days! yeah.gif

    Bob N9LVU scool.gif
  • allanlane5allanlane5 Posts: 3,815
    edited 2006-01-13 14:54
    AI has several approaches.

    The first is programming a computer to do something a human used to do. As Parnas says "once we see how the program works, and understand the problem, we will not think of it as AI anymore".

    The next is the "Inference Engine" approach, also called 'rule-based'. This uses an 'inference engine' set of code, run against a set of 'rules', to make conclusions (and then maybe take actions).

    The next is the 'neural net', where a parallel linked set of 'nodes' are 'trained' (through changing the 'weighting' of signals passed through the node connections) to 'recognize' certain conditions.

    I haven't seen much in the literature lately on major advances in AI. The 'rule-based' approaches don't scale well -- the more knowledge you have, the more rules you have, the harder it is to prevent confusion and deadlock in the rule-base.

    The 'neural-net' approaches are also limited in application. Once you've trained your 'neural-net' to recognize something, that's all it's good for.

    I find the BS2, and the PBasic language, and the BOE-Bot, make a very nice testbed for the first approach, though. You add sensors, control motion, control LED's, even control a speaker. With all these resources, the relatively simple programming language can let you create a robot which 'seems' pretty intelligent.
  • Paul BakerPaul Baker Posts: 6,351
    edited 2006-01-13 15:57
    You should investigate Subsumption Architecture, it can produce very interesting results from relatively simple algorithms, in many situations it can approximate neural networks without the complexity involved with neural nets. Its one of those things Ive done brief research on and sometime want to pursue further.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    ·1+1=10
  • SN96SN96 Posts: 318
    edited 2006-01-13 18:00
    I did a search on Subsumption Architecture, and it all sounds very interesting. I wish there was a simple code example that one could possibly convert into a Pbasic program.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Mike

    ·
  • Bruce BatesBruce Bates Posts: 3,045
    edited 2006-01-13 19:38
    Mike -

    Try this on for size: http://www.bluebelldesign.com/FSM_explain.htm
    It LOOKS interesting any way smile.gif

    Regards,

    Bruce Bates
  • SN96SN96 Posts: 318
    edited 2006-01-13 20:29
    Nice find Bruce!

    I tried it on for size, and it fit nicly! smilewinkgrin.gif

    Thanks

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Mike

    ·
Sign In or Register to comment.