The Thing Growing - Diary Sept/Oct 98

Tues 20th Oct	Dan advice - don't go for thing autonomy Dave advice - don't get hung up on intelligence get to end of story ->so the next thing will be to model and do so rather than to work out the checker stuff ho-hum - makes me feel not so great, cos I had such hopes for its intelligence, but maybe it will still come but later! first just to fix those couple of trailing things rocks run can call from button press in scenefile not getting called from code xpNode not called from app !!! finding rocks to hide under - how to fix make a faux class rocDCS and just find these? added back all the current action stores to thingscene checked in newclasseses with alteredworld and makefile and I will check the reset as I put in the next scenes etc check sound levels - as I make new sounds check motions - as I add motions
Fri 16th Oct	networked thing OK make rocks run again!! I seem to see all the event occurred things in the right placed but...
Thurs 15th Oct	put the menuRelay stuff in teachdance altered the scenefiles - put menuscene inside thingscene.net networking seems OK - except we may want to change it so that clock simply times the amount of time we have doing the teachdance? why is the RUN.net loaded performer 2.1 libraries - how to fix? cavetest transition into and out of teachdance when I add the other scenes I lose the networked thing - it flashes reset
notes	What happens if the user does not stop for a second?OK but check in CAVE The recap update is clearly faulty - kicks in before the dance is done and therefore checked - ie change it! Reset and Networking and Cleanup N.B. If the recap_instruction_first is interrupted we go to recap_instruction and many of these say - we'll recap AGAIN or similar - maybe we'll need another action store or go to recap_instruct_first again(can't) finding too many avatarDCS's which aren't all rocks I will need also to test the individual stores to see if the actions are good, sound levels etc - I could do all this just insequence -need to check through and see what is temporary - what needs to get done (nextstepinstruction) new dance steps - but should wait cos with the new models I may need to make changes anyway - will also need sound
Tues 13th Oct	recheck and rethink why the checkUserDist function fucks up Made a runawayable as well as an interruptible - so that the checkUserDist at begin of new action only happens when previous action was nextdancestep, wedance or recap recap comply etc.- do I need to adapt checkCompliance so that comply is 0 and novment or badly are 1 and 2 the check_dance_ flag will reset the state after eg youdance so it will reset the state and instead of the usermovingtoomuch we get the dancedbadly(or whatever) it resets the state set from app - unset this flag if we are interrupting an action ven if I hit the n button I seem to get recap_complied
Fri	made a checkUserDistance function to replace both the userMovingTooMuch and hasUserRunAway functions - so now if the user has run off but not far enough we should get the the don't move away AS, and if they have really run we get the stop running AS. having some problems cos the interrupt send us back through the loop to get a new action, and the interrupted state gets changed wrongly -- if I've done the interrupt I don't want to do the checkUserDistance again. - another flag! Also still need to rethink only going to recapComplied perhaps I need the runafter not to go to RECAP, but too ...something dependent on interrupted state which is really last state!
Thurs	I think I'm not saving the correct interrupted state - by the time I save it, its already been changed by the updates! duh! runafter: after the hold up and the catchuser remonstration I'm getting a message that there are no more teach dance action - I believe that the counter of steps in list is not being set - not convinced this makes it work . To help me figure this out I need to know the numbers of the dancestates, assuming enum #s 1 - > RUN_AFTER_USER 0 DANCED_WELL 1 RECAP_INSTRUCTION_FIRST 2 RECAP_INSTRUCTION 3 RECAP 4 RECAP_COMPLIED 5 RECAP_FAILED 6 NEXT_DANCE_STEP_INSTRUCTION 7 NEXT_DANCE_STEP 8 YOU_DANCE_INSTRUCTION_FIRST 9 YOU_DANCE_INSTRUCTION 10 YOU_DANCE 11 DANCED_BADLY 12 DIDNT_MOVE 13 WE_DANCE_INSTRUCTION 14 WE_DANCE 15 CATCH_USER_REMONSTRATION 16 USER_MOVING_TOO_MUCH 17
Tues	danced - well happens twice also seems to get stuck in the recap, doing same motion again and again teh problem seems to be in what the recap complied is doing - ie not doing what it should and instead we get the last dance step again, after that we go on to NEXT DANCE STEP as we should I fixed it so it either goes to RECAP or to RECAP_COMPLIED when finished - 86ing RECAP_FAILED but I will need to make a comply checker that checks a series of movements as well as one that checks one action - and this will feed into danceupdate which can have two "states" or more recap_check action_check etc does not go to RECAP_INSTRUCTION but goes to RECAP_INSTRUCTION_FIRST again - firstrecapinstruction flag not unset sticking in stay close to me after Run Away says there are nomore teachadance actions rethink the two checks userRunning Away and UserMoving Away - maybe I go from one to other dependent on distance - meanwhile I wiped the userInfo call and replaced with a distance check
Tues 6th Oct	Check all dancestates in all moods state hap angry sad manic reset networked RUN_AFTER_USER X X X X DANCED_WELL X X X X RECAP_INSTRUCTION_FIRST X X X X RECAP_INSTRUCTION X X X X RECAP X X X X RECAP_COMPLIED X X X X RECAP_FAILED X X X X NEXT_DANCE_STEP_INSTRUCTION X X X X NEXT_DANCE_STEP X X X X YOU_DANCE_INSTRUCTION_FIRST X X X X YOU_DANCE_INSTRUCTION X X X X YOU_DANCE X X X X DANCED_BADLY X X X X DIDNT_MOVE X X X X WE_DANCE INSTRUCT X X X X WE_DANCE X X X X CATCH_USER_REMONSTRATION X X X X USER_MOVING_TOO_MUCH X X X X Check reset! networked all states all moods sound volume, crackles - and whether any of the complied(danced_well) soundbites mention doing the next step (I know the recap complieds do) Don't forget how I added stuff to thingscene so it could run solo think about testing things modularly
Mon 5th Oct	run RUN in src3 - watch the errors, wait for the crash make sure all of the actions in all actionstores and classes are working work out where you are vis a vis all problems globalbodies need some more-thinking ie. when to move and when to do nothing - given our new state to state logic Errors!!! finding too many avatarDCS's which aren't all rocks brain doesn't have a global body? I see references to tgStory instead of going into teachdance I get the "nomore teachdance actions" error getting the dance-badly comment twice not responding to the keyboard yes/no but I seem to have the moods don't forget you have to add the next-step-instruction store etc - i need sound and motion bites for it so I need a stand in it might be a good idea to make a check list of things that have to work
Fri 2nd Oct	the aim is to get the newbrain class up and compiled make scenefile missing if the user moves away or runs away too much ->sulknhide networking I removed the brain and story from world and makefile removed tgGlobalBody 's pointer to the brain moved xpUser.h from the cxx to the h of userInfo
Tues 29th Sept	My todo list today finish typing in the teachdance class, add to my local Makefile and a local world, make and fix the errors. while I still have a networked version - I need to be aware that I may have to deal with it from the teachDance class however, instead of using the clock to let people know when they can change the badly/well/notdancing parameter - maybe I use it to let people know how long the teach dance period is and when to start getting mad! I believe that it needs a pointer to the brain and vice-versa so I'll have to do that "class" declaration thing to get round this causing errors. also what happened to catchUserMax?? check that its working - which means ... change the brain to fit in with the teachdance class - chmod to get local version of brain I have opted to make a new class newBrain now it basically gets its action from the class - internal to the class is the decision making process. "interrupt" message functionality must be added All the other stages so far, simply serve up an action in sequence and remain actionstores. The teachdance class should serve as a model for other classes which have to do decision making before they can serve up the appropriate action. finally change brain so as to incorporate the story. I still have to do some alterations to other classes and scene files globalBody - needs a stay still, a stay near(not be glued to) userInfo - need to check how UserMovingTooMuch is being assessed.
Mon 28th Sept	Internet2 Fall meeting Part of netcast - showed Julie tapes, thing video, and thing sirius out showing Tuesday too
Week of Sept 21-26	Worked on Thing models with Joe - notes I need to make sure that the new bits "work" with the old movements then: make a smaller innner head - maybe less transparent in base green color. Maybe this inner head chages color subtly: lighter/darker the outside head will change color when speaking flipping between yellow and green for happy red and green for for angry blue and green for sad flip all colors plus white for manic I'll have to adapt "flasher" class to do this And I'll have to be able to send an event message? to the parts so they'll do this when ordered to The four things can be in the four colors (which four) - so I need arms and body and tail in all colors I had also thought of putting inside another shape that moves up and down Worked on Brain - make teachdance into a class merge brain and story Put some of the app to tape - sirius out from the networked POV and from a "second" networked user - ie we see the thing and the user's avatar but not the menu, and lastly from user POV
Friday Sept 18	Lessons learnt at the AEC festival. The Thing models needs work Some people see tail not arms and head and try to copy it, swaying to and fro like snakes! (yellow color of tail comes forward and blue of head, body and arms receed.) Head needs to be more obvious a HEAD - I have asked Joe Alexander to help me with the head - giving a new texture map plus a chin that wobbles? The way that the thing chases after the user needs refinement People would try to drive closer with the joystick when it asked them to stay close - they were confused that driving was then disabled - also they would still be driving when the thing stopped that soundbite and driving was re-enabled, which the Thing would take as further proof that they were trying to get away. That the Thing is reset to a different place in its logic when the user runsaway (the user can run when the thing is dancing), than when the user moves away (the thing detects moves while it is conversing with the user and checks them) - this is confusing. (says Dave) Along with this Dave says that the GlobalBody move that pastes the thing to the user is bad. Things to change in scene files some people click on key immediately and this freezes everything up and the key doesn't go to the box before the cattaswoosh starts Creak says the door repeats Tail not attached to body properly - this for Dave to fix. The rocks are not perceived as threatening and people do ot try to run from them - one even asked Do you want to dance? Networking Problems the rocks don't reset right the networked user changes viewpoint too randomly, it should always be at the cattaswoosh FTP from AEC Dave compiled here at EVL - so I need to get his executable etc. Need just to get the scenes files from AEC - cos these we changed. Other Things on the Todo List Add the next scene Models for 4 things, new plain, black hole, flashlight, gun and the blocks the thing will fabricate for the escape Text/Dialog How will the rending of the earth occur? Sound get David Nahon's sound server for positional sound process the Thing's voice record 4 Things record Thing Motion capture motion for 4 things and thing Brain reconceive User Checking code I think these tests with the networking show how subtle this checking has to be! Dave mostly drove the networked user, and basically would encourage very minimal movement at the beginning - then get picky and sad as the thing went on - however well the person was doing. So the problem is not just writing a checker that works well enough, but then deciding how best to use the info. Video Spring make video Show it where when - applications to conferences etc
AEC Sept 7-12	Showtimes Wed 9th and Fri 11th - 12:30 - 4:30 Publicity/Reviews Kirsten - Swedish freelance ?? - Italian TV Peter ?? - German Magazine ORF - AEC documentation ?? Bavarian woman I was pleased with the show - it was a little too dependent on the user doing it right! The best users were those with moderate VR/Video game knowledge. Those who were really into games assumed they had to do everything with the joystick. Those who had never used a joystick couldn't drive. Very dependent on the user "getting" the English. It showed us things that have to get fixed - see above. But I think it went well enough for a work in progress. I think some people expected that there was a lot more in there. Horst asked if we could leave it, but Dave said it was too dependent on the networked user knowing the program well. We should send it back when its autonomous!