TriWowbot - Hacking a propeller into WowWee products

Comments

• zappman Posts: 418

  2012-11-17 16:24 edited 2012-11-17 16:24

  Thanks for sharing your robot disassembly and rebuild. I own the exact same red WowWeee Red TriWowBot, but I have never taken it apart. I would like to know more details about your robot; schematic/wiring and code. Great job!!!
• Gareth Posts: 278

  2012-11-18 06:36 edited 2012-11-18 06:36

  Code_ing now inserted above..... its a first Hash version , highly stable it is (however still too linear for my like_ing)
  The key to the whole motor control uses the superb PWM2C_HBDEngine.spin OBJs.(i use just three ....one spare haha)
  
  My XBee TX unit is a radio_control box i hacked - it just reads the 4 analogue stick potentiometers and two switches and spits this serial info on the XBee transmit pipeline. The propeller simply uses the PST object to gleam the rx out of the RX XBee. (at the end of code you see i dedicated an extra cog just to do the XBee data collection).
  
  Unfortunately i cant add any more pictures (ie schematics) to this blog due to a 10 picture limit with the editor . ;-(
• Zoot Posts: 2,227

  2012-11-18 09:03 edited 2012-11-18 09:03

  Very sweet hack.
• Duane Degn Posts: 10,588

  2012-11-18 10:46 edited 2012-11-18 10:46

  Gareth, as usual for you, another great project. Man you make a lot of cool stuff!
  

  Gareth wrote: »

  Unfortunately i cant add any more pictures (ie schematics) to this blog due to a 10 picture limit with the editor . ;-(

  
  Why not just add some more pictures to later posts? Like post #6?
  
  With a little trig I was able to get my three wheeled omni robot (post #44) to travel in any direction.
  
  Here's some of my code from '09 (I like to think I could do it better now).
  

  x := _Stop - pulseWidth[1] + 9                                        
        y := pulseWidth[2] - _Stop - 6   
    
        z := _Stop - pulseWidth[3]
        x *= 2
        y *= 2
        z *= 2
        
        fX := FMath.fFloat(x)
        fY := FMath.fFloat(y)
        fTargetSpeed := FMath.FSqr(FMath.fAdd(FMath.fMul(fX, fX), FMath.fMul(fY, fY)))
        if x == 0
          if y > 0
            fTargetAngle := FMath.fDiv(f180r, 2.0)
          else
            fTargetAngle := FMath.fDiv(f180r, -2.0)
        else
           fTargetAngle := FMath.ATan2(fY, fX)
          
        fTargetFront := FMath.fMul(fTargetSpeed, FMath.Cos(fTargetAngle)) 
        fTargetStar := FMath.fMul(fTargetSpeed, FMath.Cos(FMath.fAdd(fTargetAngle, f120r))) 
        fTargetPort := FMath.fMul(fTargetSpeed, FMath.Cos(FMath.fSub(fTargetAngle, f120r)))
        targetAngle := FMath.FRound(FMath.degrees(fTargetAngle))
        targetWheelSpeed[0] := FMath.FRound(fTargetFront)
        targetWheelSpeed[1] := FMath.FRound(fTargetStar)
        targetWheelSpeed[2] := FMath.FRound(fTargetPort)
        targetTurnSpeed := z
  
  

  
  As you can guess from my code, I call the three wheels font, star (for starboard) and port. Most directions all three wheels move.
  
  I'm trying to get my omnibot up and running to show to family at Thanksgiving so I'll be working on it today.
• erco Posts: 20,250

  2012-11-18 12:23 edited 2012-11-18 12:23

  Sweet job, Gareth. Funny, I just busted out all my WowWee robots this morning (before I saw this post) and my twins were playing with Tribot, RoboQuad, RoboRaptor, and RoboReptile. The girls liked dancing with Roboquad in demo mode best, they hated Tribot. Kind of obnoxious, I'll agree.
  
  I also have a Rovio. They did a great job on that one.
• zappman Posts: 418

  2012-11-18 12:26 edited 2012-11-18 12:26

  Hi Gareth (chiprobot),
  
  Thanks for looking at my YouTube robot videos and subscribing to my channel.
  I subscribed to your robot videos.
  
  Regards,
  
  zappman
  
• Gareth Posts: 278

  2012-11-19 07:18 edited 2012-11-19 07:18

  erco wrote: »

  I just busted out all my WowWee robots

  Sounds like you would profit from a lipo battery hack at the least....... from what i remember the Tribot (i did test it ;-) liked eating batteries.
• Gareth Posts: 278

  2012-11-19 07:44 edited 2012-11-19 07:44

  Duane Degn wrote: »

  With a little trig....

  
  Thanks for the code ...and.....Yes this is my next desired code iteration direction...... Kinda "Tangent" control (i think) .. Will have to brainstorm your maths and see how i can adapt...
  
  Ah-ha yes good dodge around picture limits ...... i am still a newbie around here :-)
  
  First weighted tabled control Algorithm - Fail
  

  

  

  
  
  H-Bridge - pre soldering cigarette packet schematic.....
  

  

  

  

  Tribot 010.jpg

  640 x 480 - 158K

  

  

  Tribot 011.jpg

  640 x 480 - 147K

  

  

  Tribot 009.jpg

  640 x 480 - 174K

  Tribot 010.jpg 158.2K

  Tribot 011.jpg 147.1K

  Tribot 009.jpg 174.4K
• Duane Degn Posts: 10,588

  2012-11-19 09:30 edited 2012-11-19 09:30

  Gareth wrote: »

  First weighted tabled control Algorithm - Fail

  
  I'll see if I can explain the control algorithm I came up with.
  
  First convert joystick reading to x and y values. Positive y is forward, positive x if to the right (like high school geometry). I think in your case you'd subtract 127 from the joystick value and change the sign if needed.
  
  Now that you have your x and y values convert them to floating point numbers with F32.spin (a great object).
  
  You can compute the direction you want to travel with the Atan2 method. Unlike the Atan method (which returns an angle between -90 and 90) the Atan2 method returns any angle between -180 and 180 (full circle). The Atan2 method takes two arguments, y and x, and returns the angle (I'm using degrees here but F32 uses radians, you have to convert to degrees at the end if you want to see the angles in degrees).
  
  Now that you have the desired angle of travel you also want to know how fast to travel. I computed speed based on how far the joystick was from the center position.
  
  In the code above I called this speed "fTargetSpeed" (I used the "f" to remind me it's a floating point number). I think the code shows how "fTargetSpeed" is computed by taking the square root of the sum of the squares of x and y.
  
  One thing I didn't do well with my code and it's something I plan to change is the way this target speed was scaled. In the code above I just doubled the x and y values but these should really be scaled based on the value used by the speed control method.
  
  I think in this case the target speed would be scaled by multiplying by the maximum speed the drive method accepts (I think it's 1000 for the object you're using. and then dividing by the maximum joystick deflection (I think this is 127 in your case).
  
  Once your target speed has been scaled, will also want to compute your rotation amount (based on the second joystick). I called this "z" in the code above. z should be zero when the second joystick is centered and positive when pushed to the left (in the right hand coordinate system we're using).
  
  You can scale z the same way as you scaled the target speed. I think I called the scaled z value "targetTurnSpeed." Looking back at my code I see all I did to scale my speeds was to multiple them by two, which worked with the joystick and motor controllers I was using.
  
  This "targetTurnSpeed" value will be added to the calculated wheel speeds later.
  
  Each of the three wheel speeds is then calculated based on the angle the wheel makes to the desired direction of travel.
  
  With the coordinate system we're using, zero degrees is to the right and straight ahead is 90 degrees.
  
  Here's the code for computing these three speeds:
  

  fTargetFront := FMath.fMul(fTargetSpeed, FMath.Cos(fTargetAngle)) 
        fTargetStar := FMath.fMul(fTargetSpeed, FMath.Cos(FMath.fAdd(fTargetAngle, f120r))) 
        fTargetPort := FMath.fMul(fTargetSpeed, FMath.Cos(FMath.fSub(fTargetAngle, f120r)))
  

  
  The variable "f120r" is the angle 120 degrees as a floating point number that has been converted to radians ((2*pi)/3 I think).
  
  Lets say we want to the bot to travel to the right (zero degrees). Cosine of zero is one so the font wheel would turn at the full target speed with the two other wheels turning at Cos(120) or 1/2 this speed (these two will rotate in opposite directions to each other).
  
  After we have these three speeds you need to add the turning speed "targetTurnSpeed." The problem now is the combined speeds from moving and turning can easily exceed the maximum value so the greatest absolute speed needs to be found and scaled back to the maximum allowed. The other two speeds will also need to be scaled. The adding of the rotation speed and scaling the final speed isn't included in the above code.
  
  Hopefully this is enough to get you started, I'd be glad to help in any way you want.
• Gareth Posts: 278

  2012-11-20 10:10 edited 2012-11-20 10:10

  Duane Degn wrote: »

  You can compute the direction you want to travel with the tan2 method. .

  Many thanks again for the detailed explaination.....luckily i am familiar with the float32.obj sin and cosine functions...
  ..now time to grasp the "Tan"ning process (hopefully i do not get sun blocked ;-)
• Duane Degn Posts: 10,588

  2012-11-20 10:43 edited 2012-11-20 10:43

  Gareth wrote: »

  luckily i am familiar with the float32.obj sin and cosine functions...

  
  Gareth,
  
  Sorry but I miss typed. I typed "tan2", when it should have been "Atan2". We don't want the tangent, we want the arctangent. I'll fix my previous post.
  
  "F32" is a new and improved floating point object.
  
  F32 is a drop in replacement for the previous floating point objects. The expanded floating point object "Float32Full" used two cogs. F32 does all that Float32Full did with a single cog and runs faster than Float32Full.
  
  You shouldn't need to change any of your code when switching to "F32" from "float32."
• boboi Posts: 1

  2016-01-30 10:00

  Hi, what is the motor voltage ? I tried powered mine with 12v and they produce smoke