I'm working my way toward that, but sadly still on the construction phase of mine. I got new bearings, timing pulleys and belts, and a batch of new wood. I've started doing the layout of the parts in my CAM software, so hopefully I'll have something else to show soonish. Not spending most of my weekends at home seriously limits my hobby time.
I'll be home from vacation in a week and I'll program my arm to do a figure 8. Currently near Cincy OH and wondering whether to visit Wright-Patt AFB museum or drop in unannounced on mindrobots in Medina.
Of course I'll be disappointed if Duane or Martin_H don't take a shot at it before then.
Currently near Cincy OH and wondering whether to visit Wright-Patt AFB museum or drop in unannounced on mindrobots in Medina.
If I were looking for something to do, I'd go to Wright-Patt rather than visit me...I've never been to the museum but I've spent a lot of time with me!
There's always Orrville, too! The home of Smuckers AND PropellerPowered's International headquarters!!!
Jason, good job and the wood in that robot arm is great looking! May I ask what algorithm you used?
For my robot arm figure eight I used IK with two tangent circles. But this resulted in a glitch when looping around to the start because my final value for circle two didn't match the start of circle one.
That looks to me like a "real" figure 8 shape with some straightaways in the middle, as opposed to the two tangent circles that most of us have done. Beautimous!
I pre-compute 180 points along that path. For each point I use float IK to figure out the arm orientations required, and from those I compute the number of steps required to go from my current motor positions to the computed ones. Those step counts are stored in an array.
I continuously play back the moves in the array using a modified Bresenham algorithm. I run 500 "ticks" for each move in the array, and the Bresenham code sends out the correct number of step pulses to each of the 3 motors to move them the required distance during the incremental move.
It would be more correct to have the code make sure that my tick count can handle the number of steps required for that move, and use a higher number of ticks if necessary. At present, if a move required 1000 steps, my algorithm would only make 500 of them for that axis - It can step a maximum of once per motor per tick. I chose a tick count that was much higher than I'll ever move for this particular demo so that the motors wouldn't jitter much.
What I want to try is having the Bresenham routine interpolate not only the position values, but also the step frequency. Then I could use a single Bresenham run to accelerate or decelerate a motor between two points, so I could space the sample points out much further, but it needs to be *really* accurate so I don't lose or gain steps during a move.
I thought a figure-8 was hard enough with a pair of continuous rotation servos, sheesh.
Martin and Jason take it to a whole new level with all those moving parts. :thumb:
This is a video of my two-wheel robot, Wilson, attempting a figure-eight. It's a little bit sloppy, but considering he's only two wheels and the equipment cage rocks back and forth like a swing in the wind, I'm pretty pleased with it!!!! http://www.youtube.com/watch?v=JGC8-...ature=youtu.be
Well it should be a rule.. You guys need rules to be kept in check..
How can Amanda be a rebel if there are no rules to break?
As for me, I will toe the line and:
Start in the middle of the plates,
Make the first circle clockwise,
Stop for one or more seconds at the exact starting location,
Make the second circle counter clockwise,
Stopping at the exact starting spot.
Now we have some rules just right for the breaking!
Let the rebellion begin...
My Wild Thumper will be here tomorrow. After I get the basics going I plan to enter this challenge. Has anyone done the figure 8 with anything larger than a Boe Bot? Any suggestions or tips?
Has anyone done the figure 8 with anything larger than a Boe Bot?
Actually I think the majority of the robots in this thread have been larger than a Boe Bot. I think erco's Mecanum wheeled robot is probably the largest of the thread (IIRC).
I offered some tips on getting a robot to perform a figure 8 in post #35 of my cheap bot thread.
I looking forward to seeing a wildly thumped figure 8 in the near future.
So I've been working on the figure eight challenge with the drawbot. After several dry runs without a marker I did one run with a marker:
As you can seethe second loop was a bit disrupted by the wheel contacting the edge of the paper. But I am out of paper bags, so I'll have to post this incomplete run. Here's the FORTH code which uses Nick Lordi's robot command language.
\ start the drawbot by initializing the servo library and the
\ robot command language. The 0 GF halts the robot.
: start_drawbot
sm_start_servos START 0 GF ;
\ pen up using sevo on pin 0
\ The hex magic number is the servo pulse width determined
\ by trail and error
: PU
0 h15E0 sm_setpos 250 delms ;
\ pen down using servo on pin 0
\ Again, the hex constant was determined experimentally.
: PD
0 hCE4 sm_setpos 250 delms ;
\ Pen down for one second, pen up for one second
: PDU
PD
960 delms
PU
960 delms ;
\ put pen down and then up n times to delay
: DO_N_PDU \ ( n - )
0 do
PDU
loop ;
\ A figure eight
: FIGURE8
0 GF \ zero both wheel velocities
5 9 run \ drive the left wheel faster
17 DO_N_PDU \ pen down and up while loop is completing
0 GF \ zero both wheel velocities
9 5 run \ drive the right wheel faster
17 DO_N_PDU \ pen down and up while loop is completing
0 GF ;
Sweet fancy Moses! Great demo. I suppose we'll call that an "open loop" figure 8 for several reasons, including the fact that no P5587 sensors were used in the making of that bot or video.
Comments
Of course I'll be disappointed if Duane or Martin_H don't take a shot at it before then.
There's always Orrville, too! The home of Smuckers AND PropellerPowered's International headquarters!!!
Your Ohio Fifth Third bank name is annoying enough, and the fact that they just totally skipped Forth is adding insult to injury, huh?
UNLESS you visit the Fourth Street branch...
Name: Fifth Third Bank, Banking Centers, Fourth Street Banking Center
Street: 30 East 4th Street
Cincinnati, OH 45202-3704
Phone: (513) 579-5565
How's this?
Great work, and (per others), neatest sound on the block.
For my robot arm figure eight I used IK with two tangent circles. But this resulted in a glitch when looping around to the start because my final value for circle two didn't match the start of circle one.
Sweet!
@
I also love this high-speed delta robot's orbital antics (somewhat figure-8-ish) at the 26-second mark and also at 1:13 : http://www.youtube.com/watch?feature=player_detailpage&v=Gv5B63HeF1E#t=26s
I assume you're asking what path it's following?
x = cos(t) * sin(t) * 6.0;
y = sin(t) * 12.0;
http://fooplot.com/plot/b7n1s42t82
I pre-compute 180 points along that path. For each point I use float IK to figure out the arm orientations required, and from those I compute the number of steps required to go from my current motor positions to the computed ones. Those step counts are stored in an array.
I continuously play back the moves in the array using a modified Bresenham algorithm. I run 500 "ticks" for each move in the array, and the Bresenham code sends out the correct number of step pulses to each of the 3 motors to move them the required distance during the incremental move.
It would be more correct to have the code make sure that my tick count can handle the number of steps required for that move, and use a higher number of ticks if necessary. At present, if a move required 1000 steps, my algorithm would only make 500 of them for that axis - It can step a maximum of once per motor per tick. I chose a tick count that was much higher than I'll ever move for this particular demo so that the motors wouldn't jitter much.
What I want to try is having the Bresenham routine interpolate not only the position values, but also the step frequency. Then I could use a single Bresenham run to accelerate or decelerate a motor between two points, so I could space the sample points out much further, but it needs to be *really* accurate so I don't lose or gain steps during a move.
Thanks. I thought you were kidding when you were asking for a robot arm figure eight right after I posted one.
@Jason thanks for the description of your technique.
Martin and Jason take it to a whole new level with all those moving parts. :thumb:
-Tommy
http://www.youtube.com/watch?v=JGC8-...ature=youtu.be
Building Wilson...
http://forums.parallax.com/showthread.php/148717-Two-Wheelin
Thanks for sharing!..
Oh, and +1 for breaking the rules, With not stopping in between loops. You're a rebel..
-Tommy
That's a rule? Yikes.
It's incidental, anecdotal, and optional!
How can Amanda be a rebel if there are no rules to break?
As for me, I will toe the line and:
Start in the middle of the plates,
Make the first circle clockwise,
Stop for one or more seconds at the exact starting location,
Make the second circle counter clockwise,
Stopping at the exact starting spot.
Now we have some rules just right for the breaking!
Let the rebellion begin...
-Tommy
I suppose a stop aids in demonstrating you stopped a single loop where it began but I personally think not stopping looks better.
I usually try to add some ramping during the transition so it's not too abrupt. (Ramping is also helpful at the beginning and end of the figure 8.)
+1000 (That's binary for plus 8)
-Lawless savage
Next, I want to see a Quadrover tearing up your kitchen linoleum in a figure 8...
besides, You've seen the frenzy. http://forums.parallax.com/showthread.php/148990-Quadrovers.-The-Final-Chapter.
the only hope of seeing a Quad Rover on my kitchen floor, is if you bring your brand new one...:)
Actually I think the majority of the robots in this thread have been larger than a Boe Bot. I think erco's Mecanum wheeled robot is probably the largest of the thread (IIRC).
I offered some tips on getting a robot to perform a figure 8 in post #35 of my cheap bot thread.
I looking forward to seeing a wildly thumped figure 8 in the near future.
As you can seethe second loop was a bit disrupted by the wheel contacting the edge of the paper. But I am out of paper bags, so I'll have to post this incomplete run. Here's the FORTH code which uses Nick Lordi's robot command language.
\ start the drawbot by initializing the servo library and the \ robot command language. The 0 GF halts the robot. : start_drawbot sm_start_servos START 0 GF ; \ pen up using sevo on pin 0 \ The hex magic number is the servo pulse width determined \ by trail and error : PU 0 h15E0 sm_setpos 250 delms ; \ pen down using servo on pin 0 \ Again, the hex constant was determined experimentally. : PD 0 hCE4 sm_setpos 250 delms ; \ Pen down for one second, pen up for one second : PDU PD 960 delms PU 960 delms ; \ put pen down and then up n times to delay : DO_N_PDU \ ( n - ) 0 do PDU loop ; \ A figure eight : FIGURE8 0 GF \ zero both wheel velocities 5 9 run \ drive the left wheel faster 17 DO_N_PDU \ pen down and up while loop is completing 0 GF \ zero both wheel velocities 9 5 run \ drive the right wheel faster 17 DO_N_PDU \ pen down and up while loop is completing 0 GF ;May the Forth be with you!