WideTrack
From TheRobotGroupWiki
Contents |
WideTrack Robot Page
WideTrack at ArmadilloCon 24 Aug 16-18, 2002
Background
I started the WideTrack project as an engineering experiment in "easy to build" tank tracks in early 2000. Typically robot tank tracks fall into 3 categories:
1. Custom machined links 2. Automobile Timing Belts 3. Hacked Toy Tanks
None of these solutions appealed to me. I didn't want to learn to custom machine parts. Using timing belts didn't appeal to me and toy tanks are too small. I thought that there had to be an easier way to utilize an easy to work with material such as pvc pipe that would give me the results I desired. I wanted a material that was:
a. readily available b. did not require special equipment or tooling to work with c. would work like a "real" tank track linkage d. would scale up to large sizes
In my naivete, I thought I could knock something out in a couple of weeks. Two years later, I am still perfecting the design. In that time I have learned a few things about what works and a whole lot about what doesn't work. This has become a journey for me, not a destination. I thought that others may be interested in that journey too. This journey has not been entirely alone. I would like to thank Ed Gonzalez, Derek Bridges, Glenn Currie, Tom Davidson, Edwin Wise, Don Colbath, Tom Morin, Bill Craig, Aaron Lundquist, Catherine Lundquist, P.Y., Gray Mack and the rest of the Robot Group for their help and support in getting me this far.
Thanks!
EricLundquist Austin, Texas September 2002
What do the "Big Boys" do?
I started out by taking a trip with my camera to Camp Mabry. Camp Mabry is the headquarters for the Texas National Guard and they have a great museum on base. If you ever get a chance to go there, it is definitely worth the trip. Of particular interest to me is their fine collection of WWII armored vehicles. They have everything from German half-tracks to American mud-buggies to Patton tanks. What was most interesting to me was that no two vehicles seemed to have the same track design. This taught me that there is no one optimal design for tracks. Clearly there was as much art as science in the design of track linkages. When I ran across this vehicle, I knew that this was something that I could copy and use for my purposes.
A Real Tank Track
One interesting factoid that I discovered is that all of the drive sprockets seem to have an odd number of teeth. German and American, WWII and modern vehicles all seem to follow this rule. I am not sure why this is, but who am I to argue with the "big boys"? Here is my version of the above track design on an early frame.
Building the tracks
An initial prototype of the "bogie" built entirely from cpvc pipe next to an intermediate design that still utilized cpvc for the axles. I eventually discarded the idea of the pvc in any of the framework simply because it wasn't strong enough.
The sprockets are simply pine shelving that were hand cut with a skill saw. They were then epoxied together. Later, the exterior of the sprockets were covered with fiberglass cloth and epoxied for strength. This has worked surprisingly well throughout the entire project. It is one of the few things that has not had a material failure. In the background you can see that the pvc axles have been replaced by all-thread. At this point the only pvc remaining is the tracks and the motor support frame. The side panels are constructed out of plywood that has been fiberglassed on one side.
A couple of incredibly bad ideas I had was to use rubber tie downs as drive belts on wooden pulleys that I built using hole saws. The wooden pulleys ate rubber belts for breakfast and were very difficult to get tight enough. The drive mechanism was later changed to a chain drive. You can also see the TRAXXAS XL1 Electronic Speed Controller in the background that I used for a short time. I think I can confirm that they do not work well with 12V 70Amp wheelchair motors! After letting the "magic smoke" out of two of them and melting my 18 gauge Radio Shack wire, I changed to 12 gauge automobile wiring and built a simple mechanical h-bridge out of micro-switches and servos. In the foreground, the motors are still mounted to pvc. The entire frame had a tendency to twist when the motors engaged. The frame was later rebuilt using steel plumbing pipe (see ArmadilloCon photo at top).
A poor mans high current H-Bridge. Needed to come up with something when I let the "magic smoke" out of my Traxxas XL-1 speed controllers. I got this idea from the old book "Build a Remote Controlled Robot for Under $300". The servos trip the micro-switches to allow forward and reverse. The biggest problem is that this is an off/on scenario. At +24V WideTrack really jumps when the motors kick in!
The "new and improved" drivers for WideTrack. Replaced those ugly mechanical switches with an electronic h-bridge. In the upper left is power distribution for Ground, +5V, +12V, and +24V. Upper right has the Futaba R/C reciver. The receiver feeds into the Atmel STK200 board in the lower right. The STK200 decodes the r/c pulses and generates a 1KHz PWM to the Tecel h-bridges on the lower left. In this photo one of the h-bridges is removed. Note also two muffin fans to help keep things cool.
Track Coverings
I knew early on that I would have to cover my pvc tracks with something. The first material that I used was toolbox liner. This was a rubber mesh that was fairly strong and easy to attach to the pvc tracks with tie wraps. This worked ok for a short while but it quickly started to wear and I knew that something stronger was needed.
My Daughter Catherine and I posing with an early version of WideTrack.
While wandering around at the local Lowe's, I stumbled across a roll of carpet protector for $5.00. The stuff is very tough so I gave it a try. I spray painted one side and attached it to the pvc with pop rivets. It was better, but still not there. It just didn't hold up while driving around on the asphalt driveway in front of the Robot Group warehouse. Something even tougher was needed.
WideTrack sporting the carpet protector plastic covering.
On one of the Thursday night get togethers, someone brought in a sample of the belting material used in conveyor belts and gave me the name of a local source. I went to Purvis Bearing. Once they got around the giggle factor of robot tank tracks, they were very helpful in locating a suitable material. Single ply pvc 120lb conveyor belt is what I needed. They ordered it from their San Antonio location and I had it a couple of days before ArmadilloCon. A couple of very late nights later, the new tracks were ready.
WideTrack and a couple of admirers at ArmadilloCon2002. A couple of other things of note here. Notice the addition of another battery on top. Going to 24V gave me full power on the wheelchair motors. Also notice the addition of the clear acrylic disks (thanks Ed!) on the sides of the wheels. This helps keep the tracks from skidding completely off the robot when it turns.
Videos
Catherine driving an early 12V version of WideTrack (2.8MB) Taking the 24V version for a test drive (2.1MB) WideTrack Cruising at ArmadilloCon 2002 (2.4MB)
Latest Work
(Sept 2003) This is one of the Tecel D100 h-bridges that I am in the process of repairing. Added a socket to the PAL in case we blow it again. Added as much heat sinking as I could fit on the MOSFETs.
(Sept 2003) Underside of the D-100 board. Added three 0.1uF filter caps across the motor terminals.
(March 2003) WideTrack tricked out with ghetto string lights. Added these to make it more visible for the InsomniacwithDaveAttell video shoot.
Adding a U-shaped slot to the pvc tracks
One section at a time.
How the slot fits into the new guide wheel. The guide wheel was made from Medium Density Fiberboard that was then covered with fiberglass and epoxy on both sides. A bearing was added to the wheel and it rides on the main wheel shafts in between the tires.
P.Y. grinding the little pvc support tabs in the U-channel.
Going Forward BLOG
(September 2005) Sadly, Wide Track is no more. I scavanged the wheelchair motors for use on another project. However, I recently learned about this group of tracked vehicle enthusiasts. VERY COOL and well worth a look!
(December 2003) WideTrack rides again! Got the second controller hooked up and it seems to work. DerekBridges has the Tecel D-100 controllers slightly modified for additional protection. He added some filter caps, zener diodes and heat sinks to the board and they work great. Derek also put a scope on the Futaba radio receiver and noticed that the signal we get from it goes from 1.3-1.7ms. This is significantly different from the theoretical range of 1.0-2.0ms. I think that explains the +/-50% issue with the Atmel decoding software. I'll need to step back and rethink how to address this in the software. In the meantime, I just altered my PWM table to bring the range in. It now goes -100%, -90%, -60%, -30%, 0, 30%, 60%, 90%, 100%. Derek also put a scope on the Atmel to see my PWM signals coming out. He thinks I may actually have a 100khz cycle instead of a 10khz cycle. Even at full throttle, it doesn't look like we are driving the motors at full bore. Again, I'll have to go back and look at the code. I may have to investigate using the hardware PWM on the Atmel. Despite these problems, WideTrack is mobile once again. Thanks to Louis from Tecel and Derek from The Robot Group! Without their support, this great progress would not have happened.
(November 2003) I think we are on to something with the controllers. The code and cabling have been changed to use all 3 control lines. A quick test was very promising. Forward and reverse both seemed to work smoothly. After about 15 minutes of playing, the mosfets were still intact and cool to the touch. At this point, WideTrack is still up on blocks so the motors weren't being stressed at all. We will need to get the second controller hooked up to test running both motors at once. I also need to tweak the code a little bit to capture the full range coming from the r/c radio. The software is only giving me +/-50%. I remember reading that you must sample at twice the rate that you are trying to measure. In this case, I don't think I can sample any faster with the 4 Mhz processor, so instead I'll need to reduce the number of time slices to be measured from 20 to 10. This will give me a throttle in 20% increments instead of 10% increments but it should be workable.
(November 2003) We think we may have made a breakthrough on the controllers. After much experimentation and many blown MOSFETS, Derek now believes that he understands why we keep blowing them up. You see, I had tied the ENABLE line high so that the controller was always enabled. I controlled speed with PWM'ing the FORWARD/REVERSE lines. Derek thinks that using a FORWARD and BRAKE sequence is causing a condition known as "shoot-through" that is giving the MOSFETS indigestion. I'll need to rewrite a little code for the ATMEL board and rig up some new cabling to use 3 lines instead of 2. Hopefully after the Thanksgiving holidays we'll be able to try again.
(September 2003) Derek replaced the blown PAL and MOSFETS and added filter caps to the D-100 board. The board still only operates the motors in one direction. Derek suspects that we have a short somewhere on the board. Hopefully we will track the problem down quickly.
(August 2003) I've learned a lot about h-bridges since the excitement of blowing up my controllers. Repairs are underway. Derek has really been a savior in diagnosing the failed components and replacing them for me. I have also been in contact with Louis at Tecel who has been most helpful in sending me replacement parts. Louis also offered some suggestions for my control software that I have incorporated. These changes should make for a kinder gentler PWM. I'll post the souce code when I am sure that I have it debugged. At this point on one controller board, we have replaced 2 MOSFETS and added some small heat sinks to them. We had to get creative with tin snips to get the sinks fit on the board but some has to be better than none. I bought some 20 pin narrow sockets to place in the PAL location just in case we have any future mishaps. Should get the replacement PALs in a few days and hopefully we'll have a working controller.
(July 2003) I finally got the new h-bridges all hooked up and took Wide Track out for a test drive. Things started out surprisingly smooth. My home brewed RC to PWM seemed to work ok. If you are interested, the source code is here. One small technical difficulty popped up when one of the motors stalled. The Tecel D-100 motor controller caught fire! Now I have seen a lot of "magic smoke" released from a lot of components in my day, but you haven't lived until you have seen a mosfet catch fire! (Note to self: Install fuses BEFORE taking a circuit for a test drive.) Once the excitement died down, it looks like I only smoked one mosfet on each of the controllers. In fact both controllers will still drive the motors in one direction. I tested this AFTER I installed 20A fuses on each motor. A visit to DigiKey yielded replacement MOSFETS (IRF3205) for only $1.51 each, so I went ahead and ordered 10. When they come in, I'll have someone in the group who is good at soldering replace the burnt up parts.
(May 2003) I just received my high-current H-bridges from Tecel. They only operate at 1KHz but for $28 each I can't complain. I am going to have to use my Atmel STK200 board to read the rc radio servo pulses and convert that into a KHz PWM stream. Since the Atmel is only running at 4MHz, I am having to get REAL creative on the coding to make it fast enough. I'll post the source code when I get it working.
(March 2003) AaronLundquist welded the sprockets to the drive shaft for me. They no longer slip. WideTrack ran very smoothly at SouthBySouthWest2003 and at CareerDay2003. This was the first time that WideTrack ran the batteries down before experiencing a fatal mechanical failure. That's something of a milestone.
(Feb 2003) The new groove works very well. The tracks stay aligned and don't skid off any more in a turn. Now the main problem is the connections between links. Originally, I simply used a 3/8 #4 screw to hold the connections together. This allowed me the flexibility of easy disassembly and reassembly. Now that everything is working well, the large stresses on the links tend to pull out. I am now going over each track and removing each screw and using pvc glue to permanently attach each link.
(Dec 2002) At this point, I am rebuilding the tracks to add a groove. A third wheel will be added to each axle that will fit in the groove. This should eliminate the track slippage when turning.
I would really like to get rid of my simple mechanical H-bridges and replace them with a pair of Vantec RET 713 Electronic Speed Controllers or possibly an OSMC. As soon as I can shake loose $300USD from the money tree I'll be sending in my order. An alternative may be to custom build a pair of controllers. Next will be the addition of sensors. Ultrasonic and IR rangefinders probably controlled by my Atmel STK200 board connected via RS-232 to a laptop connected to an 802.11b network.
Making Contact
You can e-mail Eric Lundquist at robo@robotgroup.org
Free counters provided by Andale
