DPRG Events Shoptalk

Website design and hosting by NCC

Tankbot GTR

28 December 2006, by Ed Paradis and R. Steven Rainwater


When DPRG member Bill Cole moved out of state, he left the group a collection of robots and robot parts he didn't have room to take with him. One of those items was a radio controlled Stuart M5 tank made by 21st Century Toys. Around this same time, several DPRG members were looking for a way to get a group project started, in part to dispell the curse of GARP, and in part to begin building up a collection of interesting robots that could be kept on hand by the DPRG for demonstrations and experimentation. The M5 tank toy looked like an ideal platform for a group robot project, so some funds were allocated and we got to work.

A few thumbnail photos are include in the build log. You can click on them to see the full size image. There are many more images in the Tankbot GTR photo gallery that you may want to browse.

Before we get into the build log, we should give some credit to the many people who directly or indirectly contributed to this project. So many members have worked on this, that I'm sure we've left someone out. If your name isn't on this list and should be, please email me and I'll get it added.

Corporate Sponsors

Individual Contributions

  • David P. Anderson - Fabricated custom drive sprocket hubs, donated HD15 pins/sockets, provided design advice
  • Paul Bouchier - Donated 2GB USB drive, compass, Arduino, and video camera
  • Bill Cole - Donated the M5 Stuart Tank and a supply of Cole-Cones
  • Mike Dodson - Donated hardware, helped with fabrication and machining
  • Dennis Draheim - Donated Earthmate GPS unit
  • David Ellis - Donated 4 MaxBotix sonars, several IR sensors, 12v DC-DC PS for Mini-ITX
  • Michael Giambalvo - Donated 2 12V NiMH battery packs, lots of help with construction, worked on AVR firmware
  • Ron Grant - Fabricated motor mounting hardware based on his M5 tank
  • Mike Hernandez - Assistance with test runs
  • Martin Meier - Donated 1GB CF Card, built wiring harness, helped with assembly
  • Jason Oliver - Donated USB wifi dongle
  • Ed Paradis - built sonar array, worked on AVR firmware, primed and painted chassis, integrated motors and h-bridges with the AVR board
  • Steve Rainwater - Donated Pittman motors, 1/4" shaft hubs, RAM for the Mini-ITX, primer, paint, vinyl dye
  • Dale Wheat - Donated 3 AVR boards, LED assembly, video monitor
  • Charlie Youngblood - Assistance with test runs

Build Log

For the most part, we work on Tankbot at our Tuesday night RBNO meetings at the DPRG lab in Garland. Occasionally we get extra motivated and show up on a Thursday night or a Saturday too. The following build log will give you and idea of how things are proceeding.

November 2005

The tankbot project started with a toy remote controlled tank made be 21st Century Toys. It was fully functional with independent track controls and sound FX. Unfortunately, we found the 'guts' too be unsuitable: it used only on-off relays (so there was no speed control) and the motor and gear boxes were incredibly loud.

So we decided to strip the tank down to its mechanical components. We also decided that we should paint the tank, and do it right. So we laid down a coat of white primer.

The wheels and track parts of the Tank were fairly robust and we disassembled them as much as needed for painting.

The tank originally had a large lip that could get caught doing certain extreme manuvers, such as driving up steps or down curbs. A hacksaw made quick work of that.

We used black vinyl dye to stain all the wheel components besides the tracks. Vinyl dye worked great because it really doesn't build a layer when used on plastic. It sort of 'soaks into' the plastic. This meant that the wheel assemblies could be painted without worrying if they would bind afterwards.

We went ahead and put a second coat of primer on the chassis. If we were going to paint this, we were going to make it pretty!

December 2005

We decided upon using Pittman 9xxx series motors, as other DPRG members had used them on similarly sized robots. First, we studied how Ron Grant mounted his motors to the same model toy tank. He let us know that the hubs on the drive wheels would need reinforcing.

We also examined how David Anderson mounted the same type of motors on his J-Bot. He has a lot of custom machined components, but had no problem with his method.

Steve Rainwater found four Pittman 9413 motors on ebay. From what he could get out of Pittman, the motors had a reasonable torque and rpm combination.

David Anderson machined some hubs for the tank which provided a sturdy way of fastening the motor to the drivewheels. They consisted of a round plate which bolted to the wheel, and collars which bolted to the plate. The collars have a set screw to secure it to the output shaft of the gearheads on the motors.

January 2006

The next step was to figure a way to mount the motors to the chassis. Ron Grant had used a sort of "Floating Tee" to mount his motors. A single peice of all-thread went along the length of the tank, in the center. Then a peice of angle aluminum was secured to the all thread, forming a "T". The angle aluminum could rock back and forth, so there was some mechanical give. He then fastened the motors to the angle aluminum. It forms a sort of floating axle.

We decided to copy his idea, and Ron generously made some duplicate mounts for us out of spare material from his tank.

Michael Giambalvo donated a pair of 14 volt NiMH battery packs. We planned to connect these in serial to have a 28 volt system. Also, we determined that the maximum speed of the robot would be about two miles an hour, which follows David Anderson's rule: "Never make a robot which you can't outrun."

Ed Paradis and Michael Giambalvo did the handywork to modify the chassis to accept the motor mounts. A test fit of the motors showed that we need to do some trimming to make the drive wheels the correct distance away from the chassis. If this distance is incorrect, the tracks will bind, causing a lot of noise, wear, and wasted power.

At the January meeting, we demonstrated the robot moving under its own power. No sensors were connected and a rather inefficient H-bridge was used. An Atmel microconroller controlled the motors to slowly move forwards, then slowly move backwards. Not exactly exciting, but it showed that the tank was mechanically feasable.

We decided on a color: a metalic candy apple red. Steve picked up a few cans at the auto-parts store and we proceded to paint the chassis. All the parts were removed from the chassis and we determined that a few had been incorrectly installed previously.

Over the course of a week or two, we proceded to paint more and more layers of candy apple red metallic paint on the chassis. We then decided what it really needed was a gloss coat, so several layers of gloss were slowly built up on the chassis. The end product was a very very shiny tank. Many brain cells died in order to obtain this shiny tank.

With the dimensions of the tank increased somewhat due to the many layers of paint, we found it necessary to sand down some of the parts where bogeys and wheels mounted. A generous application of powered graphite also helped. We determined that with the graphite the wheels were much more free spinning.

The tank treads and wheels were reattached, and the motors remounted. We took this opertunity to rethink some of the distances on the motor mounting, and made sure everything lined up. We did some testing to isolate points of friction, adding more graphite or adjusting the drivetrain.

February 2006

We displayed the newly painted tank at the Feburary meeting. We ran the tank on a single 14 volt battery but did not alter the program. The tank still slowly rolled forwards and backwards.

Randy Dumse of New Micros Inc. donated a pair of excellent H-bridges to use with the tank. These bridges feature debug/status LEDs which make them a pleasure to develop with. Also, VIA donated a motherboard for the project which would eventually make its way on to the robot.

We decided to test the robot under full power using both 14 volt packs. The results were very encouraging. Michael Giambalvo developed some firmware for the AVR to allow for a tethered control mode, meaning we could finally command the robot to turn and go different speeds. Since the ultimate goal of the tank is to run as an entry in a RoboMagellan-style outdoor competition, we decided it was finally time to try it outdoors. The tank performed admirably on the gravel, grass, and sand behind the DPRG World Headquarters. Turning in place proved difficult in some circumstances, but by adding a small amount of forward speed, we could turn very tight arcs.

March 2006

A very wide range input voltage switch 5 volt regulator was donated by Ed Kaufman of RC Model Works and was integrated with the New Micro H-Bridges. A small shelf was built to keep these parts and mounted near the motors.

We devised a standard 15 pin connector to sit between the H-bridges and the Atmel. This should allow other controllers to be used with the tankbot, without extensive re-wiring.

May 2006 (Pictures 0070-0076)

David Ellis donated four MaxBotix sonars which were built into a small sonar array on the front of the tank. One pointed forward, one to each side, and the fourth was saved to point straight backwards, to prevent getting stuck in tight situations. Each sonar returned a range from 6 to 255 inches, and were polled individually by the AVR.

The ability to turn in place is important for many schemes of navigation, so we performed tests to determine the tanks ability to turn in place.

The eventual design of the tankbot was to have a computer communicating with the AVR, gathering sensor information and sending motor commands. This interface was developed using laptops instead of the actual computer because we did not yet have a proper power supply for the computer.

The tank went on to win its first prize: First Place at the Spring DPRG Show-and-Tell. It demonstrated collision avoidance using a small laptop. The laptop ran Linux and the command program was written in Perl. The communication method between the AVR and the computer was designed to be easily implimented in any language which can manipulate text, so Perl, C, Java, or even BASIC can be used.

June 2006

David Ellis donated a 12 volt ATX power supply to match the VIA motherboard, and a spare 800MB laptop harddrive was appropriated from the DPRG pile of parts. Red Hat Linux v9 was installed on the laptop drive.

August 2006

Steve Rainwater visited E & D Plastics, a local plastic fabrication company, and collected many lexan and plexiglass pieces from their discard bins. One of these pieces was selected for use as an upper deck and mounting plate for the electronics on the robot. Mike Dodson and Will Kunhle trimmed and machined the plastic to the right size.

September 2006

Martin Meier drilled mounting holes in the new plexiglass upper deck plate and and mounted the Mini-ITX and microcontrollers. Having the electronics on top of the tank allows easy access to the computer, with the batteries fitting below.

Martin also simplified some of the wiring mess that had accumulated over the months into a clean terminal strip. A plastic drive case from a discard IBM Thinkpad was also modified to protect the hard drive and prevent its metal surfaces from touching the underside of the Mini-ITX.

A monitor, keyboard, mouse, and ethernet can be attached to the Mini-ITX, allowing it to be used like a normal desktop computer. When this is not convenient, a laptop can be connected to the Mini-ITX via the Ethernet port and ssh used to access. A full tool chain and development environment were available on the robot at this stage, minimizing the work needed to build and test the code.

Steve started working on using the encoders for proper PID control of the motors. A simple test program was developed on the Mini-ITX, written in C, that communicated with the AVR using the serial port.

The robot participated in the DPRG's Outdoor Robot Contest, held at the Science Place in late September. The robot correctly oriented itself to attain the first waypoint and stopped halfway to the goal due to an encoder problem that remained undiagnosed for some time. We also experience some trouble with the robot turning in place on uneven surfaces. This had happened before. The concensus is that slightly more torque is needed than the Pittman 9413 motors are providing.

October 2006

Steve obtained a female IDE to CF card adapter on eBay. This was the first step in replacing the mechanical hard drive with a solid state flash device. The female adapter plugs directly into the IDE connector on the Mini-ITX, allowing a CF card to be added without the need for additional cables and mounting hardware.

November 2006

Martin Meier donated a 1GB CF card. The Red Hat Linux 9 setup on the hard drive was transferred directly to the CF card using dd. It booted from the CF card successfully on the first try. Afterward, a lot of time was spent reconfiguring to Linux to make it reasonably non-disk intensive to avoid unnecessary writes.

The robot participated in an outdoor robot exercise organized by David Anderson at the December DPRG meeting. The encoder problem experienced during the September Outdoor contest resurfaced. Experimentation revealed the problem was in the right encoder only. Later testing back at the DPRG lab pointed to a loose connection in one pin of the 15 pin connector that carries the encoder and h-bridge signals. One of the connectors is replaced, hopefully fixing the problem.

December 2006

Ed began a rewrite of the AVR code to incorporate the velocity and distance calculations that were being done upstream in the Mini-ITX. Steve got a new set of Pittman motors on eBay, which were swapped out with the original Pittmans. After some testing it was determined the new motors were not performing as well as the old ones, most likely due to custom windings that resulted in less torque than predicted by the Pittman specs for that model. If motors with more torque don't turn up on eBay soon, plan B is to swap out the 19.1:1 gearboxes on the Pittman 9413 motors with 38.3:1 or 65.5:1 gearboxes.

January 2007

Steve obtained a new set of 60:1 gearheads for the Pittman 9413 motors and swapped them out with the old 19.1:1 gearheads. The result was much improved torque while still maintaining a reasonable amount of speed. Tankbot can now rotate in place on most low friction surfaces (e.g. cement, asphalt, level grass, etc.) but tends to slide or stall on irregular ground.

February 2007

Ed delivered completely rewritten AVR code that includes an improved serial protocol, interrupt support, and new PWM code. He also spent some time debugging the sonar code. David Anderson built some custom wiring harnesses for the batteries that include banana plugs to simplify the process of charging from the Triton charger.

May 2007

Tankbot was displayed and demonstrated at the International Space Development Conference (ISDC) in Dallas along with other DPRG robots.

December 2007

Scott Sumner added a Linksys WiFi router to Tankbot.

November 2008

Paul Bouchier contributed a new 8Ah lead-acid electronics-battery, to replace the 4.5Ah battery previously in Tankbot, and constructed a bracket to mount the batteries in Tankbot's hold. While this will help battery life, the 3.2 Amps it draws (2.85A from the Via motherboard) will still limit runtime without the charger. Paul also built charging cables and plugs on Tankbot which allow the 13.6V power supply and the Triton charger to be plugged to Tankbot for battery charging & power during FW development.

Will Kuhnle built a mount & brackets for the servo and steerable sonar (for which Paul Bouchier contributed the servo).

Tankbot competed in Roborama 2008b the weekend before Thanksgiving, and placed 4th in out'n-back on a double-scale course.

December 2008

Paul Bouchier contributed an Arduino board to run the sonars and compass, monitor battery voltage, and drive the steerable sonar's servo, and Will Kuhnle mounted it. The sonars were disconnected from the AVR board and wired to the Arduino. Paul build an Arduino shield for sonar connectors etc, and mounted one of Dale's tiny cylons on top of the steerable sonar.

Paul contributed an 802.11b PCI wifi card, which replaced the Netgear router previously strapped to the 'bot. The router can now be used as a battery-operated router for field use, while the wifi card gives tankbot access to the internet.

Paul replaced the ancient version of Linux (2.4 kernel, Red Hat based) with a recent version of Puppy Linux, at Steve Rainwater's suggestion. This provided a 2.6 kernel and USB-serial port & wifi card support, and is optimized for running out of the compact flash card. Along with VNC, it allows for GUI-control over tankbot. However, typically for Linux, getting connected to the club wifi network is proving problematic.

Paul contributed a USB webcam, to be used in tankbot's vision system.

Paul built a tether with a mast, which allows tankbot to wander around a little while remaining AC-powered.

January - February 2009

Paul tried to get wifi card working. He ordered a different wifi card, which didn't work in Tankbot's PCI slot either (yet it worked in Paul's home PC). Ed found that the originally contributed wifi card doesn't have linux support. Problem still not solved - possible problem with PCI slot?

Paul contributed a 2GB USB thumb drive, & club donated a USB hub to accommodate the increasing number of packages

Filled out Puppy Linux package suite: added sshd, samba server exports the robot-FW part of the filesystem as a share, devx package adds gcc + development environment. Rebuilt & tested the apps that ran on the previous RedHat linux.

Paul contributed an electronic compass after reading how to use them properly, and how dead-reckoning turns can't work on a tracked robot

Dale Wheat contributed a video monitor, which Paul mounted and connected to Tankbot's VIA Mini-ITX board.

March 2009

Tankbot appeared at All-Con, attracting attention by rolling back and forth on the DPRG display table, as well as appearing in the DPRG petting zoo event. Despite not behaving well at times, Tankbot was well received by attendees

Jason Oliver contributed a USB wifi dongle which worked great on Puppy Linux. This was fortuitous, because Tankbot's ethernet port started to exhibit intermittent failure symptoms.

Paul wired up the compass connector to the arduino on the home-made shield and installed some Arduino code to continuously poll it & print the heading.

April 2009

Paul constructed a mast to mount the compass on and keep it away from metals and current.

May 2009

Paul entered Tankbot in the RoboRama 2009a Square Dance contest, an event roughly based on the idea behind the Borenstein Square or UBMmark. Despite nicking a couple of the corner markers, Tankbot did remarkably well and returned almost precisely to its original starting position. The improvement in navigation is due almost entirely to Paul's work with the new compass.

July 2009

Having completed his goals with Tankbot, Paul decided to move on to other projects and pass Tankbot on to another member or team. He posted a "message from Tankbot" to the DPRG mailing list on July 8.


Vendors contributing to this project

Article Copyright © 2006-2007 by Dallas Personal Robotics Group, all rights reserved.