Author Topic: Heavy-weight tank  (Read 7454 times)

0 Members and 1 Guest are viewing this topic.

Offline ilovetabascoTopic starter

  • Jr. Member
  • **
  • Posts: 16
  • Helpful? 0
Heavy-weight tank
« on: September 11, 2006, 02:31:15 PM »
First I would like to thank the moderator for his superb contributions to the hobbyist robot community!

I am a CS major at the University of Texas at Dallas, currently researching with intent to design and build a “heavy-weight” (2 ft wide, 5 ft long, ~200 lbs) robot tank. It will be built on a custom aluminum chassis sporting a paintball marker (I have the luxury of a full CNC machine shop at my disposal). The end goal of the project is to provide a platform for research in weapon stabilization systems and autonomous vehicles (so eventually I’ll be needing gyros and inertial guidance sensors) but for now I’m concentrating on the platform itself.

The programming is practically trivial for me, but as this is my first foray into electrical/mechanical engineering and robotics I am a bit apprehensive about my design. I am posting it in the hopes that someone could offer me advice and/or identify problems I undoubtedly have overlooked.

1. The HID controller will consist of 2 COTS Windows based computers, the first for the driver and the second for the gunner (though this may be run from 1 computer with dual head display).
2. The onboard controller will consist of a COTS XP Embedded computer linked to the remote controllers through WiFi. It will be responsible for agglomerating and scrubbing all sensor data, and eventually executing autonomous programs.
3. DIO will be from a PCI controller card (http://www.accesio.com/go.cgi?p=../pci/pci_dio_24d.html). I plan on using a few of the outputs to drive H-bridges with PWM (I’m a bit leery that handling PWM from the central processor will drain too many computer cycles; can anyone confirm or deny this?). Also, these boards source 32 mA at 5VDC: should I op-amp it to drive the MOSFETS?
4. The H bridges will control a set of motors for locomotion, turret traversal, and the azimuth of the gun (the last may be controlled w/ a stepper). The primary drive motors are Magmotor S28-150s (http://www.battlekits.com/robot_motors.htm).
5. The power source for these beasts will be 2 Hawker Odyssey PC680s, wired in series for 24 VDC at 32 AH (I believe these should provide me with at least 30 min of cruise time, no?)
6. The tread is a point of contention currently. I have the ability to custom manufacture a tread from aluminum, but I don’t want to unnecessarily increase the complexity of this project. I may also use a cog belt or attachment chain. Any advice here would also be appreciated. I expect this thing to move at 10-15 MPH and the treads will be about 4.5 inches wide, 114 inches long.

If anyone sees any problems, tips, advice, flames (I’m sure I can get one for not using Linux or a RTOS) please let me know.
-Chris
« Last Edit: September 12, 2006, 03:37:52 PM by ilovetabasco »

Offline dunk

  • Expert Roboticist
  • Supreme Robot
  • *****
  • Posts: 1,086
  • Helpful? 21
    • dunk's robot
Re: Heavy-weight tank
« Reply #1 on: September 13, 2006, 07:34:51 AM »
wow. quite a project.
i think the best bit of advice i could give would be to build a prototype before you start spending money on the final components.
i guarantee that some aspect of your design will change as you start building.
building a prototype will save you from having to build more than one version of a mechanical component or circuit.
i have used a variety of materials for prototyping mechanical components. Mechano constructor sets are great for prototyping: http://www.constructiontoys.com/store/erc-motion.php

Quote
2. The on-board controller will consist of a COTS XP Embedded computer linked to the remote controllers through WiFi. It will be responsible for agglomerating and scrubbing all sensor data, and eventually executing autonomous programs.
have you thought about power supply for your on-board computer? i have used these with good results: http://www.mini-itx.com/store/?c=10#p1835

Quote
3. DIO will be from a PCI controller card (http://www.accesio.com/go.cgi?p=../pci/pci_dio_24d.html). I plan on using a few of the outputs to drive H-bridges with PWM (I’m a bit leery that handling PWM from the central processor will drain too many computer cycles; can anyone confirm or deny this?).
i would hope that these cards will be able to handle PWM in hardware. the specs talk about "Counter/Timers (Optional)" but i don't see PWM functionality mentioned anywhere which would be disappointing.
i suspect you will run into problems if you try to get accurate PWM controll from a windows program. when the system gets busy you will have longer pulses than when it's resources are free. i'm not saying it won't be possible (or that i know much about windows programming, (that's your field right?)) but it would be far easier if you used a I/O device that could handle the timing for you.
i am working on a very similar project. i am using a microcontroller for a motor controller, attached to the PC via a USB port. (a serial port would be *far* easier but i'm using USB for fun.) all the PWM stuff is handled by the microcontroller. the PC just has to send device_number and pulse_width info.

Quote
5. The power source for these beasts will be 2 Hawker Odyssey PC680s, wired in series for 24 VDC at 32 AH (I believe these should provide me with at least 30 min of cruise time, no?)
the bigest power draw on your bot will be the motors. from the info that comes on your motors you can expect them to draw around 37Amps @ 24 Volts. so with a pair of these, running most of the time, yes, i would concur that your batteries will probably last around 30 mins.
the 2nd highest power draw will be from your on-board computer.
the higher performance a computer you use, the higher the power draw. once you get things up and running, consider removing all unused components.
memory uses a lot of power. downclocked CPUs will use less power than ones running at full speed. laptop hard disks will not only put up with bumps better but use less power too. (linux, not needing a GUI to run, will be less CPU intensive, therefor use less power than windows....)
i think you should be able to get a PC with limited components, running a limited OS drawing less than 500mA @ 24Volts. so it would be possible to get quite a long standby life out of this thing.
(i'm nearly ready to test power draw on a mini ITX mobo at home. i'll let you know how i get on.)

Quote
6. The tread is a point of contention currently.
although i don't have any experience, from reading on various forums, tracks are notoriously difficult to make reliable.

Quote
flames (I’m sure I can get one for not using Linux or a RTOS)
dam right. what's that windows stuff doing on there......

good luck with your project and keep us posted!
Quote

Offline Admin

  • Administrator
  • Supreme Robot
  • *****
  • Posts: 11,703
  • Helpful? 173
    • Society of Robots
Re: Heavy-weight tank
« Reply #2 on: September 13, 2006, 09:20:52 AM »
Firstly . . . whoa make me one too!!!

If this is your first bot, I strongly suggest to always take the easy way in everything you do. Give up features for simplicity.

Ok now for comments:
I agree with dunk, use a microcontroller to handle the low level stuff and let your PC interface with it. The microcontroller would have the motor drivers and I/O that you need.

Cost . . . how much money do you have for it? Guessing by the components you are using, it will come out to over $1000 . . .

Consider using timing belts or assembly belt material for treads instead. Basically you get grooved wheels, stretch the belt onto them, and your done. But I think you would be better by using a 4 wheel robot, where the wheels on either side drive together. Basically a 4 wheel differential drive system - I have made one before that worked really nice (see image). Remember that treads are not only harder to make, but less energy efficient, make turning a lot harder, and reduce your forward speed. A wheel can climb over anything lower than it's radius, perhaps using large wheels might be easier?

And lastly my rants . . .
My first robot was big, heavy, and a failure because of being big and heavy. Do everything you can to keep the weight down as much as possible. More weight means bigger motors which mean bigger batteries and motor drivers which mean more costly . . . and the cycle continues . . .

And design your bot 100% in CAD (perferably 3D CAD)! A badly planned robot is a badly built robot.
« Last Edit: September 13, 2006, 09:31:02 AM by Admin »

Offline ilovetabascoTopic starter

  • Jr. Member
  • **
  • Posts: 16
  • Helpful? 0
Re: Heavy-weight tank
« Reply #3 on: September 13, 2006, 12:01:02 PM »
Quote
I agree with dunk, use a microcontroller to handle the low level stuff and let your PC interface with it. The microcontroller would have the motor drivers and I/O that you need.
OK. I was thinking about that so this pretty much confirms it.

Quote
Cost . . . how much money do you have for it? Guessing by the components you are using, it will come out to over $1000 . . .
I've budgeted $2500 for "phase 1," the goal of which is to have a bot that drives, a turret that turns, a gun that shoots, all controlled from the remote computer.

Quote
Consider using timing belts or assembly belt material for treads instead. Basically you get grooved wheels, stretch the belt onto them, and your done. But I think you would be better by using a 4 wheel robot, where the wheels on either side drive together. Basically a 4 wheel differential drive system - I have made one before that worked really nice (see image). Remember that treads are not only harder to make, but less energy efficient, make turning a lot harder, and reduce your forward speed. A wheel can climb over anything lower than it's radius, perhaps using large wheels might be easier?
I've considered that as well (including using wheels in phase 1) however the primary environment this will be running in is loose sand, mud, and grass. So minimizing my ground pressure and maximizing my "bite" into the ground is advantageous. The terrain will also be very rough; imagine driving it around a cunstruction site while they're still doing the dirt work.

Quote
And design your bot 100% in CAD (perferably 3D CAD)! A badly planned robot is a badly built robot.
I'm currently designing it in AutoCAD (3D model). I'm hoping this can avoid the need for a prototype, and a friend of mine can do FEAT analysis on the structure. I'll be creating a website for this soon where I'll post up the drawings.

Offline Admin

  • Administrator
  • Supreme Robot
  • *****
  • Posts: 11,703
  • Helpful? 173
    • Society of Robots
Re: Heavy-weight tank
« Reply #4 on: September 13, 2006, 12:11:53 PM »
Quote
I've budgeted $2500 for "phase 1"

Please tell us your secret in getting robot funding! Phase 1 sounds like a huge project in itself . . . what are the other phases? Usually when I build a robot, Phase 1 is designing it, Phase 2 is building the mechanical parts, and Phase 3 is electronics/programming/testing . . .

Quote
minimizing my ground pressure and maximizing my "bite" into the ground is advantageous.

What about really wide tires? And/or tires with deep groves in them (sorta like a gear)?

Quote
I'm currently designing it in AutoCAD (3D model). I'm hoping this can avoid the need for a prototype

No matter how much you design, there will always be mistakes. Its more about minimizing mistakes and the loss of money/time to correct them . . .

Offline ilovetabascoTopic starter

  • Jr. Member
  • **
  • Posts: 16
  • Helpful? 0
Re: Heavy-weight tank
« Reply #5 on: September 13, 2006, 12:29:56 PM »
Quote
i think the best bit of advice i could give would be to build a prototype before you start spending money on the final components.
I'm hoping that modeling it with AutoCAD / Mechanical Desktop will eliminate this necessity for a full prototype (though some things may require real mock up)


Quote
have you thought about power supply for your on-board computer? i have used these with good results: http://www.mini-itx.com/store/?c=10#p1835.
Thanks! Those should work well. I'm also going to use separate batteries for the computer. I hadn't given it much thought because I was planning on using whatever my universities IEEE robotics team came up with for their bot and AUV (both use ITX PCs, one ran Windows XP, the AUV Gentoo linux) http://auv.utdallas.edu/

Quote
i would hope that these cards will be able to handle PWM in hardware. the specs talk about "Counter/Timers (Optional)" but i don't see PWM functionality mentioned anywhere which would be disappointing.
i suspect you will run into problems if you try to get accurate PWM controll from a windows program. when the system gets busy you will have longer pulses than when it's resources are free. i'm not saying it won't be possible (or that i know much about windows programming, (that's your field right?)) but it would be far easier if you used a I/O device that could handle the timing for you.
i am working on a very similar project. i am using a microcontroller for a motor controller, attached to the PC via a USB port. (a serial port would be *far* easier but i'm using USB for fun.) all the PWM stuff is handled by the microcontroller. the PC just has to send device_number and pulse_width info.
It mentions PWM *MEASUREMENT* in the manual, nothing about output. I think I will call them up and find out. But a microcontroller is starting to sound like the way to go, I just want to keep complexity minimized.

Quote
the 2nd highest power draw will be from your on-board computer.
the higher performance a computer you use, the higher the power draw. once you get things up and running, consider removing all unused components. memory uses a lot of power. downclocked CPUs will use less power than ones running at full speed. laptop hard disks will not only put up with bumps better but use less power too. (linux, not needing a GUI to run, will be less CPU intensive, therefor use less power than windows....)
i think you should be able to get a PC with limited components, running a limited OS drawing less than 500mA @ 24Volts. so it would be possible to get quite a long standby life out of this thing.
(i'm nearly ready to test power draw on a mini ITX mobo at home. i'll let you know how i get on.)
I've got a few ideas here. First is using an ETX mobo like http://www.accesio.com/go.cgi?p=../custom/etxnano.html (but this won't be in phase 1). Second, I plan to use a pentium M processor. Third, I will eventually go to Windows XP embedded which will allow me to strip down the OS to just the components I need.

Quote
6. The tread is a point of contention currently.
although i don't have any experience, from reading on various forums, tracks are notoriously difficult to make reliable.

Quote
dam right. what's that windows stuff doing on there......
First, I'm primarily a windows programmer, but I've done enough linux to make it an option. Second, I have a close relationship w/ Microsoft (I write large enterprise aplications for the health care industry, primarily based on Microsoft technology) and with their upcomig Robotics Studio I might be able to swing some deals w/ them.

Offline ilovetabascoTopic starter

  • Jr. Member
  • **
  • Posts: 16
  • Helpful? 0
Re: Heavy-weight tank
« Reply #6 on: September 13, 2006, 12:48:12 PM »
Quote
Please tell us your secret in getting robot funding!

I'm funding the first phase by myself (the secret is that the health care industry pays REALLY well). I have received pledges for subsequent phases contingent on the first phase getting off the ground (and these are LARGE pledges).

Quote
Phase 1 sounds like a huge project in itself . . . what are the other phases? Usually when I build a robot, Phase 1 is designing it, Phase 2 is building the mechanical parts, and Phase 3 is electronics/programming/testing . . .
Phase 1 (or iteration 1), as described above, basically entails designing, building, coding and debugging the working bot which will serve as a platform for reasearch.
Phase 2 will consist of upgrades to the weapon system. Simply put its supposed to mimic how the M1A2 Abrams stabalizes its weapon, so they can fire while running at full speed and hit their target. So I'll eventually need inputs for gyros (I'll need to know my roll, pitch, yaw), encoders (I'll need to know my velocity, turret position, barrel azimuth), a gimbled camera, and a laser rangefinder.
Phase 3 will consist of making the full system autonomous (with the exception of a user held KILL switch). I've done a bit of work with neural networks, A* pathfinding, and computer vision, so this isn't impossible (just really really hard).

Quote
What about really wide tires? And/or tires with deep groves in them (sorta like a gear)?
Hmmm... the only thing is that I want to keep my profile as small as possible. Eventually I'll become the target.

Quote
No matter how much you design, there will always be mistakes. Its more about minimizing mistakes and the loss of money/time to correct them . . .
Agreed.
« Last Edit: September 13, 2006, 04:03:17 PM by ilovetabasco »

Offline Admin

  • Administrator
  • Supreme Robot
  • *****
  • Posts: 11,703
  • Helpful? 173
    • Society of Robots
Re: Heavy-weight tank
« Reply #7 on: September 13, 2006, 01:00:06 PM »
I guess it is still early in your design to ask, but what about target identification? I mean, how would your robot tell the difference between a 'bad guy' and a 'good guy'?

My assumption is that your system will be semi-autonomous, and work under a similar structure as the Predator (that flying drone blowing up the crazies in Iraq) - the robot handles everything except target ID and the firing button. Correct?

Where are you located? I might be interested in joining/supporting your project at some point in the future.

Offline ilovetabascoTopic starter

  • Jr. Member
  • **
  • Posts: 16
  • Helpful? 0
Re: Heavy-weight tank
« Reply #8 on: September 13, 2006, 01:17:06 PM »
I've had a few ideas for target identification/selection. First is computer vision (pattern matching). Second, contingent upon some gracious aid from my sponsors, would be thermal imaging. I guy that works with Raytheon mentioned a form of radar but I don't know much about it. Of course radio telemetry (cause all these guys should be sending out signals) could work as well. But the competitions will be mostly you against the world (1-on-1 combat, and seek and destroy a predefined list of targets) so friend-fo detection won't be a concern.

Quote
My assumption is that your system will be semi-autonomous, and work under a similar structure as the Predator (that flying drone blowing up the crazies in Iraq) - the robot handles everything except target ID and the firing button. Correct?
In phase 3 it will be fully automous, controlling positioning, evading, target aquisition, firing, etc. The only thing the operator will control is a remote kill switch in case it goes haywire and starts shooting at people.

I'm located in north Dallas, TX. As soon as I get the website for this up and going I will post it here.

Offline elsint

  • Jr. Member
  • **
  • Posts: 8
  • Helpful? 0
Re: Heavy-weight tank
« Reply #9 on: September 13, 2006, 02:11:19 PM »
Quote
I'm located in north Dallas, TX.

Tell me when you started testing that I won't get close to Dallas!

I don't want to be killed because of some kind of software bug!  ;D



Offline dunk

  • Expert Roboticist
  • Supreme Robot
  • *****
  • Posts: 1,086
  • Helpful? 21
    • dunk's robot
Re: Heavy-weight tank
« Reply #10 on: September 14, 2006, 06:54:00 AM »
one more point i didn't think of before,
Quote
2 ft wide, 5 ft long,
you know this guy will turn far more easily if you increase the width commaired to the length.
3ft wide by 3.5ft long would give you a similar foot print but put far less stress on your tracks as you turn.

dunk.

Offline ilovetabascoTopic starter

  • Jr. Member
  • **
  • Posts: 16
  • Helpful? 0
Re: Heavy-weight tank
« Reply #11 on: September 14, 2006, 06:03:26 PM »
That's a good point. I'm modeling it in AutoCAD now, and the current x-y bounding box is 29" by 52" for a ratio of .55 (a bit better than the oringinal guestimate). And of course the closer you get to a square the greater your area utilization :). After I get all of the large components modeled I'll play with rearranging them. Thanks!

I do have a question for all you electronics gurus: How would I go about constructing an H-Bridge for my motors? A quick chat with an IEEE guy yielded 3 options: synchronous MOSFET rectification, Schottky diodes, or a large MOSFET. Unfortunately I didn't have enough time to make sense of which was best for my application. If anyone has any knowledge about this it would be greatly appreciated. My requirements are as follows:

nominal 24VDC
~37 amp continuous, theoretical 180 amp spike
16kHz

I found a single MOSFET that should work (Digikey 497-3252-1-ND) though I would probably want to boost the output of my microcontroller to 10V to lower their RDS.


Offline Admin

  • Administrator
  • Supreme Robot
  • *****
  • Posts: 11,703
  • Helpful? 173
    • Society of Robots
Re: Heavy-weight tank
« Reply #12 on: September 15, 2006, 04:23:42 AM »
Since you have the money, it would be better to buy a DC motor driver than to make something on your own. Probably rated at 30V and guessing about 50 amps (use a big capacitor to suppress current spikes). Check out products for battlebots, as there should be a good selection of motor drivers for them. Expect to spend like $100-$200+ on them, and make sure its 2 channel so it can handle 2 DC motors (or buy two of them).

Offline dunk

  • Expert Roboticist
  • Supreme Robot
  • *****
  • Posts: 1,086
  • Helpful? 21
    • dunk's robot
Re: Heavy-weight tank
« Reply #13 on: September 25, 2006, 09:00:57 AM »
so just following up on the PC power draw point from earlier.
so my mini itx motherboard (http://www.mini-itx.com/store/?c=2#p1641) running a fairly minimal Linux OS with no GUI enabled is drawing around 1400mAmps @ 12Volts.

i haven't tried much optimization yet but i imagine using some of the processor stepping options and HDD spindown options i could bring this down to around 1000mA @ 12V.

running windows, CPU and memory intensive as it is, would have the opposite effect but i'd imagine it would only make a few hundred mA difference from an unoptomised Linux system.

let's pick 1500mA @ 12 Volts as an average figure.
this equates to 750mA @ 24 Volts.
so realistically, using ilovetabasco's 24 Volt at 32 AmpHour batteries, an on-board computer like this could run for 32,000 / 750 = 42.666 hours, with no other powered on equipment.

that's some set of batteries you have on board there. close to 2 days worth of standby battery life if you do a good job of power management on all your hardware interfacing.

imagine your bot, nothing powered up except the on-board computer and a webcam. waiting for the opposition to make a move.
http://www.hackaday.com/2005/09/21/robotic-sentry-gun/

dunk.

Offline ilovetabascoTopic starter

  • Jr. Member
  • **
  • Posts: 16
  • Helpful? 0
Re: Heavy-weight tank
« Reply #14 on: September 27, 2006, 12:02:56 AM »
Work is proceeding slowly and steadily. I was modeling the whole thing in AutoCAD, but threw that out in favor of SolidWorks. SolidWorks= teh shiznit. I believe the suspension and base frame is complete. It looks like I will be using cog belts as the tracks to begin with (and perhaps replacing them in a future iteration).

Dunk,
First, I will be using CF instead of a hard drive. I also plan using a pentium M processor, and Windows XP embedded. My build of XPe should be fairly small, featuring no shell. I'm not sure what kind of power requirements this will have, but as soon as I find out I will let you know. Also, I will probably be powering my electronics from a separate set of batteries (so I don't have to worry about spikes/brownouts). It will probably be 24V as well, so if I wanted to get really fancy I could switch the electronics from bus A (main power) to bus B (isolated).

For those not familiar with XPe, its a componentized version of Winxows XP. I call it Microsoft's answere to Gentoo Linkux. Don't need DHCP? Throw it out! Who needs a shell or graphical user environment? You can toss it as well. The absolute minimum configuration has a footprint of 4.8 MB. A full configuration, which is practically a full installation of Windows XP, is 300 MB. Oviously boot time can be significantly decreased.

I get XPe from my school for free, along with non-commercial runtime licenses. Suprisingly enough its not even available in my MSDN subscription, though a trial version is available publicly.
« Last Edit: September 27, 2006, 12:04:27 AM by ilovetabasco »

Offline SP

  • Jr. Member
  • **
  • Posts: 16
  • Helpful? 0
Re: Heavy-weight tank
« Reply #15 on: January 22, 2009, 08:40:36 PM »
you could use vex parts from www.vexrobotics.com to make a protype. they actualy have treads to put on your prototype.

 


Get Your Ad Here

data_list