PDA

View Full Version : Autonomy And Mech-Warfare



JonHylands
07-14-2008, 08:30 AM
The question of autonomy has come up a number of times in this forum in various threads, so I thought it might be interesting to start a new thread on the topic.

Some of the things I'd like to discuss in here:

- targeting
- onboard processing only versus having a remote PC
- detecting things like buildings
- strategy

And anything else anyone else comes up with.

A-Bot
07-14-2008, 08:46 AM
The question of autonomy has come up a number of times in this forum in various threads, so I thought it might be interesting to start a new thread on the topic.

Some of the things I'd like to discuss in here:

- targeting

I assume autotargeting is best accomplished with Roborealm + custom code, but I still need to work it out.

- onboard processing only versus having a remote PC

I'm for onboard. It's possible to get a pico-itx in a mech for those who want that power. If we use an IR beacon then video processing is not required which opens the doors to affordable microcontroller options.

- detecting things like buildings

I was planning to use sonic rangers for this.

- strategy

"Seek and destroy the opposing mech!" :mad:

Strategy is where the real autonomy comes in. This is where us programmer geeks can really start to have fun. :veryhappy:

lamont
07-14-2008, 03:31 PM
My plan was to have an directed mech with AI like onboard processing for finer control. My fear is that the onboard video will not be high quality enough or low latency enough for fine-grained aiming, so I plan on having the ability to paint a target via the on screen display from my laptop, and have the mech take care of actually doing the firing.

Similarly, I don't plan on servo-level control from my operations laptop, but to be able to direct the mech to "walk forward" or "dodge sideways" and have the on-board processing take care of those goals.

So I plan on using a mix of onboard and offboard processing to direct my mech, with the idea of pushing what decisions I can closer to the mech.

I'll post details when I have them. So far it's just pseudo-code and notebook drawings.

Adrenalynn
07-14-2008, 03:40 PM
"Fly by Wire", Lamont? "Pilot Augmentation"? :) Great goals, IMHO.

lamont
07-16-2008, 05:05 PM
Great, lofty and perhaps unobtainable goals. But I'm making some small progress. I've got a working ruby library with the low level commands, and progress is accelerating now that I've got the building blocks.

Plus I finally got my airsoft cannons yesterday, so I can start experimenting with them this weekend.

sam
07-17-2008, 02:35 PM
My plan was to have an directed mech with AI like onboard processing for finer control. My fear is that the onboard video will not be high quality enough or low latency enough for fine-grained aiming, so I plan on having the ability to paint a target via the on screen display from my laptop, and have the mech take care of actually doing the firing.

Similarly, I don't plan on servo-level control from my operations laptop, but to be able to direct the mech to "walk forward" or "dodge sideways" and have the on-board processing take care of those goals.

So I plan on using a mix of onboard and offboard processing to direct my mech, with the idea of pushing what decisions I can closer to the mech.

I'll post details when I have them. So far it's just pseudo-code and notebook drawings.

Well... That isn't AI or automous bots, is it?

You can have lots of preprogrammed stuff in you miscrocontroller on your bot and then just use a controller to tell the bot wich one to run.

Isn't what John is talking about is a fully automous bot that does all the aiming/walking/aboiding obstacles stuff? Or maybe half automomous (self aiming with firing, you take care of the walking/avoiding)

DresnerRobotics
07-17-2008, 02:41 PM
Have to start somewhere. Honestly if anyone shows up to Year 1 with a fully autonomous bot capable of competing with the human operators, they need to be selling bots to the military :P

sam
07-17-2008, 02:48 PM
We'll see about that! ;)

I would like to try tought. Have a small BRAT syled robot and maybe a pico ITX with the SEQ software and roborealm and some range sensors, laser and camera... Perfect :D

DresnerRobotics
07-17-2008, 02:58 PM
You'll be very hard pressed to fit a pico-itx on anything close to the size of a brat. Would be incredibly top heavy.

Adrenalynn
07-17-2008, 03:07 PM
I'm going to take a shot at it myself. Different plan, though.

sam
07-17-2008, 03:12 PM
You'll be very hard pressed to fit a pico-itx on anything close to the size of a brat. Would be incredibly top heavy.


Yeah, I just realized that. I will maybe go with a quad instead to get enough place for batteries, the motherboard with guns and camera system. Will need good servos tough.

Adrenalynn
07-17-2008, 03:13 PM
Or wirelessly get your processing off the 'bot and over to a laptop somewhere else.

sam
07-17-2008, 03:17 PM
But wouldn't sending all the inforamtion (video feed, all the sensors) make it sometimes kinda lag and it wouldn't work well. Since the laptop would have to send signal to fire, but if it lags, well it might just miss evrything.

JonHylands
07-17-2008, 03:25 PM
If you've got a decent wifi connection, you can get the round trip lag down in the 30-50 ms range. Faster than your reaction time, in any case...

A-Bot
07-17-2008, 03:32 PM
I'm going to take a shot at it myself. Different plan, though.

It's not clear what the "it" is that you're going to take a shot at... :robotindifferent:

Adrenalynn
07-17-2008, 03:37 PM
Knowing autonomy - whatever "it" is probably won't get hit, but all the other "its" better watch out!

I don't expect the wireless to be any slower than the wireless for remote control.

MSpurk
11-28-2008, 07:18 PM
Is there some way we can get a standard colored target for autonomous targeting? What I'm thinking about are those little fluorescent colored circle stickers you use for like yard sales and stuff. If I could put a fluorescent yellow sticker on the approximate center of the target would make auto targeting much easier this first year. Having the shape and size known in advance would allow for ranging as well.

-Matt-
Cutting-Edge Robotics

Sienna
11-29-2008, 12:52 PM
If we are putting florescent anything on the targets, I would like to know sooner rather then later. Because as soon as we said "Oh hey, this piezo works well on a lexan shell", I changed my design to make the front canopy (which contains the cameras) to be target.

elios
11-29-2008, 01:17 PM
lol yea, that sounds like an oh no moment!

MSpurk
11-29-2008, 09:56 PM
If we are putting florescent anything on the targets, I would like to know sooner rather then later. Because as soon as we said "Oh hey, this piezo works well on a lexan shell", I changed my design to make the front canopy (which contains the cameras) to be target.

That's fine I don't mind. :veryhappy:

In all seriousness it shouldn't be that big of a deal. The target is 3x3 and the circles I'm thinking of are 3/4" in diameter, so it shouldn't be too hard to not stick them over your camera.

Like these (http://www.officemax.com/omax/catalog/sku.jsp?skuId=07022624&searchString=labels&productId=ARS25484&category_Id=654)

-Matt-
Cutting-Edge Robotics

devilDroid
01-07-2009, 04:40 PM
I'm playing catch-up here, so please forgive my ignorance... I thought that remote control was sort of the point of a 1st person point-of-view mech. A *real* (i.e. full-sized) mech would be piloted, correct?

devilDroid
01-07-2009, 04:42 PM
Then again, I suppose there's a lot of automatic assistance (like auto-targeting) to piloting a jet.

Adrenalynn
01-07-2009, 04:51 PM
Then again, I suppose there's a lot of automatic assistance (like auto-targeting) to piloting a jet.

FAF - Fire And Forget
FBW - Fly By Wire
Intelligent Radar
AWACS in "overseer" mode
intercept by ship radar
... The list is near endless.

devilDroid
01-07-2009, 06:07 PM
Still digging my way through the old posts... Adrenalynn, are you divulging your autonomy vs rc control plans?

Adrenalynn
01-07-2009, 06:35 PM
I think that will be decided on how much longer it takes me with this other robotics competition design. All depends how much time I have.