www.robowars.org

RoboWars Australia Forum Index -> General Chatter

Towards Autonomous Combat
Goto page Previous  1, 2, 3, 4, 5, 6, 7, 8, 9  Next

Post new topic   Reply to topic
  Author    Thread
marto
Experienced Roboteer


Joined: 08 Jul 2004
Posts: 5459
Location: Brisbane, QLD


 Reply with quote  

This is getting repetitive.

Might see if I can get proof of concept going this weekend.
_________________
Steven Martin
Twisted Constructions
http://www.botbitz.com

Post Wed Dec 01, 2010 12:38 am 
 View user's profile Send private message Send e-mail MSN Messenger
Valen
Experienced Roboteer


Joined: 07 Jul 2004
Posts: 4436
Location: Sydney


 Reply with quote  

If somebody wants to test it at home they setup the same system that would be at the arena.

What is simple about driving a robot in a combat arena?

What is your proposed system?

An IMU is ~$100 from sparkfun and you would use it for high rate correction of angle, the PC would give it longer term information.
Its basic control systems theory, you use a fast noisy sensor and correct it with a slow absolute sensor.

Especially as getting accurate orientation information from the video is going to be very difficult, commands from the PC would be of the form, more left, more right based on velocity vector.

Trying to put the sensors on a moving platform that has scissorhands trying to bash its face in is doomed to failure/lots more work for no real benefit.
You change your control system problem from a 2D image manipulation into 3D spatial awareness, your sensor system now needs to scan its entire environment and account for the sensor platform being deformed and destroyed whilst its in use. Anything that can see outside the bot, is going to get destroyed eventually.
you have now drastically increased the processing power required, and required that this processing power be in perhaps the most hostile environment on the planet for electronics. (ok the stuff they put in copperhead laser guided tank shells might have a higher G-loading, but its only once and only in the one direction)
It also means that you aren't going to have autonomous beetle or ant weight bots, well nothing thats better than a turtle anyway.
_________________
Mechanical engineers build weapons, civil engineers build targets

Post Wed Dec 01, 2010 8:28 am 
 View user's profile Send private message Visit poster's website AIM Address Yahoo Messenger MSN Messenger ICQ Number
chris...



Joined: 20 Oct 2010
Posts: 28


 Reply with quote  

quote:
Originally posted by Valen:
If somebody wants to test it at home they setup the same system that would be at the arena.

You're proposing that all builders will need a similar setup at home as for the arena. Because builders will have to test it somewhere, thats one way to make a costly hobby even more costly.


quote:
What is simple about driving a robot in a combat arena?

You have to write the anticipations of situations into the software, the human controller doesnt


quote:
What is your proposed system?

Using a kinect to verify an IMU is not a good idea
Have a different league where its self contained bot vs bot that way all disadvantages are equal


quote:
An IMU is ~$100 from sparkfun and you would use it for high rate correction of angle, the PC would give it longer term information.
Its basic control systems theory, you use a fast noisy sensor and correct it with a slow absolute sensor.

You plan to filter for adjustment of input with a high rate noisy sensor, and can determine the difference between noise and vibrations.
The kinect doesnt suffer from noise, paralax errors, clipping and is accurate enough at the distance to cover the entire arena?

Another question will their be any obstacles, know or unknown?
If unknown then youll need more snesors on the robot
If known, then avoiding them will be trivial

Post Wed Dec 01, 2010 9:09 am 
 View user's profile Send private message
marto
Experienced Roboteer


Joined: 08 Jul 2004
Posts: 5459
Location: Brisbane, QLD


 Reply with quote  

On iPhone so hard to quote.

If you think you can build a reliable control system onboard for less than $200 go for it. Look at my proposed weight bonuses in the first post.

That different class you speak of is called robot sumo. It's lame.

I am not sure what you mean by error. Yes there is error you do your best to account for it. Other than that you just plan with the uncertainty. The main point of the IMU is to interpolate between positions for the kinect. I would not bother with IMU if a tracking system (kinect or visual) is sufficient.
_________________
Steven Martin
Twisted Constructions
http://www.botbitz.com

Post Wed Dec 01, 2010 10:00 am 
 View user's profile Send private message Send e-mail MSN Messenger
Jaemus
Experienced Roboteer


Joined: 01 Apr 2009
Posts: 2674
Location: NSW


 Reply with quote  

chris, i think you need to see the results of damage say scissorhands or demon can do, up close and in person on the bots they've hit, before you go installing delicate sensors on a bot that might be facing anything even close to them
_________________
<Patrician|Away> what does your robot do, sam
<bovril> it collects data about the surrounding environment, then discards it and drives into walls

Post Wed Dec 01, 2010 1:24 pm 
 View user's profile Send private message Send e-mail MSN Messenger
chris...



Joined: 20 Oct 2010
Posts: 28


 Reply with quote  

quote:
Originally posted by marto:
On iPhone so hard to quote.

If you think you can build a reliable control system onboard for less than $200 go for it. Look at my proposed weight bonuses in the first post.

Not likely going to cost $200 but by doing it the way you are proposing is forcing every automatic robot builder to do it the same way. You'll need to set some framework for the builders to follow.
Do the programs run on the users computer or a shared one?
Can you conduct cyber warfare at the same time?
What kind of information do the players get from the kinect system?



quote:
That different class you speak of is called robot sumo. It's lame.

really, introduce rules to not make it lame, what does robot sumo end up being? and what's stopping someone from making a lame sumo robot?


quote:
I am not sure what you mean by error. Yes there is error you do your best to account for it.

IMU data is relative to the last reading, its like measuring out 1km with a 1m stick. With the vibration from the machine and noise from the device you'll need to have some kind of filter or moving average which will require a few assumptions.

So first up your saying if you make it all onboard and automated you end up making a robot sumo, therefore an automated scissors hands or demon will loose against an automated sumo?

Post Wed Dec 01, 2010 4:01 pm 
 View user's profile Send private message
chris...



Joined: 20 Oct 2010
Posts: 28


 Reply with quote  

Can you post a compressed bag file of the kinect with the camera static amd looking down at a moving object, and make it short, like a few seconds with PC data

Last edited by chris... on Sat Dec 04, 2010 1:00 pm; edited 2 times in total

Post Fri Dec 03, 2010 4:27 pm 
 View user's profile Send private message
Glen
Experienced Roboteer


Joined: 16 Jun 2004
Posts: 9481
Location: Where you least expect


 Reply with quote  

Got the time to read the thread properly finally.

Intruiging concept! If i understood it right then you'd be keeping all the delicate sensors off the robots and use a kinect mounted to the roof looking down and do the actual processing on a computer that would transmit the info to each robot through the standard radio gear.

Makes autonomous robots at least viable as the sensors are realistically never going to stand up to the insanity of combat conditions + would i be right in saying that you could literally adapt any current robot to work with a system like this?

Could also use the power of a computer to get some pretty complex strategies and the like going as well vs microprocessors on the robots themselves, which would always seem to be the downfall of autonomous combat robots IMHO, that they cant do much more then ram into each other sumo bot style.

If it could be made to work (and with all this over the counter stuff like kinect happening now its probably not all that unreasonable) then it could be a massive progression for the hobby Cool


quote:
quote: Originally posted by marto:
On iPhone so hard to quote.

If you think you can build a reliable control system onboard for less than $200 go for it. Look at my proposed weight bonuses in the first post.


Not likely going to cost $200 but by doing it the way you are proposing is forcing every automatic robot builder to do it the same way. You'll need to set some framework for the builders to follow.
Do the programs run on the users computer or a shared one?
Can you conduct cyber warfare at the same time?
What kind of information do the players get from the kinect system?


quote: That different class you speak of is called robot sumo. It's lame.


really, introduce rules to not make it lame, what does robot sumo end up being? and what's stopping someone from making a lame sumo robot?

quote: I am not sure what you mean by error. Yes there is error you do your best to account for it.


IMU data is relative to the last reading, its like measuring out 1km with a 1m stick. With the vibration from the machine and noise from the device you'll need to have some kind of filter or moving average which will require a few assumptions.

So first up your saying if you make it all onboard and automated you end up making a robot sumo, therefore an automated scissors hands or demon will loose against an automated sumo?


Why dont you just do up a nice long post of your system then. This wrestling forum-esque rhetorical questioning and qoute-replying is growing old rather fast.

Post Fri Dec 03, 2010 8:50 pm 
 View user's profile Send private message Send e-mail Visit poster's website MSN Messenger
Valen
Experienced Roboteer


Joined: 07 Jul 2004
Posts: 4436
Location: Sydney


 Reply with quote  

there is no real need to use a kinect specifically, a $5 webcam will probably suffice though the kinect might add some more detail/flexibility to working out the orientation of the bots without having to try and do stereo imaging
_________________
Mechanical engineers build weapons, civil engineers build targets

Post Fri Dec 03, 2010 10:41 pm 
 View user's profile Send private message Visit poster's website AIM Address Yahoo Messenger MSN Messenger ICQ Number
chris...



Joined: 20 Oct 2010
Posts: 28


 Reply with quote  

Parts of the scene are white when they should not be, what's causing it?

Post Sat Dec 04, 2010 11:28 am 
 View user's profile Send private message
marto
Experienced Roboteer


Joined: 08 Jul 2004
Posts: 5459
Location: Brisbane, QLD


 Reply with quote  

No return. It's where it's absorbing IR
_________________
Steven Martin
Twisted Constructions
http://www.botbitz.com

Post Sat Dec 04, 2010 1:57 pm 
 View user's profile Send private message Send e-mail MSN Messenger
chris...



Joined: 20 Oct 2010
Posts: 28


 Reply with quote  

Like dark items,
offset/shadows
glass

Getting PC data from a depth image is not very good
Their no calibration information and the data is only 8bit
Can you post a compressed bag file single frame of PC data from the kinect, looking down at 2 objects on the floor

The compressed depth image file you posted goes down to 24Mb almost 10 times smaller

Post Sat Dec 04, 2010 4:33 pm 
 View user's profile Send private message
marto
Experienced Roboteer


Joined: 08 Jul 2004
Posts: 5459
Location: Brisbane, QLD


 Reply with quote  

Interesting descriptor which could do exactly what I want for robots.

http://www.ros.org/wiki/vfh_cluster_classifier

Would give you a 6DOF position.

Steve
_________________
Steven Martin
Twisted Constructions
http://www.botbitz.com

Post Mon Feb 21, 2011 2:32 am 
 View user's profile Send private message Send e-mail MSN Messenger
marto
Experienced Roboteer


Joined: 08 Jul 2004
Posts: 5459
Location: Brisbane, QLD


 Reply with quote  

http://info.ee.surrey.ac.uk/Personal/Z.Kalal/tld.html

Off the shelf solution to robot tracking. however no rotational component Sad
_________________
Steven Martin
Twisted Constructions
http://www.botbitz.com

Post Wed Apr 13, 2011 4:23 pm 
 View user's profile Send private message Send e-mail MSN Messenger
marto
Experienced Roboteer


Joined: 08 Jul 2004
Posts: 5459
Location: Brisbane, QLD


 Reply with quote  

Had reasons to get the above algorithm running for uni.

http://dl.dropbox.com/u/815267/test.mp4

Now to apply it to robot.

Steve
_________________
Steven Martin
Twisted Constructions
http://www.botbitz.com

Post Fri Jun 17, 2011 4:23 pm 
 View user's profile Send private message Send e-mail MSN Messenger
  Display posts from previous:      

Forum Jump:
Jump to:  

Post new topic   Reply to topic
Page 4 of 9

Goto page Previous  1, 2, 3, 4, 5, 6, 7, 8, 9  Next

Forum Rules:
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum

 

Last Thread | Next Thread  >
Powered by phpBB: © 2001 phpBB Group
millenniumFalcon Template By Vereor.