PDA

View Full Version : Radar for Mech Warfare?



mannyr7
04-29-2010, 05:30 PM
Thought I'd share this link. Enjoy!

http://luckylarry.co.uk/2009/11/arduino-processing-make-a-radar-screen-part-3-visualising-the-data-from-sharp-infrared-range-finder/

darrellt
04-29-2010, 06:44 PM
Wow that's pretty cool. My GUI is already in processing and I have an IR sensor so i may have to try that.

mannyr7
04-29-2010, 06:51 PM
Yeah, I thought of you specifically, when I saw this.

jes1510
04-29-2010, 07:29 PM
Search the forum for "Mechdar". It's a really clever project that Fergs started but it got put aside in favor of the scoring system.

lnxfergy
04-29-2010, 07:49 PM
Back before I was even building a mech last year, Andrew had asked me to build something like this. MechDAR consisted of a Mini RoboController (same as the target transponder), 5 IR sensors (one to each side, 2 off the rear, and one on a panning turret), and an XBEE link back. The (unfinshed) code & docs can be found here (http://vanadiumlabs.com/mechdar/) (it was a python/Pygame gui).

Some thoughts from having played with this setup:


Putting a forward and rearward facing IR sensor, each on their own pan servo, would give awesome coverage.
If you're using XBEEs, you'll need a seperate XBEE for the MechDAR from your control -- the turnaround time from RX to TX to RX may quickly slow down your link, causing erratic behavior (unless you are sending movement commands at a really slow framerate)
Speed of a scan is critical, and the part missing from the code linked is the update for movement. That code Manny linked to will take 3.6s to pan 180 degrees. At 3.6s, the bot will move during a scan, causing points not to be aligned correctly. You could tie in your control outputs, to apply a transform to the points, or speed up the panning.
Please, please, make the bot move enough that a 3.6s update rate makes this unusable before bothering building such a system. (I'd really like to see mobility, more than smarter turrets, and I would think most people agree).

I actually planning to revive this system for next year, because I plan to build a biped. Issy was too fast moving to make this worthwhile, but a slower quad, which can only move 10cm/s or so, would probably benefit a lot. I do like their method of extending the dots out using that radar appearance -- I'm thinking I may go that route (rather than the original, wall extraction/localization scheme I had been working on).

EDIT: Jes beat me to the punch.

-Fergs

DresnerRobotics
04-30-2010, 07:34 AM
Fergs, perhaps we should bring up the 'other' radar idea?

mannyr7
04-30-2010, 09:12 AM
I'm thinking the scan could be sped up considerably as you only need to scan once every 8"-12" at 15' maximum. My trig fails me at the moment as to what angle this comes out to.

lnxfergy
04-30-2010, 09:57 AM
Fergs, perhaps we should bring up the 'other' radar idea?

Sure,

The discussion has been (between Andrew, I, and a few others), that we need some way for competitors to figure out where the heck they are. The cameras aren't that great, and if we were to get two very mobile mechs running around, they could easily miss each other.

The proposed solution was to install a camera over the arena, which would track the location of the two mechs, each competitor would be shown a screen that always shows their location, relative to the arena. Occasionally (1s on, 5-10s off?) they would see a blip that tells them were their opponent is. There are, of course, issues with such an infrastructure:


How do we provide this heads up map to competitors? (who are on separate networks, running different GUIs, etc)
Does it make things too easy?

Thoughts?

-Fergs

mannyr7
04-30-2010, 10:38 AM
How about a small java applet which could be easily integrated into the users choice of GUI?

gdubb2
04-30-2010, 11:29 AM
How about just using a common monitor hardwired to the camera. Have the competitors sitting a little closer togther than we were, and just share the monitor, have the blips update the same for each one. An occasional update would still be better than we have now, and this would eliminate any problems with everyone having to have some sort of connection to the camera.

It's cumbersome as all get out, but I have a 19" CRT monitor I would donate to the cause.

Gary

Stobs
04-30-2010, 11:45 AM
How about just using a common monitor hardwired to the camera. Have the competitors sitting a little closer togther than we were, and just share the monitor, have the blips update the same for each one. An occasional update would still be better than we have now, and this would eliminate any problems with everyone having to have some sort of connection to the camera.

It's cumbersome as all get out, but I have a 19" CRT monitor I would donate to the cause.

Gary

If that's the route to go then if splitting the signal between two monitors works with that I have an old LCD screen I can bring along too so the combatants don't need to share one monitor.

[EDIT] PS: It's a 17" with pretty crappy resolution but (it) should suffice for the purposes. http://www.westinghousedigital.com/details.aspx?itemnum=29

MSK Mech Commander
04-30-2010, 02:56 PM
Actually, if two mechs were confused about where they were, maybe we should just put colored flags for every quadrant of the arena. It certainly would be a lot less work.

I don't know if this might be a problem, but the ten second blip period might be distracting to some pilots who are trying to navigate through their surroundings. It might be hard trying to predict where your opponent is going to move next while you yourself are piloting your mech. In ten seconds, an agile mech could easily move away from the location it was spotted in. If it was every three seconds, it could be proven more effective as you can see which way your opponent is going.

Plus trying to get a camera mounted from a top-down view and getting it hooked up to two displays might be more work than necessary. I also think that this would take a lot of the fun away and make it too easy. Instead of focusing on combined strategy, stealth, etc, we might end up making this game about who has the better aiming system. In a sense, the battles would be a lot more direct, and less balanced between direct and indirect.

The colored flags could be put in the center of the quadrant. That way competitors won't have to memorize every landmark, just a few.

DresnerRobotics
04-30-2010, 03:08 PM
Plus trying to get a camera mounted from a top-down view and getting it hooked up to two displays might be more work than necessary. I also think that this would take a lot of the fun away and make it too easy. Instead of focusing on combined strategy, stealth, etc, we might end up making this game about who has the better aiming system. In a sense, the battles would be a lot more direct, and less balanced between direct and indirect.

Just want to add; you'd be surprised at how easy it is to get lost in there when you're piloting. We saw multiple mechs getting pretty disoriented. It's an entirely different ballgame when you've actually built a mech and tried piloting it in an arena against an enemy, I think your perspective of what's easy might change. Hell, I lost track of where S.A. was a few times, and he was right under my nose (almost literally).

The output display of the proposed mech would just be a grid with the buildings/borders marked, and a blip-dot of where you and your opponent are flashing once every 10-15 seconds or so. Given that radar was standard issue equipment for most mechs in the Battletech universe, I don't see an issue with an implementation of this.

Another thing to note is that it probably wouldn't be effective against stationary mechs, so you could still employ a bit of stealth.

MSK Mech Commander
04-30-2010, 03:24 PM
Just want to add; you'd be surprised at how easy it is to get lost in there when you're piloting. We saw multiple mechs getting pretty disoriented. It's an entirely different ballgame when you've actually built a mech and tried piloting it in an arena against an enemy, I think your perspective of what's easy might change. Hell, I lost track of where S.A. was a few times, and he was right under my nose (almost literally).

The output display of the proposed mech would just be a grid with the buildings/borders marked, and a blip-dot of where you and your opponent are flashing once every 10-15 seconds or so. Given that radar was standard issue equipment for most mechs in the Battletech universe, I don't see an issue with an implementation of this.

Another thing to note is that it probably wouldn't be effective against stationary mechs, so you could still employ a bit of stealth.

Good to know, just proposing a devil's advocate point of view.

Even though I have pretty much no experience whatsoever, hooking up a HUD to several models of computers, networks, codes, yada, would be really hard. So, we'll most likely will end up with a satellite imaging monitor(s). Colored flags for landmarks might be something to consider on top of the "satellite" imaging. That way pilots could navigate without craning their heads to the side to look at the satellite monitor if they wanted to. But it's just an idea should we need it if ever. By that, I mean I doubt we'll ever need it. But hey, things happen.

darkback2
04-30-2010, 03:32 PM
I have an IR sensor on the front of each squidwords front feet. While walking him around the house it really helped because each foot sweeps almost 90 degrees per step. I'm thinking of adding a front and back IR sensor...

Then again...I should probably just focus on getting him to stop walking in circles.

DB

mannyr7
04-30-2010, 06:42 PM
Just want to add; you'd be surprised at how easy it is to get lost in there when you're piloting. We saw multiple mechs getting pretty disoriented. .

Let's have Fon put in street signs next year or themed parts of town, ie. Chinatown...:veryhappy:

darkback2
05-01-2010, 01:08 AM
I know I'm in the minority here when I say I thought it was sort of fun having people watching screaming...he is to the left...Oh!!! turn...turn

It was sort of like bootleg radar.

A top down view...I sort of think people-dar is more fun...we would just have to limit what the people are saying to things like...Left, right, your hitting a building...that sort of stuff. It also makes it so that we can have plants yelling out the wrong directions. :p

DB

lnxfergy
05-01-2010, 01:28 AM
I know I'm in the minority here when I say I thought it was sort of fun having people watching screaming...he is to the left...Oh!!! turn...turn

It was sort of like bootleg radar.

A top down view...I sort of think people-dar is more fun...we would just have to limit what the people are saying to things like...Left, right, your hitting a building...that sort of stuff. It also makes it so that we can have plants yelling out the wrong directions. :p

DB

DB -- have you ever watched some of the fire-fighting matches? Those are hilarious displays of "people-dar" trying to tell the robot "no, go back, you missed the fire!"

-Fergs

MikeG
05-01-2010, 08:22 AM
If you guys have the overhead area camera figured out, then what about shooting the image to a web server. Most high level languages have libraries for hitting the Internet/Intranet. The Mech commanders can either integrate into their existing UI or use a separate browser instance during the match.

lnxfergy
05-01-2010, 10:35 AM
If you guys have the overhead area camera figured out, then what about shooting the image to a web server. Most high level languages have libraries for hitting the Internet/Intranet. The Mech commanders can either integrate into their existing UI or use a separate browser instance during the match.

Yeah, the issue is that competitors are each on different physical networks -- not many (any?) laptops have dual network cards, so we're looking at dragging a full PC with dual network cards around for this.... unless there's something I'm missing?

-Mike

darkback2
05-01-2010, 11:01 AM
I guess it was when we were first figuring this out...I suggested using spotters. If each participant had a person, sort of a copilot that could see the arena directly, but had a limited number of words they could use, it would eliminate a whole lot of the necessary tech, and solve the problem of robots not being able to find each other, or getting hung up on things that they can't see.


This would in essence speed up a match, but would eliminate some of the problems.

One thing would be having it so that the spotters could only tell where the other mech was, and not where it was going. An overhead camera would make it so that sneeking up behind a mech would be virtually impossible, and would increase the likely hood of face on shootouts, especially with quads that can straif.

Having an overhead camera would fundamentally change the game by eliminating line of sight issues. A pilot could look at a screen and know where the other guy is, and what he is up to. Having spotters or copilots separates this out so that the pilot's responses are not as direct.

Just thinking aloud.

DB

lnxfergy
05-01-2010, 11:08 AM
Well, the other reason I like this ceiling camera radar system is that we could use it to determine when a mech has "moved" for scoring purposes... thus taking the guess work out of it.

-Fergs

MikeG
05-01-2010, 11:10 AM
Sorry, I made an assumption about the network setup.

Upgrayd
05-01-2010, 11:59 AM
If I am following this correctly the radar proposed is to be used for three things..

1. determine movement for penalty purpose
2. allow pilots to easily determine where they are in the arena
3. allow a pilot to easily track his opponents movement

Without knowing how the will be implemented or how the information may be relayed to the pilots (video or simply blips on a screen) it seems like the last two would take a bit of strategy out.

Speaking for myself I felt there were more then enough clues in the arena to determine where I was at all times. First thing I did was memorize the color/shapes of the buildings along with some of the features (cop cars, crashed Humvee). I could combine the visual clues with the amount of possible paths I could move (two paths = corner, three paths = side, four paths = center) and always keep my bearing straight.

Robonaut
05-01-2010, 12:26 PM
The output display of the proposed mech would just be a grid with the buildings/borders marked, and a blip-dot of where you and your opponent are flashing once every 10-15 seconds or so. Given that radar was standard issue equipment for most mechs in the Battletech universe, I don't see an issue with an implementation of this.

Another thing to note is that it probably wouldn't be effective against stationary mechs, so you could still employ a bit of stealth.

In reading this it sounds like you would be using a regular camera and some motion detection to tracks the robots. If so I was thinking about a slightly different set up.

I got the idea from watching some of Johhny Lee works that he has done with the Wii remote. He made a Wii remote into a head tracker ( amoung other cool things) for 3D environments. The Wii remote is a really cool remote, it has a IR camera and accelerometers in it for a decent price. http://johnnylee.net/projects/wii/

If you had an IR camera mounted above the arena it could track IR sensors that would be attached to the upper part of the robot pointing upwards. Maybe a few could be used to better detect the robot depend on it's given pose. The IR sensor would be low weight, low cost and would have a small foot print. I would also think radio frequencies would not be an issue it the camera was hard wired.

sthmck
05-01-2010, 12:34 PM
I am with Ryan on this one. There was really only one time when I got lost in the arena, and that more to do with my frame rate dropping to right around 4 or 5 than anything. I think I ended up getting turned around while I couldn't see. Other than that the buildings were different enough that I was able to figure out where I was.

sthmck
05-01-2010, 12:38 PM
If I am following this correctly the radar proposed is to be used for three things..

1. determine movement for penalty purpose
2. allow pilots to easily determine where they are in the arena
3. allow a pilot to easily track his opponents movement

Without knowing how the will be implemented or how the information may be relayed to the pilots (video or simply blips on a screen) it seems like the last two would take a bit of strategy out.

Speaking for myself I felt there were more then enough clues in the arena to determine where I was at all times. First thing I did was memorize the color/shapes of the buildings along with some of the features (cop cars, crashed Humvee). I could combine the visual clues with the amount of possible paths I could move (two paths = corner, three paths = side, four paths = center) and always keep my bearing straight.

This seems to be a pretty reasonable solution. The only problem I could see would result from a building shielding the IR cameras view.

Robonaut
05-01-2010, 12:58 PM
This seems to be a pretty reasonable solution. The only problem I could see would result from a building shielding the IR cameras view.

More than one camera?

darkback2
05-01-2010, 07:21 PM
I am willing to sit out next year and official so long as I get to participate in an exhibition match or two. If someone is willing to write a scoring guide I am more than willing to follow it to the best of my abilities. I'm sure Maya would also be willing provided it doesn't interfere with her entering the humanoid competitions.

Allowing pilots to see the arena fundamentally changes the game.

Having spotters / copilots and an official solves these problems without fundamentally changing the game, or requiring more crap (I mean stuff) that can go wrong or that we have to worry about.

Just an offer,
DB

MSK Mech Commander
05-02-2010, 03:09 PM
An overhead camera would make it so that sneeking up behind a mech would be virtually impossible, and would increase the likely hood of face on shootouts, especially with quads that can straif.

Having an overhead camera would fundamentally change the game by eliminating line of sight issues. A pilot could look at a screen and know where the other guy is, and what he is up to.
DB



Allowing pilots to see the arena fundamentally changes the game.
DB

That's what I was thinking. I have to admit I was sort of surprised to hear that you guys had navigating issues right after seeing some of the videos from Robogames with all of those tanks, crashed cars, etc. If there's still problems navigating, we could just put up billboards advertising different things around the arena.

Having an overhead camera seems like a lot of overkill to figure out where you are. Not only that it looks like we're re-inventing the wheel by doing this, but I will drastically change the game. It will all come down to who has the faster mech, aiming system, and piloting skills. I think that this would kill all of the strategy of the warfare involved. Although I agree that mechs that far in the future would have radar, their enemies would also have a jamming device countering their radar.

I don't know about you guys, but it seems that this conversation is running in circles and we should just leave it the way it is for now. Maybe we could do some testing of this satellite imaging during the exhibition matches at Robogames. But for now I think we can leave the fundamental gameplay the way it is. As more competitors join and as time passes, there will be more and more advances on all the upgrades our mechs have currently. Sooner or later, someone will find out a way to incorporate mechdar that everyone can use.

Let's just leave it the way it is and not cheat the process.

sthmck
05-02-2010, 04:06 PM
That's what I was thinking. I have to admit I was sort of surprised to hear that you guys had navigating issues right after seeing some of the videos from Robogames with all of those tanks, crashed cars, etc. If there's still problems navigating, we could just put up billboards advertising different things around the arena.

Having an overhead camera seems like a lot of overkill to figure out where you are. Not only that it looks like we're re-inventing the wheel by doing this, but I will drastically change the game. It will all come down to who has the faster mech, aiming system, and piloting skills. I think that this would kill all of the strategy of the warfare involved. Although I agree that mechs that far in the future would have radar, their enemies would also have a jamming device countering their radar.

I don't know about you guys, but it seems that this conversation is running in circles and we should just leave it the way it is for now. Maybe we could do some testing of this satellite imaging during the exhibition matches at Robogames. But for now I think we can leave the fundamental gameplay the way it is. As more competitors join and as time passes, there will be more and more advances on all the upgrades our mechs have currently. Sooner or later, someone will find out a way to incorporate mechdar that everyone can use.

Let's just leave it the way it is and not cheat the process.

I guess this means we will being seeing you next year at robogames?

MSK Mech Commander
05-02-2010, 04:53 PM
I guess this means we will being seeing you next year at robogames?

I certainly hope so. This time I'll be prepared for April. I couldn't compete in this year's Robogames because I thought that it would be conducted in June again. The change in months, unfortunately, wrecked my plans and postponed them.

I spent a lot of time last summer trying to scrape up enough cash to buy a base for a mech. Now that I'm done with the huge purchases (for now), I will be able to spend a lot more time on outfitting my Phoenix with the necessaries and slightly edit the code to suit my preferences.

mannyr7
05-02-2010, 07:18 PM
.... But for now I think we can leave the fundamental gameplay the way it is. As more competitors join and as time passes, there will be more and more advances on all the upgrades our mechs have currently. Sooner or later, someone will find out a way to incorporate mechdar that everyone can use.

Let's just leave it the way it is and not cheat the process.

The point of my original post. A "Gee, wow! That's cool" element to add to your GUI/HUD.

DresnerRobotics
05-02-2010, 07:37 PM
This thread is here for discussion, nothing is being changed within the game rules without some serious thought towards it.

A few things I want to add if we were to try something like this: It wouldn't be a straight up video feed. It would likely end up being a 64x64 block output that shows border, buildings, and occasionally (once every 10-30 seconds) a blip of each mech. Not quite the same thing as a continuous video feed.

We are not going to have 'spotters'. In fact, we're going to be setting rules in place that prevent ANY communication to the pilots from bystanders outside of a ref calling out approved feedback (such as, "Odin, you have left the square your weapons are hot, etc). Not meaning to sound like a hard ass on this, but we want to focus on improving technology here rather than people yelling info back and forth.

Another thing we might end up looking into once again is MechDAR, which is currently 100% allowed per the current rules. This might help considerably with piloting to the point where we wouldn't need feedback from an external radar system.

mannyr7
05-02-2010, 07:44 PM
Back on topic. Thank you! :)

MSK Mech Commander
05-02-2010, 09:29 PM
This thread is here for discussion, nothing is being changed within the game rules without some serious thought towards it.

A few things I want to add if we were to try something like this: It wouldn't be a straight up video feed. It would likely end up being a 64x64 block output that shows border, buildings, and occasionally (once every 10-30 seconds) a blip of each mech. Not quite the same thing as a continuous video feed.

Another thing we might end up looking into once again is MechDAR, which is currently 100% allowed per the current rules. This might help considerably with piloting to the point where we wouldn't need feedback from an external radar system.

Of course, I'm just participating in said discussion. If I'm not, please correct me.

Although it wouldn't be a straight up continuous stream of video, it would dramatically change the game without any individual prowess required to do so. But considering the speed of the mechs, if you were going to do this, you might want to make the blip interval 30 or 45 seconds long. This way it won't be a clear feed of what your opponent is up to, but you also get a view of where you forgot to look.

Either we can have the interval blip, or each pilot can have three "peeks" which allow him/her to get a spy satellite image. I think limited peeks might be better than the interval blips, but I'm not sure.

DresnerRobotics
05-03-2010, 11:23 AM
Of course, I'm just participating in said discussion. If I'm not, please correct me.

Although it wouldn't be a straight up continuous stream of video, it would dramatically change the game without any individual prowess required to do so. But considering the speed of the mechs, if you were going to do this, you might want to make the blip interval 30 or 45 seconds long. This way it won't be a clear feed of what your opponent is up to, but you also get a view of where you forgot to look.

Either we can have the interval blip, or each pilot can have three "peeks" which allow him/her to get a spy satellite image. I think limited peeks might be better than the interval blips, but I'm not sure.

Was a general statement, not directed at you. I was just trying to explain that we're opening it up for discussion rather than telling people it was something that was without a doubt going to be implemented.

The more I think about it the more I might want to shy away from universal radar and move towards onboard radar for obstacle avoidance, such as mechdar.

Connor
05-03-2010, 11:43 AM
With 2-on-2 matches, universal radar could be set it up so we could maintain a active picture of where our team mate is, so that we could coordinated our strategy etc.. Something that couldn't be done with Onboard mechdar very easily (if at all). This is very much in the realm of the BattleTech Board game, and Mech Warrior series of video games.

Thanks, Connor

master_of_robots
05-03-2010, 12:27 PM
Here is a thought. I really like the idea of an overhead "satellite" view, but of course you could see where the opponent is. How about using Augmented Reality to not only provide a cool looking terrain but also show/hide each mech depending on the person's view. That might sound confusing at first, but here is a video to help:

YouTube- ARhrrrr - An augmented reality shooter

Player 1 could see his mech, and the entire map, but the AR view would block out Player 2.
Player 2 could see his mech, and the entire map, but the AR view would block out Player 1.

By putting AR markers on the top of each mech, you could easily keep track of each mech.

(FYI, in the video above, it is understood that the camera would be fixed above the arena)

darrellt
05-03-2010, 09:00 PM
There is a complicating factor with trying to use a single AR camera. If you assume the mechs are 1ft tall, and the buildings are two feet tall, and you will tolerate a 2 inch view shadow from the buildings, the camera would have to be at least 16ft high. Some of the buildings are taller then this, esp with the billboards etc, so perhaps even higher. This makes an easy to set up, secure mounting solution somewhat tricky. Also you would need a pretty high resolution cam to pick up an AR marker at 16+ feet away.

I was thinking a tracking system could be used for automating the "must not stand still for more then 20 seconds rule", but their are a lot of things you can do if you know the XY pos of each mech. If the camera mounting problem could be solved I could run the video through processing and have a blocky mechdar type display fairly easily, but l think mounting the camera could be a pain.

I personally did not find getting lost to be much of a problem, but my mech never walked very far or very fast, and it had a compass.

Stobs
05-11-2010, 12:40 AM
I'm researching on whether I want to go with a flavor of the Arduino controller or that of the ArbotiX (presuming that I'm comparing apples-to-apples of some variation, and not to the proverbial orange instead), and came across this three-part series on Arduino-based radar. I have to say very little of this makes sense to me as yet, although I understand the general concepts. In another thread I'm going to ask about Arduino vs. ArbotiX - after I do some more research on both, but for now I figured I'd post this as food for thought, once again presuming it hasn't been considered as yet.

Arduino + Processing: Make a Radar Screen to Visualise [sic] Senosr Data from SRF-05 - Part 1: Setting up the Circuit and Outputting Values (http://luckylarry.co.uk/2009/11/arduino-processing-make-a-radar-screen-to-visualise-sensor-data-from-srf-05-part-1-setting-up-the-circuit-and-outputting-values/)