Well, they could do a small-scale GPS:

have 3 broadcast beacons - in 3 corners of the field - broadcasting an identical pseudorandom code, and have a reciever that calculates the differences in the signal to calculate it's position, then the OI could ask it for a position from the reciever whenever it needed it. it would be precice and fast.

or ultrasound
 

GPS works because there is a measurable time delay between when a signal is sent from space till the time it hits the ground. Beacons on the field would be too close to measure that delay.

You could do it with ultrasound though. Each RC gets a module that listens to commands over RF (or IR) and sends a chirp when it hears its name. Ultrasonic receivers are placed at the four corners of the field. The field controller cycles through the robots connected to the competition port, calling each one in turn, measuring the time till it hears a chirp at each receiver. It processes the data and sends the coordinates through the competition port.

Or you could reverse it and have the robot listen for chirps from the four corners. Each corner would send an RF or IR signal identifying itself and a chirp at the same time. The RC would measure the time between that signal and when it hears the chirp. I read a paper from CMU about a team of mini autonomous robots that use that method to determine their position.

That will be a real surprise to all the high school teams that already compete in the BotBall program competitions. BotBall is based on the MIT 6.270 Introduction to Robotics competitions. They have teams of high school students (no engineers), six weeks to build their robots, complex problems, multiple robots playing at the same time, and the robots are fully autonomous (no human control at all). And they have been doing it for the past seven years.

These teams have already addressed virtually every problem outlined in the referenced message (autonomous orientation and localization, hazard recognition and avoidance, dynamic environments [moving opponent robots], fault recognition and recovery, navigation error recovery, etc.). If they can do it, then I would see no reason that it could not be accomplished by FIRST teams as well.

To be clear, they use some different hardware. The BotBall control system is based on the Handyboard processor running Interactive C. But with a few modifications, a version of the IFI controller could have the same capabilities. They combine some very cheap sensors with some very smart students creating some very impressive code, and away they go...

-dave

Re: or ultrasound
 

A group in the Aerospace Robotics Lab at Stanford (currently run by Dr. Steve Rock, who is one of the mentors on Team 115) developed "GPS pseudolite" (pseuo-satellite) technology that takes care of this. They basically use ground-mounted Differential GPS transmitters to create an artificial GPS constellation, which can be placed within a small facility. Off-the-shelf commercial Trimble GPS receivers are then used to capture the signal and determine position (with one receiver) or position and orientation (with two receivers). As long as you know the position of the transmitters with good accuracy, you can get centimeter-level positioning accuracy out of the system (they are currently working on a self-calibrating pseudolite array that you would just plug in and let the pseudolites communicate between themselves to determine relative positions).

Because DGPS relies on the phase differences to determine range, rather than strict time delay, the transmitter-receiver distance can be quite small. The lab at Stanford uses the system to track the position and orientation of free-floating robots on their air bearing tables, and localizing Mars rover prototypes in demo fields. They frequently have transmitter distances less than two meters.

The only real cost of the system is in the pseudolite transmitters to establish the localization context (i.e. buy a few transmitters and place them on the edges of a robotic play field). The robots only need one or two receivers, which are cheap, and they get precise localization and orientation information that can plug straight into a control system.

-dave

what about having human inputted autonomous modes? I agree FIRST probably isn't ready for a full scale autonomous comp, but it might be able to handle small scale. Rather than having joysticks, have buttons/ toggle switches/etc, with each operating a different autonomous mode. for example, if we were still playing this years game, have one button hunt for the reflective tape on stacks, have another center the bot on the ramp, a third stack boxes, etc. basicly, each button would have the bot go into a different subroutine. this way, it would still be human controled, but autonomously operated! Also, in order to stop someone from having 4 buttons and emulating a joystick (ie, one goes forward, one goes back, one left, one right) have it set up so that the driver can only change the robot's subroutine every 5 seconds or so. a team could them use a joystick if they really wanted too, but it wouldn't work very well. just an idea which i doubt will ever happen, but i can see a lot of interesting strategies developing is it did.

I have no doubt that there are teams out there who can do it, but there are also a large number who could not (this is just my speculation, however). Without the right mentors, I think it would be very difficult. HOWEVER, what's the point in doing something if it's not difficult? ;)

I think it might be great to try it some year. FIRST doesn't seem afraid to try new things (see 2000 game)... And if it doesn't work out, they don't hesitate to change things back.

If FIRST analyzes this idea and determines it can do it in a way that will not turn away half the teams, let's go for it. I'd love to help mentor a team with this type of stuff!

- Patrick

Re: Blah
 

Can I join your Battle Bots team? Please dont make this full auto, thats what FLL is for.

Middle??
 

This would be a very difficult challange. Autonomous in the middle would lead to a situation where the initial state of the machine during this mode would be unknown, or at least a set of situations. THis is a ver difficult condition to deal with when planning your actions during this mode.

I'm sure some teams would drive to a 'known' position prior to the start of the autonomous mode, but this would only be the approximate starting condition and difficult to plan for.

But, I'm sure it would be interesting.....

leave it in the beggining until teams get past dead reckoning.
and let the drivers touch the joysticks during auto so your ready to go during human control mode.

i say get rid of autonomous, even though it offered a new challenge, everything happened during autonomous mode. during autonomous, nearly all the stacks were knocked over, then it became a war for the king of the hill position and pushing the opposing alliance's boxes out of their scoring zone. i prefer the 100% human control, that way a team's opening move isnt ALWAYS the same, going nearly perfectly everytime, making it difficult for teams w/o autonomous. full driver control tests the driver's control overall since it could end up different, with the driver miscalculating a distance or another robot's speed. my $.02

Give me a better controller with C, some better sensors, and a lot of caffeine, and I won't care when or how long autonomous is. It would give some members of my team have to redeem ourselves for failure to complete our autonomous programming independent study course...

Does this evil man work for NASA? :D

And I think having both autonomous and driver control is a good direction to go, because autonomy is fun (along with something about the definition of "robot") and driver control allows for extreme unpredictability.

Frankly I like the mix of autonomy and human control and think that it should stay that way. The complete autonomy would be a challenge, but it would also be very uneventful.

They would loose almost all none-participant spectators! 4 robots driving into wells and getting stuck on bins isnt very exciting for 2 minutes. 15 seconds of autonomous keeps things interesting...

If it whole thing was autonomous, what would the drivers do? And in the middle would be great. Teams would be forced to add sensors to make the robots actual robots, instead of just glorified remote controlled cars.

if we had better sensors, a better language, and a slightly faster chip (for instance, the light sensors ran so fast, you may have missed a signal from them), i think autonomous anywhere would be at least feasible. at the moment though, uh, no, it won't work very well.