Finding the way to docking station

[Tomomastchi-project]

Hi,

I want my robot to be able to find it's way to the docking station where it can recharche it's lead accu. Do you guys think it is possible to let it find the docking station through IR or should I use another system(i.e. laser, sonar)?

Thanks

[airman00]

The problem with IR , laser, and sonar is that the docking station has to be in "view" of the robot

Why not try RF ? Triangulate your position with 3 transmitters in a room and you can find your way around to the docking station. Then use IR when the robot is close and needs to align to the charger.

[Tomomastchi-project]

I think it's going to be too complicated then. It doesn't have to have a huge range. A couple of metres is OK. Do you think IR is fine for that?

[Admin]

Why not try RF ? Triangulate your position with 3 transmitters in a room and you can find your way around to the docking station. Then use IR when the robot is close and needs to align to the charger.

won't work, because you're asking your sensors to measure the speed of light in an area not bigger than a room . . .

Just do the iRobot Roomba method of IR, but make it really high powered and modulated for good range . . .

[airman00]

Why not try RF ? Triangulate your position with 3 transmitters in a room and you can find your way around to the docking station. Then use IR when the robot is close and needs to align to the charger.

won't work, because you're asking your sensors to measure the speed of light in an area not bigger than a room . . .

why not ? You detect your distance from each transmitter . Remember that toy that I wanted to mod that had RF ranging on it?

[Admin]

well, a clock cycle at 16Mhz is every 6.25 × 10^(-8) seconds

however, light travels at about 1 foot per nanosecond . . .

lets say the room is 10 ft x 10 ft for simplicity . . . so light would travel across the room in 1 x 10^(-8) seconds . . . about 1/6th of a single clock cycle

and your timer isn't running nearly as fast as 16MHz!

to get any meaningful measurement, the room would have to be at least 1000ft x 1000ft


edit:

Remember that toy that I wanted to mod that had RF ranging on it?

that toy works on intensity, not distance measurement . . . you'd have to use three devices emitting entirely different frequencies for the triangulation to work . . . so in this particular case, triangulation will work on the condition the intensity gradient is sufficient enough . . .

[airman00]

oh I understand now

I meant intensity . But can't you determine your relative position based on the intensity levels of multiple transmitters?

[Admin]

But can't you determine your relative position based on the intensity levels of multiple transmitters?

well, you'll run into issues of poor intensity gradients . . .

for example, with an 8 bit ADC, and a sensor far from a light source, a single value could represent several meters

so error exponentially increases as the sensor distance from the emitter source increases linearly

but if you did the timing method, error decreases as distance is increased

[pomprocker]

What about starting your robot off at the charging station, and using shaft encoders to count your clicks on each wheel as you go. Then through formulation, you can get back to start when it needs to charge.

[airman00]

But admin can't you tune the RF device to have a single value represent inches or feet , etc. I like RF because you don't have to face the charging station and you can things in between the station and you.

[Admin]

pomprocker, you'll have encoder drift . . . error just builds up all the time

But admin can't you tune the RF device to have a single value represent inches or feet , etc.

nope

example:
ADC distance_away
0     1 inches
1     2 inches
2     5 inches
3     9 inches
4     13 inches
5     20 inches

So as you can see, the resolution worsens as distance increases. And objects in the way will reduce intensity too.

Another example below . . . the change in intensity per distance changes as you get closer to the source:

[Tomomastchi-project]

My original plan was to use two IR dependent sensors and one source. The difference between the intensety indicates the direction. But then I realised that IR is in the spectrum of sunlight. I gues it isn't possible to work with the intesity?

[Admin]

But then I realised that IR is in the spectrum of sunlight. I gues it isn't possible to work with the intesity?

All light is in the spectrum of sunlight

But not a problem, you'd have to use modulation. Search around for how to modulate an IR signal (or better yet, a sensor you can buy that already does it for you).

[gamefreak]

Not to throw a monkey wrench, a monkey with a wrench, or a wrench that looks like a monkey into the works, but is the use of multiple sensors not possible?

By combining things such as wheel encoders and IR you could create a system that is fairly accurate, get the robot within the vicinity of the station with encoders and then pinpoint the exact location using IR sensors on the bot and a beacon on the base.

It is impossible to avoid error, computers only have so much space reserved for data and anything extra is simply tossed into the ether. So having redundancy or even entirely different yet coexistent systems provides for a reliable way to return to your original location.

To err is human, and to something is something else.
Since robots are made from humans naturally they must Err as well.

Also, how does localization/GPS work? Dont they use RF? And GPS can be extremely accurate....

[Conscripted]

Also, how does localization/GPS work? Dont they use RF? And GPS can be extremely accurate....

If you are asking how gps works. As I understand it... (very oversimplified)

GPS uses a series of 24 satelites for postion determination. Civilian systems use the frequency of 1575.42 MHz. The reciever uses the almanac and Ephemeris data to know what time it is(within a few nanoseconds) and where each of the satelites are. The signal includes an ID number for each satelite. Once you have a lock on at least 3 statelites you can triangulate your position. The accuracy is very good on a global scale. Not great on the smaller scale that most deal with for robotics. The accuracy of gps is typically within a few meters.