Wossname... and whoever else

[dsheller]

What's up with the grumpy snail? I'm working on my own first robot project and was curious how the grumpy snail is going. Also, did he ever talk over Wi-Fi?

[kregg]

You mean the Linux Robot project wossname was writing about in his blog? I never heard much of that.

[Shaggy Hiker]

My robot talks nicely over bluetooth, but my sonar sensors turned out to produce pure crap beyond about 60cm, which pretty much made them useless. Therefore, the robot sits while I try to figure out a new kind of sensor to work with that gives me some kind of meaningful distance reading beyond 1m range.

[wossname]

Finished it, sold it to NASA. Now I'm rich.

...no wait, that was in a cartoon I drew.



Funny you should mention this actually, I've spent the last week or so rewriting the core daemon system for GS. Since quite a long time has passed since I last blogged, I have learned much about linux system programming. So it will be a great deal more stable and fault-tolerant than my original design. Basically it will consist of a single binary that forks into probably 6 completely separate daemons that all have an important task to do (motor control, environmental sensor processing, AI, database handling etc...). Then the parent process has the job of coordinating and controlling the other forks in response to user / envorinmental input.

My main problem is my profound lack of electronics/hardware fabrication skills. I can do simple circuits with really simple integrated circuits but things like stepper motor controllers and safe power supplies are something I will need to overcome if this thing is ever going to become mobile under its own power.

If anyone has any interest in this project then I'll happily resume the blogging.

[Torc]

Finished it, sold it to NASA. Now I'm rich.

Only until the revenuers take most of it in taxes.

[wossname]

No, Grumpy Snail is a registered charity. It's registered in Switzerland under the name of a man that died in WW1. It's funds are all tied up in oil interests overseas and we don't have a fax machine.

[Torc]

In that case you will probably have the US IRS on your tail.

[kregg]

In that case you will probably have the US IRS on your tail.

[dsheller]

My robot talks nicely over bluetooth, but my sonar sensors turned out to produce pure crap beyond about 60cm, which pretty much made them useless. Therefore, the robot sits while I try to figure out a new kind of sensor to work with that gives me some kind of meaningful distance reading beyond 1m range.

I'm using the ultrasonic sensor called PING))) from parallax, it works really well up to ~3 meters... I'm using it on an ATmega168 and capturing it's echo. The kicker with this one is it has only 3 pins, Vcc/Gnd/Sig... so essentially you send a 5 uS pulse through Sig and wait 750 uS (Echo Holdoff). After the signal you'll get a PWM signal which is directly correlated to how far the object is... I get good readings assuming I am using a good value for the speed of sound =)

If you want to ditch ultrasonic, Sharp makes some very good IR detectors that have decent range... I'm using one that's ranged max up to 1.5 m... I don't remember the minimum range.

EDIT: Forgot to ask you what type of bluetooh module you're using... I'm considering going with something called Zigbee myself. The only problem with that is I don't know if I could use it to talk to a laptop... for robot to robot communication it looks great though.

[dsheller]

Finished it, sold it to NASA. Now I'm rich.

...no wait, that was in a cartoon I drew.



Funny you should mention this actually, I've spent the last week or so rewriting the core daemon system for GS. Since quite a long time has passed since I last blogged, I have learned much about linux system programming. So it will be a great deal more stable and fault-tolerant than my original design. Basically it will consist of a single binary that forks into probably 6 completely separate daemons that all have an important task to do (motor control, environmental sensor processing, AI, database handling etc...). Then the parent process has the job of coordinating and controlling the other forks in response to user / envorinmental input.

My main problem is my profound lack of electronics/hardware fabrication skills. I can do simple circuits with really simple integrated circuits but things like stepper motor controllers and safe power supplies are something I will need to overcome if this thing is ever going to become mobile under its own power.

If anyone has any interest in this project then I'll happily resume the blogging.

What kind of hardware were you using again? Obviously something a little more beefy than a microcontroller. As for the motro controllers you can ditch those and go with servos, which can be modified to do continuous rotation, and then all you have to do is send them a PWM signal to dictate a speed... but I don't know the kind of weight requirements you have... and you might have motors already. If you do already have motors, a good way to control those is with something called an H-Bridge, which can be made out of a handful of passive components and transistors (probably MOSFETS if you're using beefy motors). Power supply shouldn't be too bad, if you only need one voltage all you need is a regulator and some filtering capacitors... and that SHOULD be fairly decent for the motors... another thing is to seperate the motors from the processor as a voltage drop will likely ensue under a few heavy current conditions (changing direction is one such condition)... but once again a large capacitance capacitor could help to alleviate such voltage drops if you place it between Vcc and Gnd rails.

[kregg]

If anyone has any interest in this project then I'll happily resume the blogging.

Please. Also I won't spam it, despite what I may have said in the past.

[BillGeek]

Wossy, where is this blog located?

[dsheller]

http://lcav.blogspot.com/

L2 Google, noobs. =)

[kregg]

L2?

[Shaggy Hiker]

I'm using the ultrasonic sensor called PING))) from parallax, it works really well up to ~3 meters... I'm using it on an ATmega168 and capturing it's echo. The kicker with this one is it has only 3 pins, Vcc/Gnd/Sig... so essentially you send a 5 uS pulse through Sig and wait 750 uS (Echo Holdoff). After the signal you'll get a PWM signal which is directly correlated to how far the object is... I get good readings assuming I am using a good value for the speed of sound =)

That's good to know. I'm using the SRF08 sensors. They are supposed to return the first 17 echoes, but what I have found in testing is that they appear accurate for less than 1 meter, and all the echoes beyond that remain basically the same, regardless of what's out there. The one real advantage to them is that they work off the I2C bus, which means that I am using only 1 IO port for five sensors. The other good thing about them is that power consumption is minimal. Those Sharp IR sensors draw an insane amount of power, even in standby. I was trying to use a couple FETs to toggle the IR sensor on and off, but it made the sensors a bit quirky.

EDIT: Forgot to ask you what type of bluetooh module you're using... I'm considering going with something called Zigbee myself. The only problem with that is I don't know if I could use it to talk to a laptop... for robot to robot communication it looks great though.

I bought the bluetooth chip from SparkFun, along with a bluetooth dongle. I forget the specs on it (BlueSoleil, I think), but last time I looked, they had changed their lineup somewhat. The chip is on a small IC board with a simple pinout. You can get it with and without an external antenna.

[dsheller]

That's good to know. I'm using the SRF08 sensors. They are supposed to return the first 17 echoes, but what I have found in testing is that they appear accurate for less than 1 meter, and all the echoes beyond that remain basically the same, regardless of what's out there. The one real advantage to them is that they work off the I2C bus, which means that I am using only 1 IO port for five sensors. The other good thing about them is that power consumption is minimal.

Could you be forgetting to account for an overflow on the timer you're using? Or do the SRF08 sensors send you some digital data... I'd look it up but i'm side tracked with some other stuff currently.

[Shaggy Hiker]

They send digital data. 17 WORDS showing distance in cm for the first 17 echoes. The problem is that if I use my straight up sensor in the middle of the room, I sort of get an echo in the vicinity of where the ceiling should be, but half the time I also get one at about half the distance. When I use the forward looking sensors, I soon found that I was getting echoes at the same distances wherever I moved the bot. What that means, I can't say.

[wossname]

I'm going to start a thread for Grumpy Snail (LCAV to give it it's official name) in the project forum so it doesn't get spammed into oblivion.

et voila http://www.vbforums.com/showthread.php?t=528273

[wossname]

Shaggy, What hardware/OS (if any) and programming languages are you using for your robot? Just out of interest.

I'm using a Pentium 3 (1Ghz), smallish desktop ATX mobo (nothing fancy), CentOS linux 5.1 and I'll be coding virtually everything in C/C++.

I haven't decided what the final "task" is that grumpy snail will perform, but it will likely be something like terrain surveying (contours, shallow ground metal detection) combined with simple pathfinding and collision avoidance. Oh and don't forget driving it around manually with a wireless xbox controller

[dsheller]

How are you demodulating the signal the controller is sending you?

[kregg]

Shaggy, What hardware/OS (if any) and programming languages are you using for your robot? Just out of interest.

I'm using a Pentium 3 (1Ghz), smallish desktop ATX mobo (nothing fancy), CentOS linux 5.1 and I'll be coding virtually everything in C/C++.

I haven't decided what the final "task" is that grumpy snail will perform, but it will likely be something like terrain surveying (contours, shallow ground metal detection) combined with simple pathfinding and collision avoidance. Oh and don't forget driving it around manually with a wireless xbox controller

A linux based robot and an Xbox controller does not go hand to hand.

Think rationally!

[Shaggy Hiker]

Shaggy, What hardware/OS (if any) and programming languages are you using for your robot? Just out of interest.

I'm using a Pentium 3 (1Ghz), smallish desktop ATX mobo (nothing fancy), CentOS linux 5.1 and I'll be coding virtually everything in C/C++.

I haven't decided what the final "task" is that grumpy snail will perform, but it will likely be something like terrain surveying (contours, shallow ground metal detection) combined with simple pathfinding and collision avoidance. Oh and don't forget driving it around manually with a wireless xbox controller

The bot is using an OOPic board, which is a microcontroller with a very-pseudo OO language. Actually, the only thing actually OO about it is the two Os in the name, but it is based on the PIC microcontroller and the OO language is a bunch of objects that wrap functionality.

Therefore, there is no real OS on the board, and it's all pretty low level. Memory and program space is very limitted, but that's not an issue. I hooked a bluetooth chip to it from SparkFun.com, and the bot is nothing but a sensor platform with a motor controller. The program loop responds to sensor read requests, and movement commands (duration and speed for each track on a tank tread chassis). The brain will be spread across three or four computers via ethernet and WiFi. At least, that's the plan, but the fact that the sensor data is garbage and that I am pouring all my money into my back yard has combined to put the project on hold till fall.

At this point, the communication with the bot, communication in the peer-to-peer network on the various computers, and motor control all work. The touch whiskers are working, too. However, since the sensors are only effective at sub-meter range, the bot is effectively so myopic that my designs went right out the window.

I'll be off on a long walk-about for the next month, during which I'll consider whether I can just live with the myopic bot, but I feel that I need at least one medium range, narrow beam, detector.

@dsheller: The output from the sensor is a series of up to 17 two byte WORDS holding the distance in cm to the object. There's no processing taking place of any sort on this information. I feel that my problem has to do with side lobe echoes, which is a known issue with this particular sensor, but I didn't find that out until after I had bought them. They would be more effective in a large space and set well above the ground. Interestingly, they do a fair job at detecting items on the floor at a distance of about half a meter, which is of some benefit, but they aren't the medium range sensors I was expecting.

[dsheller]

Thanks Shaggy. I think I'll probably get the bluetooh module as well for mobile communication to my machine. I was considering the zigbee module as it is a little cheaper.. not sure if it would be more useful.

[wossname]

A linux based robot and an Xbox controller does not go hand to hand.

Think rationally!

It does when you realise that xbox controllers are just plain USB devices. Just plug the controller's receiver into the USB on the PC and off you go. This is nothing new, people have been using xbox controllers on PCs for years now. If its an old XBOX 1 wireless controller you have to cut the cable up a bit to remove the yellow wire but otherwise it's fine.

You don't honestly think I'd cut up my precious 360 controller do you?

[wossname]

How are you demodulating the signal the controller is sending you?

I'm using SDL to read the gamepad device. I'll be using SDL for the vision system when i finally work out how to get and tune a raw webcam signal.