Mindboards

First, a little about me: I started writing code 50 years ago on an IBM 650 (look it up). I worked for IBM as a programmer/tester/performance analyst from 1965 to 1995, when I retired. Then I did contract work, mostly in MS Access. Now I'm fully retired and my wife and I enjoy traveling in our fifth-wheel trailer. I started "playing" with NXT last summer. I've built the robots using other people's instructions, then "enhanced" their code. After using just NXT-G for a few months, I started programming the PC with http://www.norgesgade14.dk/bluetoothlibrary.php, both console and, using QT, windowed interfaces. Now that I've discovered NXC, I'm using it to program the NXT.

I'm currently working on "Odin - a robot for odometry", found at http://www.philohome.com/odin/odin.htm. I found one assembly error, in the 3x5 L-beams that attach the turntable to the motor that drives it. (His instructions show it one way, the video shows it reversed, which is correct.) The NXT-G code he provides for the script driven code doesn't work for me because he wrote it in NXT-G v1 (integer) and I have v2 (floating) and 2 math blocks (sqrt and atan2) that he used don't seem to be available any more. So, I'm rewriting it in NXC. (That's my excuse, anyway.)

With his version, it is not precise in the distances nor angles, so I am trying to understand John's routine for ramp up/down and how to get my measurements precise.

What sensors will you plan to use except odometry?
I'm already using a navigation robot, based on odometry, compass, and gyro. An additional accelerometer is planned (some implementations will have to wait until my BT based chess robot is finished).

The navigation robot made to "map and explore" the environment, and once having explored it it may use an A-star for finding the shortest path to a target.
The map is based upon squares of 40x40cm² each up to a total size of about 40x40 of those squares (total size 16m x 16 m =~ 300m²).
(40cmx40cm is about the sice of my robot, so he can stand, and drive, and turn on each single sqr.)
A bug2 algorithm for path finding in yet unexplored areas is plannend to be implemented.
For outdoors each sqaure might be enlarged to 1m x 1m and so the total map will be up to 40mx40m (1600 m²) and can be combined with a GPS sensor.

The Odin robot doesn't use any sensors, except a touch sensor to initialize the "foot" it uses to do a precise pivot/turn. You feed it a script with x,y coordinates and it navigates using only dead reckoning (http://www.tfd.com/dead+reckoning).

Thanks for the interesting information!

My Grandma (who passed away in January) used to be a programmer for an IBM computer (rented from IBM by the local news paper company). I don't remember the specific model she programmed, but it was vacuum tube based, and she worked with it from about 1970 - 1985.

mcsummation wrote:The Odin robot doesn't use any sensors, except a touch sensor to initialize the "foot" it uses to do a precise pivot/turn. You feed it a script with x,y coordinates and it navigates using only dead reckoning (http://www.tfd.com/dead+reckoning).

but to my observations odometry on it's own is far too uncertain to determine a robot's position...

Mine is doing fairly well. The primary error I have now is in my initial angular positioning. By the time it has wandered off through the house about 4 meters, it may be off by about 10 cm to the left or right. The distance is very close. I'm testing it on a vinyl floor with one foot "tiles".

well, ok, but now let it go for 10 times exactly round a square of 1m x 1m and then let it go it backwards the same route...

I have been working for a couple of weeks on Odin and finally got it so it will accurately navigate through the house. The secret to this type of robot is accurately calibrating the distance and angles navigated. The distance is easy - measure off 10 feet and see how many degrees the drive motor turns to go that distance. As to the angle, Philo had provided the plans for a big L and a Reset AZimuth (RAZ) routine to get it aligned very well. Two trips through the RAZ L would only get about a 1 degree correction the second time.

I am now working on the fourth incarnation of HiTechnic's RC Kart. Although they control theirs with a Lego IR controller, I have been using Bluetooth. For this incarnation, I am using Philo's 2-axis joystick. I am also using some new Bluetooth knowledge gained here on Mindboards to make the communication robust and high speed. (More on this later.)

but honestly, Odin is no real driving robot, is it?
To my opinion steering has to be managed by the wheels itself or a steering wheel, and not by a lifting and turning punch or plunger (no idea how you may call this).