This weekend while I was working on The Peanut Tin of Terror Spencer, my son, mentioned that he learned what sandpaper was made out in his woodworking class. “Do you think we could make our own sandpaper Dad?”

I want to encourage him to try things and to teach him that it is alright to make mistakes-- something that people who visit this website know I am uniquely qualified to do. So we took some thick paper from a junk mailing, some hot glue, and some sand from the sandbox. We ended up with this:

My wife thought it was so funny she took a picture.

Anyway, I have not been updating the main page but I have been working on my robot for Combots. You can see the details here.

I have started to build a small autonomous combat robot for combots. I will keep a page for the 2009 Combots competion, and you can read about the development here.

The Combot Cup IV is December 19-20, 2009. I realized that this may be one of those times when it is easy for me to go over to the Bay Area for a weekend. Spencer (my older son) loved the Robogames and he really wants me to do combat robots. Combat robots isn't really my bag, but maybe I could do an autonomous fighting robot? I am considering going. Most importantly I have gotten the go ahead from the wife. But I have no robot to enter as of yet.

Registration has begun for RoboGames 2010. I will be there. I have not decided in what events, although I am sure I will do the walker challenge. I will update this site as I prepare.

I have been having a lot of problems getting an accurate reading from an internal compass in Sterylite 6000. One alternative would be to create an external mount for the compass, taking it away from the electromagnetic interference of the robot. Another alternative I am exploring is using a small RC helicopter gyro to determine direction.

I got an inexpensive GWS PGS-03 Gyro from Tower Hobbies. I had to experiment a lot to discover how it works. If anyone reading this realizes I have gotten something wrong, email georgeecollins at a yahoo mail account to correct me. The way the gyro seems to work is that it takes a servo pulse as input and then lengthens or shortens that pulse based on the angular velocity of the servo in one dimension. Tracking the angular velocity from moment to moment can give you a pretty good approximation of the direction gyro is pointing, with some important exceptions noted below.

I created a little balsa wood mount for the gyro so I could sit it on top of a servo. I put the gyro on a servo so that I could turn it by consistent amounts and for consistent periods of time, to test and tune tune the measurement. The servo and the gyro were connected to Fluffy the robot, who output the results to a television.

This is a test of the GWS PGS-03 Gyroscope. I am using the propeller chip from my quadruped robot to drive a servo moving the gyro back and forth. The same chip measures the variation in the pulse width coming from the gyro as it turns. The prop uses a TV signal to display an arrow showing where it thinks the servo is pointing based on the gyro.

The TV display has a lot of interference and may be hard to read. You can see that the gyro is pretty reliable moment to moment, but it has a tendency to drift over time.

One of the challenges of the Walker Challenge is to get your walker to travel in a straight line autonomously as you cross uneven terrain. Initially I did not think this would be a big problem since Fluffy uses a HB55 compass chip that can tell in what direction the robot is traveling. I had already taken the chip and put it into a custom card for a Spin Studio board. That meant that as soon as I had a prop board working in Sterylite, I could swap in the compass board.

Unfortunately what worked reasonably in one robot completely failed in the other, probably due to issues of power and scale. Sterylite uses much more power than Fluffy, with larger motors that are relatively closer to the electronics. I did not realize this would be such a big problem. At first I just stuck in the board and wrote some code that caused the robot to navigate to a particular compass heading. When this did not work I assumed their was a bug in my code. After much testing of the code I discovered how badly the compass was performing in the new robot.

A simple test for the compass is to plug in a TV output board and run a program that displays a compass needle. Then you can spin the roboit around and watch the compass turn at the same time. When you do this with Fluffy, you can see when Fluffy faces North the compass points South because it is mounted backward, an easy problem to fix in software. When you spin Fluffy in a circle the compass turns in a circle. A quick test with the same set up inside of Sterylite shows an obvious problem. You can spin Sterylie in a full circle the compass needle only moves about 115º. A compass that cannot travel a full circle has an obvious problem.

I have used the HB55 on Fluffy, and while I am not sure exactly how accurate it is, I know that it at least can point to all compass headings, and changes reading to reflect the attitude of the robot. To figure out what might be best for Sterylite, I am developing a more accurate logging system to keep track of how accurately the robot can track its direction. I am also going to try an inexpensive RC helicopter gyro, to see if I can use that as a more accurate way to keep straight.

August 8, 2009 - Sterylite with Ankles
I have gone back to ten motors on Sterylite. Instead of mounting the four additional servos high on the legs-- as I did previously-- I put them at the feet. I also used feet from Lynxmotion, stealing the whole layout from their bipeds. I really like Lynxmotion products, although they are not designed for a robot as heavy as Sterylite. I have had enough problems with my own designs. I am trying to take more proven components.

Here is the foot:

Here is the leg and foot together:

I have put up videos of the ankles in action. You can see that the robot walks more gracefully. I also slowed down the gait and decreased the range of the servos. As I test the robot I am running it in a more conservative mode to reduce the chance of damage.
The robot actually damages itself a lot. Before I did these videos the robot bent the "C" servo bracket on one of the arms while I was testing walking. So I decided to slow it down, at least until I race it again.
You can see the videos here.