New Website

Because we (my wife and I) use this website for different things, I am migrating my robot projects to the site www.backyardrobots.com . That site can be all robots. I will update this site for a while but I do plan to move the links over to the other site.

I got twelve servos for a new quadruped robot. To try different servos, I got HS-5485s and HS-5685s. These are digital servos, faster and more powerful than the standard size servos I have used in the past. I came into a frame for a quad in some unusual circumstances. I am hoping that during the Christmas break I will be able to build the robot. That is what I am looking forward to for the holidays.

The Peanut tin of terror battled Otto at Combots V this year. The event was great. We actually had pretty big crowds for the small weight classes. We watched a whole bunch of exciting robot fights and I even brought a small remote controlled fighting robot. As for the Peanut, well you can see what happened in this video.

The Combots V robot competition is coming this October. The Peanut Tin of Terror was created to be an entry in this competition last year, but a personal injury prevented me from attending. At Robogames Peanut Tin did surprisingly well. Now the Peanut Tin is a veteran of autonomous robot combat and ready to improve his performance at this year's Combots. So far Peanut Tin is the only entry in the 3 lb division, but I think he will have some competition soon. I will post updates on the improvements I am making to the Peanut Tin to hopefully allow it to be a better combat robot. I plan to improve software and hardware. The hardware improvements have already begun with the installation of the ColorPAL sensor. But there is more to come.

What is this on the bottom of the Peanut Tin of Terror?

Well, Parallax had a sale earlier and I picked up three ColorPALs. The ColorPAL is a little sensor that can detect different colors on a surface. It has three LEDs that are red, blue, and green. It uses a little microprocessor connected to the LEDs and a light meter.

I created a Spin object to use it with a Propeller chip. I am looking forward to doing some new things with it. In the meantime, I think it will be able to keep the peanut tin from accidentally driving out of the arena in Robogames 2011. This was a big problem for many of my opponents. I have plans to continue to improve the Peanut Tin of Terror. I expect the competition to be much better next year, so hopefully it can keep up.

I finally have added a guide to servos for robots. As the world's greatest expert on robots I decided it was time to weigh in with an authoritative guide to which servos you should use in your robot.

I finally finished updating the Robogames 2010 page with videos from the autonomous combat. See how Peanut Tin came in second. I will continue to update this page soon with movies from the walker challenge.

Peanut is running and pretty much ready to go to Robogames. I did a weigh in with the postal scale at work. It weighed 2 lbs 11 ounces, so I can add some weight to the spinning blade.

Stery is 10 1/2 lbs. Pretty heavy for a walker. I shortened the legs a little to make it flex less.

To get Stery ready for the walker challenge I have added a mount for the compass and padding for the feet. The compass really seems to work much more accurate than the internal compass I put into Stery before Robogames 2009. I cannot tell if this is because the compass is externally mounted and away from the magnetic fields of the compass or if it is because I am using a better type of compass.

Last year I put a Hitachi HM55B on a Spin Studio card. I still use this on Fluffy, but it is not particularly accurate. On Stery it was even worse, presumably because the wiring and the motors were all around it. For whatever reason it got very poor readings.

This year I put an HMC6352 on a little stand above Stery. I wanted to get it out of the case so that it was away from the magnetic fields of the motors as much as possible. But the HMC also has an auto-calibrate feature that seems to work really well. The HM55B needs to be calibrated manually. It could be that HMC is a better performing compass because it auto-calibrates accurately. I want to take some time to look into this later.

Another problem seems to be that now that the feet bend and are smooth, there is a tendency for them to slip. The foot can flex and both the front and the back foot can slide backward. To give the foot more grip I put furniture foot pads on them, which seems to help.

I still seem to be having a lot of trouble getting over obstacles. Worse, the flat feet seem to tend to tuck themselves underneath obstacles (like boards) getting themselves stuck. It really may have worked better to just have broomstick feet.

I am trying to get Sterylite ready for the Walker challenge. I added new brackets to hold the servo wires.

More proof that I am a terrible fabricator, if any were needed.

So far I have had no luck in getting the robot to clear 5 cm. It walks pretty well, but it is not clear that it has better ground clearance with articulated feet then it had with broomstick feet. We will see.

I am getting exited for Robogames 2010. This year I am entering The Peanut Tin of Terror in the 3lbs autonomous combat competition. That looks to be an active competition with six entrants so far. I am entering Sterylite 6000 into the Walker Challenge and once again I am waiting until the registration deadline to see who I am going up against.

Every year the walker challenge has been won by Ziggy, a terrific hexapod built by Nick Donaldson. Last year his Mindstorms robot Flik won second. Nick Donaldson is a terrific robot builder and a very nice guy. Ziggy in particular is an amazing robot. Currently neither robot is registered, but last year they registered near the end. It would be too bad if they didn't compete.

A new robot has registered for the Walker Challenge, Dexter. Dexter comes from Vanadium Labs, a team that also competes in the very cool, very elite MechWars challenge. Although I don't know anything about Dexter, I can definitely say that they make very cool robots. I think one of the builders is the CTO of Trossen robotics.

I think Sterylite can improve its performance over last year, but can it beat anybody? I think if everything works well it will outperform Flik. I think I can keep up with Ziggy, but would have to get lucky to get over the stairs at the end. So I think Ziggy is likely to win vs. Sterylite. But what about Dexter? I don't know. You can bet I will be checking their blog a lot. I think the best thing about the uncertainty is that it inspires me to improve Sterylite.

I still have to make an external mount for the compass on Sterylite. I have gotten a compass to work pretty well on Fluffy. There is still some programming I want to do to improve the averaging of the compass vector. But a compass chip has never worked well in the body of Sterylite, probably because it has a lot more power and a lot wires creating magnetic fields. I would also like to build sleeves to hold the servo cables on the shoulders.

Peanut Tin will face a lot of veteran robots in its autonomous combat class. I don't think its chances are very good, but it will be fun to try a new event.

To get better at keeping my robots in a straight line, I have been working on getting Fluffy to navigate via its internal compass. I tried using a gyro, but it was too unreliable on a walking platform. A compass also gets a lot of funky readings that you have to average. Here is a demo of Fluffy going South, West, North, and East in a rough rectangle.

Obviously there a number of problems. One problem is that the compass isn't really calibrated. I think that is the main reason for the crookedness on the second turn. Fixing calibration actually isn't that big of a deal.

Another problem is that I am storing the compass heading as a scalar (degrees) which does not work well for averaging erratic readings. A compass heading of : 355, 5, 358, 2 is possible when going North. But that averages to a heading 180, or due South. Averaging it as a vector takes care of that problem. It also has some advantages if you want to use the compass vector to make minor course corrections. But if you average vectors, sometimes you have to normalize them to work with them. Then you are doing pretty much floating point math on a microcontroller but there is currently one of the eight processors set aside for floating point.