2010 Robogames

Before the Event

Here is Peanut Tin ready to go before the competition:

You can read more about Peanut Tin here. The strategy of Peanut Tin was to fire sonar over the opposing robots to detect walls and use downward facing motion sensor to detect opponents. It started by turning and then going forward until it found a wall. Once it got to a wall it pivoted and then paused. It checked for two things: 1) if the way forward was clear 2) if something in front of it seemed to be moving. If condition one was true the Peanut would charge with its blade spinning. If condition two were two it would also charge, even if the way ahead appeared blocked.

The point of this strategy was to mostly keep Peanut close to walls and preserve power. I also learned from experimentation that while the motion sensor was very unreliable at detecting the motion of an opponent, it was pretty good at detecting when the Peanut was getting moved by somebody else. In other words, if the robot expected to be still and the motion detector went off, it could be that the robot was getting pushed or hit. This is actually a useful piece of information and a good reason to scoot forward.

First day of competition - April 23rd

I didn't know what day Peanut Tin would compete so I needed to go down on Friday to San Mateo to see when autonomous combat would be. This was my first visit to San Mateo event center. They had a firetruck and ambulance outside. Nothing says fun like an ambulance-- although I noticed they were somewhere else for Saturday and Sunday. You could see EMTs on the premise at the large combat arena.

The first day of the competition was very informal, with people standing around setting up competitions and testing robots. I headed for the small robot combat arena where I met a guy, Kevin, who was organizing the event. I checked to make sure I would be OK without a radio trigger for my robot. He said it would be fine if I had a five second delay. I asked him if the arena I saw would be the place of competition. He said there would be an insert.

He began diagramming the insert to the arena. It would be smaller, with low walls. There would be a white stripe around the edges, and red markers at the gap in the walls. The gap in the walls would be a place where your robot could push out another robot. I wasn't aware of these specifications before.

The philosophy of Dave Caulkins is to have Robogames be an event like the Olympics where different groups take charge of different events, like sanctioning bodies. Robogames tries to bring in diverse groups. The organization levels of the events vary a lot. Some competitors like the fire fighting and Mech Warfare seem super organized. Some others seem more loose.

If you want to know how your event is going to go it often is best to contact the person who is running it. The trick is figuring out who is running it. I tried to make a point of meeting my competitors this year because they could be the people we arrange the event with next year. Beetleweight autonomous combat now seems like an event that I understand pretty well, but I understand mostly from having done it. It did not seem like anyone sent the arena specifications to the competitors. John Frizell-- a more experience competitor-- knew all about the arena, so maybe there was a communication I missed. Anyway, not having the arena dimension right was the biggest problem for Peanut tin.

Second Day of Competition April 24

My nephew Krishna competed this year in the Lego bowling challenge. He created a Lego robot to throw a bowling ball. There were a lot of robots in this competition. The winning robot used a rubber band to shoot the ball. It didn't break the rules, but there really didn't seem much that was robotic about it.

Here is Krishna's robot:

His friend from school, Sumner, also created a robot. It came in third place and worked well.

Round 1 : Otto I vs The Peanut Tin of Terror

My first match was against last year's winner: Otto 1. Because this was my first match, and my friends were watching the Jr. Division Lego, I did not get a very good video of it. Peanut turned around immediately at the start and turned into the corner. The sensors were above the wall and failed to detect it. There was no time out condition for the going forward state, other then detecting a wall. So for the whole match I was just driving into the corner. But Otto drove himself out of the arena. So I won. It was a cheesy win, but Otto got his revenge later.

My first match was against last year's winner: Otto 1. Because this was my first match, and my friends were watching the Jr. Division Lego, I did not get a very good video of it. Peanut turned around immediately at the start and turned into the corner. The sensors were above the wall and failed to detect it. There was no time out condition for the going forward state, other then detecting a wall. So for the whole match I was just driving into the corner. But Otto drove himself out of the arena. So I won. It was a cheesy win, but Otto got his revenge later.

As it turned out, there were no red markers were at the exits of the arena. As far as I could tell, no robot actually used the white lines to detect the edges of the arena. The white lines were actually kind of silvery. The exits of the arena did play a crucial role, you really needed to avoid them.

In the next match you can see me turning on the robot with a five second timer. Having a radio starter is actually a disadvantage in some ways because it is another point of failure. Your radio could be a hundred percent accurate in tests but fail in the environment of Robogames, where you have the shielding of the combat arena between you and your robot, as well as many stray electromagnetic fields. You may need to have one because they don't like you to go in and separate robots without turning them off. Next time I will try to have a radio.

Round 1: Otto II vs Mini Toro+

Here is the other Round I match between Mini Toro+ and Otto. Mini Toro was definitely the toughest bot but it had sensor problems. It seemed to just charge into things. It bumped into Otto II later, but just got stuck pinning it. Anyway, it was a decision for Otto II.

Otto II had this high sensor that was supposed to detect the edges of the arena by peeking over them. Otto II had an active weapon, but it was not really very powerful.

Round 2: Peanut Tin of Terror vs. Otto II

In round two I went against Otto II. In round one I won because my sensors were two high for the wall. For round two I angled the sensors down so that they were pointing down more. This actually worked worse in some ways because they shook into my blade. In this video you will see what was probably the greatest damage done by any robot to any robot. Who did the damage? The Peanut Tin of Terror. Who was damaged? Peanut Tin of Terror!

But once again I got lucky. After staying in control in the first match, Otto II went out of the ring and the peanut picked up another lucky win.

Round 3: Peanut Tin of Terror vs. Otto

In Peanut's last round it faced Otto I again. I can't explain the brackets, but it would have been crazy to win without doing much except getting stuck. And really if I could have I would have wanted to fight every robot twice. You work so hard making a robot you just want to see it go as much as possible.

Since the sonar was a problem in both of my preceding matches, I adjusted it again. I wanted to get the sonar pointing down, so it could catch the wall. But bending it down caused it to get caught in it's own blade. So this time I bent the sensor stalk back so that the sensors were at a lower height but not bent over the blade. The problem with this was that it defeated the point of the stalk, which was to give the sensors a clear view past the blade.

Peanut confused its own blade with a sensor. I put a time out to keep from gong straight forever. But there was no time out in the code to turn, it just kept turning until it thought it was clear of a wall, which was never. Peanut was counted out. So peanut lost in the finals, but ended up in second place.

One thing I learned is that you have to program for sensor failure. My sonar failed because it was poorly positioned. But if it had worked better there would be a good chance that they would have been knocked out in combat. Peanut pulled the wires off of its own motion sensor, but it could have easily been the sonar wires that got pulled out. It would have been easy to put in code that detected when a sensor got the same reading for a certain period-- like 360 degree turn. At that point you could just ignore the sensor and get yourself out of the deadlock. Its the kind of thing I know to do now. The Peanut Tin was not strongr or faster then the other robots, but I think it had a much more powerful microcontroller. I should have done more to take advantage of that.

I put some other videos of autonomous combat on my youtube channel. If you are interested in seeing it, you should check out youtube.com/georgeecollins

Third Day of Competition April 25 - The Walker Challenge

This year I entered the Walker Challenge again, which was the only event I did last year. Sterylite6000 had new feet added since last time. It also had a compass, but the compass was too unreliable for me to trust using it. The shaking is very bad, and you need to recalibrate often. Calibrating a compass in the environment of a race at a robogames did not seem like a good idea.

Nick Donaldson was running the race agian this year. He said he was a little nervous when he saw that Michael Ferguson of Vanadium Labs was entering. He did two things. First, he entered a robot in the lego tube push event, which he thought he could win. He wanted to make sure he took home a win. He made a very clever lego robot and as he expected, he won. The other thing he did was make Ziggy walk even faster.

Michael Ferguson was not able to compete. He showed me an excellent frame for an AX-12 hexapod that he has been working on. Apparently he did not have time to finish up programming and getting his robot together. I ready later that he had a lot of problems with his quad Issy in the MechWarfare event. MechWarfare is really a premiere event. It is one of the most difficult challenges because you have to build a walking robot that can carry a heavy payload. You have to put a camera on it, weapon hit sensors, and a gun. You have to be able to control your robot reliably in an environment with a lot of radio frequency noise. Michael's robot eventually had its camera fail, which means you can't compete.

As usual, Ziggy did great and won. It walked better this year than last year. It really is an excellent robot.

Sterylite6000 also did much better. You can see it here going much straighter than before, though still not straight enough. I guess what I am proud of is how fast it is, and what good ground clearance it has. The obstacles of the course aren't really a problem.

I don't know what I will do for the Walker Challenge next year. I do not think I can make Sterylite6000 much better. It's biggest problem is that it is too big and has a hard time keeping straight over obstacles. I asked Nick about having a wall on one side of the course. If there was something a Ping sensor could measure against for navigation, it would help. A smaller robot would also have an easier time staying on the track. But the only other person who does the event is Nick, who organizes it. It may make more sense to put the effort into an event with more participation like MechWarfare.