Building a minisumo robot

I will now post some of the important parts for building this robot.

First and probably most important are the motors. These are the most expensive and hard to find parts for any robot and in sumo especially they play a vital role. I am planning on using motors I got from eBay. These are used motors from some industrial machine and they were made by Think Engineering. The best motors are usually made by either Maxon or Faulhaber and these can cost a fortune event when used. The best place to hunt for motors for these robots is usually eBay as you can get some 10 motors for the price of one new motor. The motors I am using are actually 24 volt motors and that is not very good for a robot because the batteries we usually use are 11.1 V lithium polymer or LiPo and they are about 12 volts when fully charged. I am planning on using either these baateries and having to live with a slow robot or use some different batteries. One option would be three 9 volt rechargable NiCad or NiMH batteries but they are bigger and don't give out as much power as LiPo batteries. The main reason for choosing these motors was their length. They are just 51 mm long including the shaft and everything. I am going to cut of a little bit of the shaft and then these two motors can fit side by side inside the maximum 10 cm allowed. My last robot used longer motors and thet were offset by a little distance and that caused problem when the robot was accelerating or pushing and one of the wheels lifted off the ground.

The second and equally important part are the wheels as they are the only way for the motors to transfer their power to the ground. My last robot used wheels I bought from an RC-racing store. They had good grip and were the right diameter. I am going to use them again this time at least for now. A much better solution would be to cast urethane wheels myself but that is a lot of work as they need a very good mold and also the compononents for this rubber are very expensive. When some people from our robotics club buy more of this compound then maybe I will alos use some to make new wheels.

For the brains of this robot I will use the mbed controller I got from the Roadtest event. This is a gross overkill for a minisumo robot but for the sake of simplicity and just because I can I will still do it. The mbed is a development and prototyping platform based on an ARM Cortex M3 microcontroller and runs at 96 MHz. The fasttest sensors on this robot take about 10 ms to update a reading so the controller will have a lot of time to just wait around and come up with strategies based on previous readings. Some of the robots in Standard category actually try to predict the movement of the enemy and also continously calculate their own speed and position and inertia to just stay inside the ring but for this little guy that will probably be a distant dream. One thing I can test beacuse of the mbed and its power is logging every match and later trying to run simulations and maybe improve something.

Detecting the enemy as fast as possible is very important to a sumo robot because the robot is very weak and vulnerable from the rear and the sides. Prettey much all sumo robots use infrared distance sensors to find the enemy and lock onto him. Most robots here used to use analog distance sensors made my Sharp and these were fine until the robots started moving fast enough to actually outrun these sensors. Some robots in the standard class and lately even in minisumo are so fast that the 20-60 ms that these sensors need for every reading is too much. When looking for opponents and turning at the same time the sensor has to react very fast or the robot will overshoot or even miss a target. So the sensors I am going to use are faster but they only have a digital output. This means I can not read the distance of the enemy - I can only know if there is something in front of the sensor or not. These sensor are also smaller and more resistant to noise. I may use on of the slower sensors for long range detection to compliment the fast short range sensors.

The ring for our sumo fights is black and has a white circle around it. This is to help robots stay inside the ring but allow pushing of opponents over the edge. A robot is allowed to cross the white line but touching the area below the raised ring means you lose. Most robots use infrared reflective sensors to detect the white circle and then do a manouver to avoid going outside. Some very simple robots in Roomba category use the built in cliff sensor to detect the fall at the very edge of the ring but that is very often too late especially for faster robots. We have never measured the response time for different reflective sensors but that is something that may need some looking into when more time becomes available. For now I will use some cheap sensors that have the IR LED and detector in on package and also have a very nice mounting hole. These are a bit on the large side but I hope I can fit them in there somwhere. One of our standard class sumos used a distance sensor angled down to the ground some 20 cm in front of the robot to detect the edge of the ring from far away and that proved very useful because a 3 kg robot has a lot of inertia and is very hard to stop fast. For my minisumo I hope this will not be a problem.

I have one L293 driver for the motors right now and I will build something to test the parts and code using this. The problem with chips like these is that they have a large voltage drop and so will use up part of the available battery voltage. A better solution would be something with DMOS technology as these have a low Rds in the on state and because the currents are small they would have next to no losses. I have yet to find a perfect driver for these small motors but the search continues. I would prefer not to use SMD devices because I am planning on building this robot with no custom PCB design. For a faster and better robot I would use high current motors and discrete FET bridges to control them.

These are the most important parts for a minisumo robot but one very important part that is not included here is the actual chassis and scoop. I have seen many examples of good robots losing because of bad mechanical design especially regarding weight distribution and scoop design. The scoop is one part of the robot that is never perfect. Someone will alway have a sharper and straighter scoop. There are a few common designs for minisumo robots that have proved to be effective and ano of them is the classic wedge. This design uses two wheels at the back of the robot and the front rests on the scoop. Most minisumo robots use this design and so will I. Batteries are another thing that I have not figured out yet. LiPo-s would be good as we have a lot of them and we have the chargers and experience but there is the problem with voltage. I could use two batteries in series but then I would need to fit them somewhere. There are actually 22.2 V LiPo packs available but they are big and I have yet to fing a good supplier for them. The motors I am using require very little current - only some 100 mA when stalled and usually a lot less so I think 9 volt batteries in series is also a possibility.

This video is from the previous event that was held in Riga. The video was recorded by a competitor (and organizer) from Lithuania.

I am a member of a local robotics club "TUT Robotics Club" and for most of the robots we work in teams of at least 2-3 people and for bigger projects like one of our standard class sumos we have had some 10 people and worked on it for more than a year. For this minisumo I work solo with maybe a little help here and there. This will be built mostly from what I have at hand and work will be done in whatever spare time I can find.

I will post a link to our club homepage for those interested but it is mostly in Estonian because we have had no use for English translation so far.

http://www.robotiklubi.ee/

There is a small introduction in English also on the page.