Combat Robots

To avoid the trouble of pouring another motor mount, we bought a new mount from the Robot Marketplace. It’s pretty nice, but pretty simple. It makes me wonder how hard it would be to avoid all this casting mess with a mill. A mill would be nice.

It’s interesting to see how far we’ve progressed on this project by investigating the mounts. We started off with some jigsaw’ed 2×4’s that clamped down with rubber sheets for friction. Many facepalms were had. Working on this robot was another educational experience.

It was like communicating with myself last summer and wondering why in the world I chose to do things the way I did. Axle too large? Hammer it in. Discount possibility things will have to be taken apart later.

Here’s a nice garage-spanning table I made with my dad as a bonus. It’s nice that we don’t have to work on the floor anymore.

Here are the failed attempts at shaft collar wheel hubs that sheared off during our previous run. We started with JB weld, moved to epoxy/gorilla glue? mixture, and settled with some extra strength epxoy. Still didn’t work.

However we have opted to mount the reduction sprocket directly onto the wheel. By the sheer weight of the things we were afraid the hardened steel would be impossible to machine with our inadequate tools. However, we found out that the inside is extremely soft most likely cast iron. Machining worked well, and this solution seems like a confident final solution to our wheel hub problem. It’s a wonder we didn’t think of this painfully obvious solution earlier. More facepalms all around.


Last spring semester (2011) I made this pan/tilt setup to mount a turret for Mech Warfare. It’s been stagnant for a while but I might be able to buy the airsoft tank cannon soon. I have an incredibly small 150′ range camera system set up so I might be able to test the full setup soon. However, I haven’t worked on the hexapod for a while as I don’t yet have the funds to buy the nice Dynamixel servos I wanted to use. Hobby servos just won’t cut it for me for a walking robot.

Mechanically, its based off the ServoCity pan/tilt system made of ABS. I made this one out of blue laser cut clear acrylic scrap, and modified it to use the lazy susan turntable as reinforcement for the pan servo axle.

Some time next summer after I’ve gotten used to programming quadrupeds I might consider rebuilding that hexapod for competition 2013. Control is shaky on the standard RC airplane transmitter, and hopefully serial control with the SSC-32 might be better. However, I used some pretty cheap Towerpro servos from Hobbyking for this pan/tilt, so I’m not really confident in the lifespan of this pan/tilt. Hey, at least they were cheap.

With my digital camera taped onto the mounting platform.

There it is. The rebirth of Spirit MK1. It is essentially a moving DDR platform box on wheels as of now. Our old solid steel frame was way too heavy and extremely overbuilt. It was good welding practice, but that old one piece frame just didn’t like to be modified. We tried to change that with our second frame iteration. This frame was made from 1″ aluminum square bar that we initially planned to braze or weld together.

The only welder we encountered was some random guy standing in line behind us at the metal supply store, and our HTS-2000 (advertised to be “stronger than the surrounding material”) failed miserably to stand up to any impact tests. and thus, the bolting began…

Six bolts per corner, three module boxes (drive left, drive right, and center) totaling hundreds of bolts. After we put it together, it started vibrating apart. Assembly number two included hundreds of tension washers. I would not be surprised if the added weight of the steel bolts has negated any weight/strength benefit from using aluminum, but at least the frame will stay shiny without sealing now. Our goal was simply to make the new config drive, which is what we did.

If you look closely you can see three cast motor mounts. Yes! They were a success. However the last mount took forever to make (too many unsuccessful pours) so we might just buy an official motor mount to compensate. On the bot is instead one of our unsuccessfully poured motor mounts. A permanent solution can wait.

Otherwise the configuration is relatively similar to Spirit MK1. We have divided the robot into functional sections. The drive modules just include the transmission. The center module will house the batteries, motors, and electronics. A separate weapons module will be installed at the front/top, while armor plating will cover the exposed sides.

We spent quite a bit of time attempting to make hubs that would effectively restrain the wheels to their axles. However, most of our attempts have been failing. For this iteration, we attempted to double epoxy two shaft collars (that conveniently are the same diameter as the wheel bore) to transmit the torque. After a successful run and rounds of cheering, the robot attempted a tank steer while beached on a rock (we built this thing with low clearance to avoid wedge attacks). The stall was enough to shear off some of our temporary wheel hubs and left the wheels to freely spin.

It will have to wait until winter break however, I’m going to start school soon.

The past week we’ve been trying to get a better mold making Green Sand mixture to use. Sifting the sand and using only fine particles works pretty well, but the large particles are prone to clumping still. Part of the reason for this is probably because the “bentonite” we’ve been using is probably just clay and filler from the cat litter. Clay is still supposed to work, since it is supposed to naturally bind to sand particles. Or you could say we’ve been playing with dirt for a week. “Indian Clay” in grade school nomenclature.

Lukas purchased a coffee grinder from Walmart to try to grind the insoluble clumps in the cat litter. Not only can it grind clay, apparently it can turn sand into smaller sand dust. We are using that as our molding mixture now, but it’s still just wet sand. It is very difficult to remove the negative from the mold with the wet sand adhering to it. We managed to make a rudimentary mold that is much higher resolution than the last attempt.

The issue with removing the negative however is making an uneven mold, so a double cope and drag isn’t possible yet. The next attempt will hopefully include a casting flue and a completely enclosed casting space so the product won’t oxidize as much (opposed to the muffin tin we have now).

Melting was facilitated by simply adding a steel cap to the steel grass jelly drink can. By melting smaller bits of aluminum at a time and using the cap, we were able to melt .75 lbs of aluminum in half the amount of time as attempt #1.

Our mistake was letting the coals die down by the time we were ready to cast, so the resulting pour was only partially liquid. Since we’ll be having to cast this same piece four times, I’ll probably keep the coals fully loaded the entire time.

The pour was “too cold.”

The resulting cast looks much better than attempt #1. There is little oxidation on the surface, and little slag due to the cap on the crucible while heating. There was virtually no steam from the mold, but the backside still has the impressions of the sand from the mold.

Front Side

Back Side
Hopefully the third time around will be a charm. The two part mold and hopefully improved negative removal technique will bring this motor mount to acceptable standards for use.

Old post from 7/1/11

Motor mounts have always been a problem. SPIRIT uses two 3″ Magmotors, which have no mounting screws or protrusions. They are just plain cylinders. We attempted to make a two piece wooden mount with rubber inserts that clamp down on the face of the motor. However this approach risks the rubber being sheared off the wood by the torque of the motor.

Commercial motor mounts for this motor are generally expensive. They are just pieces of aluminum that clamp down with bolt tension to apply friction to the motors. After shipping, these mounts would total hundreds of dollars. I thought of directly casting the motor mounts as a possible method of making adequately strong motor mounts. Since the dimensions were listed online for the commercial mounts, I could just copy and mold that instead.

For our first attempt, we tried to make a sand casting rig with bentonite, sand and water. Green sand for sand casting can be made that way. Too bad the source of bentonite we used, kitty litter, had so many insoluble impurities that the mold was difficult to make even. With our lack of prime moulding materials and experience, we decided to make a muffin-tin mold and just pour from the top without a casting spout or a two part mold.

The dirt-n-brick furnace I have is pretty badly designed. It’s just a pile of cinder blocks shielded by a dirt mound. The heat efficiency is terrible, and it took a long time to melt the aluminum. However, once a puddle of liquid aluminum formed on the bottom of the crucible, melting happened very quickly. And did I mention that the crucible we used was just a tin soup can. This wasn’t an emergency casting attempt #1 for nothing. Two blow dryers were directed at the furnace opening as bellows, and paraffin tea lights were thrown in the crucible for bursts of heat (at the expense of adding impurities).

Since this was such a brash procedure, we didn’t bother to scoop off the surface slag. That was later very apparent in the quality of the final product. W also didn’t wait for the mold to dry most of its water, and we had used way too much water during the green sand mixing process. As a result the pour went as expected. At parts water spot boiled leaving large vacuums where the aluminum would not flow. The slag accumulated near the top and weakened one side of the cast. The lack of a two part mold encouraged heavy oxidation on the top side. Here’s what happened on video:

The end product was pretty disfigured, but the concept works. Had it not been for the steam bubbles, the cast would have been high enough quality to be functional as a motor mount. The resulting product was around half a pound and was heavily oxidized on one side. Regardless, the attempt was just an experiment and shows promise, so we will try again with a little more effort on the mold and get things right the second time around.