Older blog entries for tbenedict (starting at number 19)

In response to the work being done by cschur:

First, thanks for undertaking this! I'm building out two new robots: one's a line follower, another is a 4WD chassis initially intended for mini-sumo, but it's also being designed as a general purpose explorer. I was planning to use IR sensors, so I eagerly await your findings.

Second, have you looked at the Sharp GP2D12 and related family of sensors? They use triangulation rather than modulated reflectivity, and seem to be much less sensitive to the reflectivity of the target surface. At one point I was trying to "stealthify" a mini-sumo robot, and found it could accurately gauge distance even when the target had extremely low reflectivity. It was uncanny.

Finally, I seriously look forward to your IR photographs. I've done IR film photography since about 1996, and switched to digital IR in 1999. These days I work at a place that does a lot of IR optics. As part of that I run spectra from 190nm out to 3200nm of any material we put close to an optical beam. As you say, the results can be surprising.

Black electrical tape is black in the IR. Colored electrical tape is not. Antistatic foam, as you pointed out, is quite dark and makes good optical baffle material. Some flock papers are extremely bright in the IR. Others (typically PVC-based ones) are extremely black. Black Delrin is black in the IR. Not all black plastics are. Most black anodization dyes are not black in the IR. Black cotton cloths tend to be quite black in the IR. Black synthetics like satin tend to be extremely reflective. Krylon Ultraflat Black paint has almost the same spectral response as higher priced IR paints (pretty sure they both use carbon as a pigment.)

I'd be curious to see what your findings reveal.

Thanks again,


In answer to Vilita's question about wheels, there are lots of possibilities for a 4" wheel less than 1" thick. There are lots of possibilities for wheels, period. If I was making a robot that needed wheels I didn't have in my shop, here's where I'd start:

Check the hardware store and see if they have any lawnmower or wagon wheels 4" in diameter less than 1" thick. Failing that, check the plumbing department and pick up some PVC or ABS pipe caps. They should be available in about that size. Cut back the flange part of the cap until you have the thickness you want. If you have a lathe, it's a ten minute job at the most. If not, you can do this on a drill press by first hacksawing off as much as possible, mounting it to an arbor, chucking the arbor up in the drill press, and finishing it off with hand files. Run slow and BE CAREFUL!

Check at the nearest Goodwill, Salvation Army, or equivalent thrift shop. If anything there has 4" wheels the right thickness, buy it, strip the wheels as part of your salvage job, and shelve your spares.

Check at the toy store and see if anything has wheels about that size. Likely suspects are small wagons and rolling toys for toddlers.

Make your own on a lathe. This sort of goes back to the PVC or ABS pipe cap idea, but takes it one step further to where you design and build it from scratch. Stranger things have happened. You never know. I made a series of 4" diameter parts on a Taig lathe. Similar work could've been done on a Sherline or Asian mini-lathe. A good setup for any of these will run you between $500 and $1000. It may seem like a lot for a set of wheels, but the lathe will still be cranking out parts long after that robot is shelved or stripped.

Design your own and have someone else make it. Strange though this sounds, it bears mentioning. See if there's a home shop machining club in your area and see if any of them is interested in taking on the job. Better yet, see if any of them are interested in robotics and you may have a teammate. (Better still, see if anyone's selling a lathe used. Then you get your own home shop and you can join their club!)

Failing that, price out the job at a local machine shop. Shop rates vary and some shops price jobs as a way of controlling workflow, so do shop around.


I'm freezing mechanical changes on Shallow Blue until after the Hilo Robofest competition. I made a second set of mechanical bits for my wife's robot, Black Dolphin. The last electrical mod should be done tonight: swapping to a LiPoly battery. I've ordered most of the bits and pieces for Speedy, the line follower I've (not) been working on. The only remaining part still un-ordered is a Baby Orangutan from Pololu. I hope to have the mechanics done by the time the Orangutan shows up.

Depending on how Speedy goes, I'll probably buy another Baby Orangutan for Mule, a 4WD chassis I'm building out. It'll fit inside the mini-sumo size constraints, but I'm thinking more in terms of projects with wider scope: maze, firefighting, mapper, etc. But it'll start life as a mini-sumo for debugging hardware since it's a role I'm familiar with.

Once these are done, that closes out my open robotics projects. Which will open the door to new things! Already people in our club are kicking around some larger ideas: slocum gliders, robo magellan, even talk of an entry in NASA's lunar lander competition (yikes!) But at least I'll have more bench space in my shop to take some of these on.

I'm curious about something for most of the "bug style" walking robots. (This question really doesn't apply to the RoboOne style biped robots.) Almost all of the bug-style walkers have their legs articulated the same way: The entire leg pivots around a vertical rotary axis at the hip joint. Also on the hip joint is a horizontal rotary axis that lets the entire leg move up and down. Past this there's typically at least one other knee joint, also usually oriented horizontally.

Earlier when someone posted about the Boston Dynamics "Big Dog" robot, someone made a comment about how odd the leg joint arrangement was. It's odd for a robot, but not for a dog. The primary joint is a horizontal rotary axis, oriented parallel to the length of the body. This lets the leg swing in closer to the body or out, farther away. Mounted on this is another rotary axis that pivots with the rest of the leg. It's neither always vertical or always horizontal, but with the leg directly under the robot it's mostly a horizontal rotary axis that lets the leg swing forward and back.

It's this hip design (a very good analog to what you see on a cat or a dog) that lets the Boston Dynamics robot have such nimble feet and respond so well to a disturbing force (e.g. a swift kick to the ribs, as you can see on the videos on their web site.)

The only other place I've seen that close an analog between nature and robot was the study done at MIT on the mechanics of cockroach leg joints. It's not nearly as obvious as one might think, and bears little to no resemblance to the hexapod arrangements available from most robotic supply houses. Among other things, the hip joints more closely resemble those of BigDog.

So I'm curious about two things: One, why haven't more people tried this hip orientation? For accelerating from a dead-stop, it places more of the loads along the leg rather than across it (much easier on the joints!)

Second, and more important, did I just ask the "but why?" question that's going to get me out of mini-sumo and into my next phase of exploration in robotics?

In a way I hope the answer to the second question is no. I love building mini-sumo robots. In another way I hope it's yes. Walking robots pose REALLY neat questions I'd like to play with.

I guess I really hope the answer is "both"! It means I get to make walking mini-sumo robots! Though hitting the 500g limit with three or more servos per leg is pushing it...


Robotics projects have been put on hold while we recover from the earthquake out here in Hawaii. My robots survived fine (some books fell on one of them, but that's it.) The shop more or less survived fine. Mill and lathe are still on the bench, though I did have a hand scraper and a couple of other tools fall. My wife had more tool avalanches than me, including some torches (thank goodness we never leave tanks attached!)

Most of my time has been spent at work trying to get things operational there. It's amazing how violent an earthquake can be. It's even more amazing to see how fast our facility has been brought back online. It's entirely due to the efforts of some phenomenal people. With any luck we'll be back on the sky tonight, maybe even before the bad weather clears.

Back to work!


I got a nice lesson in infrared optics, the necessity of reading a datasheet before using a product, and the concept of paranoia in a robot.

I've been adding sensors to my mini-sumo robot, Shallow Blue. I finally installed the last of them, a rear-facing Sharp IR proximity switch set to trigger at 40cm and closer. For now any time the sensor trips, the robot is set to spin around 180 degrees. So far so good.

I loaded the code, verified, plopped it in a ring, and turned it on. It immediately set to spinning around and wouldn't stop! Paranoia!

Of course it's simpler than that. The vertical spread of the beam was impinging on the floor, giving the sensor a reading closer than 40cm. It was doing exactly what it was told to do. I was able to verify this in a dimly lit room using an older CCD camera (Nikon 950), which could image the robot as well as the beam spot from the sensor. A quick modification to wedge the sensor at an angle to the floor solved the problem.


Our club had its third meeting last week. Two mini sumo robots and two line-followers (with one doing double-duty). There was a lot of enthusiasm for the mini sumo competition, and it looks like we'll have more active builders over the next month. Four people have declared an intention to, and two have already purchased hardware. I'm hoping we can hook a fifth.

I'm building a second mini-sumo, a stock Mark III, which I'll be using to tune up new changes to Shallow Blue. Once that's done the sensors and boards will be ripped off the new 'bot so the 4WD platform I'm building out can be finished. Not sure if that'll happen by the next meeting, but I can hope.

The rest of the stripped Mark III, a chassis two servos and two wheels, will be outfitted with an R/C receiver for remote control ala Battlebots. This isn't in an effort to develop an R/C mini sumo class, but as a way to get people interested in the club. We've got two public appearances coming up in December, and being able to hand someone a transmitter and say, "Want to try your hand against an autonomous robot?" may possibly draw new people into the club. Have to see.

The rest of the meeting went well. It looks like we may have a FIRST Robotics team forming up in town, and there was a call for mentors. There should also be several Ballbot teams forming up before January, and possibly at least one underwater ROV team as well. In addition it looks like we're getting a fair number of "big kids" interested in tackling some larger projects. Should be a busy year!

In response to svo's post about the steam engine (nice engine, by the way!) I'd say a good, solid confirmed on that one! My first steam engine was made entirely on the lathe, not using a mill at all, even for all the little rectangly bits and drill patterns. So it's nice to see the opposite situation holds true as well.

So here's a challenge... First, some background:

A recent series of articles in Machinist's Workshop (or Home Shop Machinist... I take both and don't distinguish much between them) described a 4-cycle gas engine called the EVIC Twin. What sets this thing apart is that it uses electronic solenoids for the valves, and is entirely controlled (throttle and all) from a PIC.

Now for the challenge:

An oscillating steam engine isn't the most efficient piece of equipment on the planet. (My first steam engine was an oscillator, too.) This is why oscillating engines were almost never used on locomotives. This is also why all manner of valve gear was developed over the years. So if the idea of the EVIC was extended to steam, and a steam engine was put entirely under microprocessor control, what would such a beastie be like?

Any takers? Anyone actually want to build one?

In answer to bear on building cheap robots: It always depends on where the bits and pieces come from. Good example: I've got about fifteen steppers lying around on my bench. They came from scrapped floppy drives. Cost? $0 monetary outlay, though as you point out a fair bit of time was spent removing them. Another example: Last time I took trash to the transfer station, someone had dumped a bag full of R/C toys. I picked 'em up. One just needed a battery charge and some contact cleaner on the power switch. (Free toy for the kids!) Another had a fried transmitter. Question is, do I fix or replace the transmitter, or just gut the car and stick a microcontroller in it? $0 monetary outlay, and not much more than that for time invested so far. Offroad chassis for a RoboMagellan project? Possibly so. (Though my kids would kill me if they heard me say that about "their" toy.)

Problem with this approach is it doesn't scale. If you're doing commercial robots, you can't rely on surplus or salvage. This is one reason why commercial robots tend to be a lot more expensive than what can be built at home, even given similar construction methods. It also doesn't give any gurantee of identical spares. In that stack of steppers on my bench, only two are of identical make and model.

One more question to ask when estimating cost: Do you count the cost of items previously purchased? For example, the line follower I'm building is getting its entire drive train out of spares from my ROV and mini-sumo. Total cost? I could say zero since the parts are just lying around. Total cost for someone trying to duplicate what I'm building? Considerably more, especially since I didn't buy the motors as single units.

I guess what all this boils down to is this: Do people estimate cost based on their out-of-pocket expenses, or do people estimate cost based on what someone else would have to spend in order to reproduce what they built?

Even putting it so simply still leaves you with hairy questions. For example, do you add in tooling costs? I've got a complete machine shop at home and a decent electronics shop as well. I don't add in machining or tooling costs, but someone trying to reproduce what I do would have to.

I'd be curious to hear what other people's practices are when they estimate cost on a project.


Our club had its second meeting. Bigger turnout, and three mini sumo bots (including mine). Shallow Blue got its butt thoroughly kicked. I've got a list of changes to make before the next meeting: one software, one hardware, and one electronic. Good mix.

Turns out there aren't rules for the line following event yet, so I'm planning for right angle turns on the high speed line follower I'm making. People laughed when I said I was after two meters per second. Fine by me. I doubt I'll reach that. But it's good to have a goal, and it'd be a riot if I could hit that, even if it's only on the straightaways.

The club issued two challenges to its members: Make a robot that can track a straight black line across a white field, and/or make a robot that'll stay on a table. One, of course, is in preparation for building a line-follower. The other, of course, is to get people thinking about how to keep a mini-sumo bot in the ring.

The next meeting should be fun!


10 older entries...

Share this page