Resume/CV

For those who want to know my professional history in a single page, here is my fairly recent resume:

For the rest of you who want an informal and lengthier narrative about my professional, geeky, robot-related history, just keep reading this page.

Table of Contents
  1. Italian Institute of Technology (2008-2011)
  2. Internship at ATR (2010)
  3. Research at Nagoya University(2006-2008)
  4. Nagoya University (2005-2008)
  5. Work History (2000-2005)
  6. Robots at the University of Washington (2000-2004)
  7. Homebuilt Catamaran (1998-Forever)
  8. High School Robots (1996-2000)
Italian Institute of Technology (2008-2011)

I am presently pursuing a PhD at the Italian Institute of Technology (IIT) in the department of Advanced Robotics. My thesis touches four topics of research:

  1. Nonlinear series elastic actuator design
  2. Fiber-reinforced polymers for lightweight robot construction
  3. Balancing algorithms for highly dynamic robots
  4. Rigid body dynamics algorithms and simulations

My contribution to the state of the art is mostly integration work across multiple disciplines, to combine good ideas into a coherent whole. I love applied math, programming, and building things, so my thesis is admittedly more applied than theoretical. Videos, papers, software, posters, and photos are pending publication acceptance.

Until then, you can just look at this artist's rendition of my planned robot and laugh:

An artists representation of my robot

IIT is in many ways an unusual place to get an advanced degree. IIT was conceived was as an institution where researchers in neurobiology, robotics, nanotechnology, and drug discovery could mingle in close proximity, presumably to cross-pollinate each other with interdisciplinary knowledge.

Despite its creation on paper in 2003, a building was not purchased until 2006, and by the time I arrived in 2008, most of the laboratories had just finished moving in...although the elevators didn't fully work until 2009! In fact, I believe that my group is the first batch of PhD students to spend all of our time at the IIT building in Genova -- before us, the different research groups were located in different labs and even in different cities across Europe. It looks like I arrived right at the sweet spot just when things were getting started.

As a research institution IIT is exceptionally well funded (50-100 million euros/year for 10 years). I have heard it claimed that the IIT receives more money from the Italian government than the rest of the institutions in the country combined. It's also pretty international; in my department the number of foreign and Italian researchers are about equal, although I think there are more Italians than foreigners when the entire population of ~400 researchers is considered.

Although the goal of IIT is to become the MIT of Italy, to be realistic at the time of writing (Dec 2010) they aren't at that level of world-class research yet. On the other hand, the young average age of the researchers at IIT, and the slightly chaotic atmosphere at IIT gives me plenty of flexibility and resources to pursue whatever topics interest me.

IIT is also great as a place to rub shoulders with smart people working on cool projects. For example, my ex-roommate is the designer of this robot:

The guys who sit a few desks behind me made world's first pancake-flipping robot:

Along with another ex-roommate, they also taught the iCub -- the signature robot of IIT -- to use a bow and arrow. Their work was even shown on the Colbert Report (scroll to 1:40):

The Colbert ReportMon - Thurs 11:30pm / 10:30c
Droid Rage
www.colbertnation.com
Colbert Report Full Episodes2010 ElectionMarch to Keep Fear Alive

Regardless of Italy as a country being a spectacularly beautiful, chaotic, and often bureaucratic mess, IIT remains ambitious experiment trying to bring Italian research up to a world-class standard.

Internship at ATR (2010)

I spent six months in 2010 working at the Advanced Telecommunications Research (ATR) Laboratory near Nara, Japan. My contribution was to write the initial realtime control software and for a hybrid electric-pneumatic exoskeleton providing 0 to 100% assistance when walking. The basic idea of the project was to use the low-bandwidth pneumatic actuation to compensate gravitational torques, and to use the electric actuator for feedback, stabilization, and the more difficult control aspects of balancing. My advisor, Sang-Ho Hyon, recently published a paper entitled "XoR: Hybrid Drive Exoskeleton Robot That Can Balance" in IROS 2011, if you are curious.

Exoskeleton

Another fun project I developed while at ATR was a novel monotone cubic spline-based interpolation system to quickly create whole-body posture control and complex but smooth motions using a keyframe system. This type of quick development is extremely useful during testing for isolating mechanical problems, developing whole-body motions that may later be optimized or constrained further, and even creating simple demonstrations of the technology for publicity reasons.

This postural control system was tested on two types of robots: the 'secret' robot that will be described after publication, and the 38-DOF hydraulically-actuated humanoid robot CB:

CB and joystick

Control of the robots could be accomplished in real time by creating keyframe poses using a standard Playstation-style joypad. These poses were then smoothly interpolated using the monotone cubic splines, and chains of sophisticated actions could be built within minutes.

Mostly, though, I made it because I am amused by dancing robots...

CB and I

I also learned how to use the SL simulator (developed at USC by Stefan Schaal's group), and started working on a fast-dynamics approximation system for baseball batting, but this project was quickly abandoned. Here's a screenshot showing the humanoid robot with a bat, batting box, and sweet-spot trajectory.

CB and I

Research at Nagoya University (2006-2008)

My masters thesis was concerned with the design of high-efficiency biped robots based on the passive-dynamic walking phenomenon. Passive-dynamic walking robots are legged, mechanical machines which walk down a shallow slope...and do so without a control system or motors. Although this may sound like a rather useless type of robot, some researchers hypothesize that by basing a robot off of a mechanical structure which moves naturally in ways similar to walking we may be able to construct robots which walk with better energy efficiency. The concept is similar to how the Wright brothers first developed a glider, and then added a powerplant to the glider to make the first airplane.

My contribution to the field was the development of a new measurement for quantifying the gait robustness of passive-dynamic robots, an accurate rigid-body simulation system for measuring this gait robustness quantifier, and a novel variable-stiffness series elastic actuator suitable for use in a robot based on the passive-dynamic walking phenomenon.

I developed an actuator which I call the VSSEA (Variable Stiffness Series Elastic Actuator). It is essentially similar to MIT's SEA, with the exception that the effective stiffness of the springs can be adjusted.

A vertical hang test.
A vertical hang test.
Hang test with sensors.
Hang test with sensors.

To explore the theory and practice of passive-dynamic robots, I also wrote a simulation in common lisp (SCBL) which uses the matlisp and cl-opengl libraries.

A screenshot of the custom simulation environment.
A screenshot of the custom simulation environment.
A picture of a slice of phase space of the compass biped, showing the basin of attraction of the limit cycle in green.
A picture of a slice of phase space of the compass biped, showing the basin of attraction of the limit cycle in green.

Although I am happy with the way my thesis turned out, the contents of the conference papers look somewhat contradictory when I look back on them. I suppose this is to be expected because during the course of my research, I learned a lot, read a lot, and changed my mind several times.

If you note any mistakes or misattributions in my thesis, please tell me.

Below are two copies of my master's thesis presentation, one in English and one in Japanese. At my actual thesis defense, I gave the presentation speaking in Japanese, partly because the audience was entirely Japanese and also partly because I wanted to demonstrate the level of fluency in Japanese that I had built up (JLPT Level 1).

The contents of the presentations are essentially identical. I also included some extra slides at the end which were used for answering questions.

I also made a few posters during the course of my master's degree. I created them by cutting and pasting text from the above papers, so they ended up with way too much text. On the upside, as a result they are mostly self-explanatory and may be interesting even if I am not present to narrate.
Nagoya University (2005-2008)

I was lucky enough to receive a full-ride scholarship and living stipend to the University of Nagoya, from which I graduated with a M.S. in Mechatronics. I also did research at the now-defunct RIKEN Bio-Mimetic Control Research Center as part of my curriculum.

My scholarship was provided by the Monbukagakushou (aka MEXT or 文部科学省), and I highly recommend the program for people from any country who are interested in studying in Japan. The scholarship applications are available through your local Japanese consulate, and the stipends typically last from 6 months to 3 or 4 years, with extensions available for up to a total of 6 or 7 years.

In my case, I managed to finish the equivalent of 2 years of college-level Japanese and learn roughly 1000 kanji before arriving in Japan, so after a year of living in Japan and further study I was at a level where I could take the graduate school entrance examination to Nagoya U. in Japanese. This was necessary because the second and third years of my stay in Japan I needed to follow the course lectures, which were also in Japanese.

As an aside to those people interested in getting the scholarship, although you may survive at some institutions in the country (particularly in Tokyo) without real fluency in Japanese, to really enjoy the country you are going to need to have as much fluency under your belt before arriving as possible. Many of my friends did not have sufficient fluency when they arrived in Japan, and found themselves in English-speaking bubbles without real contact with the society. So above all, my advice to you is not to trust the Japanese to teach you their own language -- they spend years learning English, yet usually have poor fluency. You will need to do as much as possible before arriving there, and take matters into your own hands when you do arrive by making Japanese friends (or girl/boyfriends, preferably).

Work History (2000-2005)

I have only worked at small, startup companies, which I think is rather unique.

Also, some recent news compels me to write a small disclaimer and explanation regarding the above work.

I was one of the first employees of A Dot Corporation in 2003, now known as MOD Systems, known in a previous incarnation as Fullplay, which was previously known as Interactive Objects during the dot-com days. I knew the then CEO/CTO, Mark Phillips, from about 1998, when he contacted me over the internet to teach him about electronics. We were in contact until about 2005, when I left his company to go to Japan, and Mark then founded MOD Systems. Looking back, during the years that I lived in Seattle, I believe he tried to act as some kind of entrepreneurial mentor to me, although I didn't fully realize it at the time.

On the one hand, I mention this because Mark was a successful entrepreneur who secured multi-million dollar angel investments and produced real products. I worked for his companies as a grunt-work college intern several summers in a row during my university days. It was interesting and a unique experience to watch first-hand how a small company grows from less than half a dozen people working from a spare bedroom in a apartment to a spacious office employing 30 people. Also, despite my age and lack of experience, Mark often gave me work designed to help me improve my engineering ability, and was clearly investing in me as an engineer for the long term.

On the other hand...well, it appears that Mark has made some mistakes since I knew him last that would make him a less than ideal reference.

His legal troubles place me in a somewhat awkward position because was an excellent professional reference for me in the past, as not many college students know millionaire CEOs and CTOs or have worked at startup companies. Of course, I don't wish for my own name to be tainted or associated with any criminal activity he may have committed, nor do I wish to in any way condone or excuse criminal behavior. Although I am not aware of the details of what he did, his actions appear to be both legally and morally wrong, plain and simple.

Still, Mark's bad behavior appears to have began years after I left Seattle, during which time I have been pursuing my Masters and PhD degrees in robotics around the world. So despite his recent mistakes, I still am proud of the cool technical work while under his employ. I believe that most reasonable people would consider it unfair for an honest employee such as myself to be penalized by association by an ex-boss's behavior years after leaving a company. Accordingly I am not omitting this portion of my work history.

Hence, all the above explanation to clarify my interactions with Mark, because I still value the work I did employed by his organizations, but I also wish to avoid any condemnation through simple association.

Robots at the University of Washington (2000-2004)

In 2004, as part of my electrical engineering capstone project, I built a pair of robots capable of moving holonomically on a flat surface. The word 'holonomic' refers to a type of mathematical constraint, but in its simplest form it roughly means that the number of degrees of control is equal to the number of freedom. For example, a car is not holonomic because you only have two degrees of control (steering and forward/backward), while there are three degrees of freedom (up/down, left/right, and rotation). The holonomic robot that I built was capable of controlling all three degrees of freedom simultaneously, meaning it could spin while driving in a straight line, or move in any direction from any orientation.

If cars had holonomic drive systems, we wouldn't need to worry about parallel parking.

As for the details of the project, my two partners and I struggled not to make the software or digital electronics, but the motor amplifiers. There was not just one, but three fires created during the course of this project. People thought it was funny to put a fire extinguisher next to our bench while we were working, "just in case."

If the pictures are intriguing, go ahead and read the full lab report, minus the 100+ pages of source code that we used to beef up the writeup, which of course was done at the last minute because we preferred to work on the robot itself for the majority of the project.

Here's the holonomic robot without any circuits, and you can see the undercarrige of the robot.
Here's the holonomic robot without any circuits, and you can see the undercarriage of the robot.

Here's the big robot, as close to finished as we ever got with it.
Here's the big robot, as close to finished as we ever got with it.

Ahh, what a beautiful, yet tragically flawed robot.
Ahh, what a beautiful, yet tragically flawed robot.

The board on the left is the motor drivers, which caused 90% of all the pain and suffering (you can see fire damage). The board on the right contained the PICs, FPGA, and IR communication circuitry.
The board on the left is the motor drivers, which caused 90% of all the pain and suffering (you can see fire damage). The board on the right contained the PICs, FPGA, and IR communication circuitry.

We hacked a nintendo powerglove, and also an original nintendo controller, and added a PIC to dump out bytes over serial IR to control the robot remotely.
We hacked a Nintendo powerglove, and also an original Nintendo controller, and added a PIC to dump out bytes over serial IR to control the robot remotely.

These motors were sent by the devil himself.
These motors were sent by the devil himself.

An early schematic of what the robot might have looked like. But, the gearboxes couldn't be fabricated easily, so we dropped the idea.
An early schematic of what the robot might have looked like. But, the gearboxes couldn't be fabricated easily, so we dropped the idea.

Here's the test robot we made in conjunction with the larger robot. The hope was that at least ONE of the two robots would be completed in time.
Here's the test robot we made in conjunction with the larger robot. The hope was that at least ONE of the two robots would be completed in time.

Another view of the lovely small robot, which did actually move around in a limited fashion.
Another view of the lovely small robot, which did actually move around in a limited fashion.

Note the reduced amount of circuitry on the robot compared to the large one.
Note the reduced amount of circuitry on the robot compared to the large one.

Aww, isn't it cute?
Aww, isn't it cute?

You can see the battery peeking out from underneath the robot. Yee haw! Ride that 9.6V of raw power, regulated down to 5V of sheer boredom.
You can see the battery peeking out from underneath the robot. Yee haw! Ride that 9.6V of raw power, regulated down to 5V of sheer boredom.

In 2003, I designed a robot-controller-on-a-chip in my VLSI class. First I designed basic logic gates, such ANDs, ORs, DFFs, etc in the silicon wells. Then I designed a the state machines, sensor interfaces, etc in VERILOG, and using an autorouter and automagic tools laid out the chip. There are extra spaces between the n-wells to make routing easier, since we were rather limited in the number of metal layers we could use. The whole chip worked fine in simulation and if we had a 90um process manufacturing plant, I could manufactured the chips and made a semi-intelligent robot vacuum cleaner! In the end, though, I've come to the opinion that although ASICS are kind of cool, FPGAs are way cooler.

A DFF.
A DFF.

The entire microchip.
The entire microchip.

In 2003, I was in several robotics classes, and one of the more fun ones involved playing with LEGO mindstorms. We designed and programmed little robots to play autonomous golf, foosball, and solve mazes. The games were run entirely automatically, and started by a robotic referee via wireless. The competition was fun, although I was very frustrated by the sensor limitations of the HandyBoard. I remember wishing desperately for real IRQs and an little FPGA for some glue logic, but not having enough time/motivation to add them myself. I settled for making my robot sing the Mario Brothers theme song in all its square-edged glory as it ran the maze.




In 2002, I built TETRIS...entirely in hardware (ie, using lots and lots of 74HC and TTL chips). That means no programming, no CPU, and no software. It was a ridiculously difficult project, and we eventually cheated a bit by using a few one-time-programmable GAL chips to implement state machines because using just DFFs would take too long to wire up. In the end, the collision detecting circuit never worked, so all you could do was draw a block on the "screen" (a collection of LEDs) and move it around. It cost over $300 and used about 100m of wire...we could practically hear the AM radio with tat much wire. Somewhat to my surprise, we somehow managed to do all of this in only 2-3 weeks, and had we had an extra week I believe we might have finished debugging it.



In early 2001, I participated in the FIRST robotics competition as a advisor to some local high school students. It was a lot of fun, although our team only placed 75th or something out of 300 teams. The competition was kind of like 2v2 basketball, with various bonus points for different tasks. We chose to exploit one of these extra tasks, designing our robot to fall forward from a vertical position, extend its width by about 50% so it was wide enough to carry another 75kg robot over a small bar about 0.5 meter high while flipping its wheels over a 2x4 board nailed to the ground. It was too complex and never worked very well, but rather amazing to see when it did work.





In Fall 2000, as a freshman I worked on building an antenna to find the source of EM and ES interference coming from a Z-Axis Pinch (ZAP) Fusion Device. The project was basically a joke because I'm sure the professor could have purchased a better antenna than the one that we made, but we did learn the basics about antennas for measuring EM and ES fields. The conclusion was, perhaps not surprisingly, that most of the interference was coming from the wiring and the connection plate (where tens of thousands of amps of current would flow for a few microseconds at a time).

It's not really a robot, but this seems like the place to put this sort of geek project.

The ZAP fusion device.
The ZAP fusion device.

You can see the source of the noise: the wiring and connection plate.
You can see the source of the noise: the wiring and connection plate.

Our little hand-made antenna.
Our little hand-made antenna.

Catamaran (1998-Forever)

One bored high school summer, I began building a catamaran with my best friend. It was several years later before we could put it in the water, and I hesitate to say that the boat is actually complete, even a decade later. One thing I learned during the project is that boats should never be placed near liquids, because nothing breaks boats faster than water. In fact, the safest place to keep your boat is in the middle of the desert, where no dangerous water can get near the damn thing.

This project, more than any other, deserves an entire book written to it. For now, you'll just have to enjoy these photos. (Warning: A few steps are omitted from the following tutorial. Particularly between steps 4 and 5, and also what happens if there is a problem at step 8...)

Step 1
Step 1: Cut out foam slices.

Step 2
Step 2: Glue them together.

Step 3
Step 3: Sand the edges smooth.

Step 4
Step 4: Cover in Fiberglass

Step 5
Step 5: Apply gel coat.

Step 6
Step 6: Paint!

Step 7
Step 7: Make metal pieces.

Step 8
Step 8: Test.

cables
Checking the cables won't kill us.

sailors
Posing like scurvy-ridden sailors.

launch
Launching the boat.

drink
Enjoying life on the water.

peace
Bringing peace to all nations.

High School Robots (1996-2000)

You are now entering a section so old and stagnant I hesitate to include it at all -- but whenever I talk to middle or high school students about robotics, they invariably want to see photos of what robots I built when I was their age.

Much of the text describing these robots is similarly ancient, although I recently tried to correct some of the most juvenile parts of it without destroying it's youthful feel. Yet, I wonder: Isn't there some internet law or statute of limitations for stupid but enthusiastic writing you did as a teenager, so that what you said can no longer be held against you a decade later? I can only hope...

So without further ado, the following photos are from the few surviving robots that I still have today.

Robot Family Portrait #1
Robot Family Portrait #1

Robot Family Portrait #2
Robot Family Portrait #2

Family Closeup #1
Family Closeup #1

Family Closeup #2
Family Closeup #2

Family Closeup #3
Family Closeup #3

Family Closeup #4
Family Closeup #4

A Carbon Fiber Hexapod Robot

In 1998, my friend Mark and I decided to build a pair of hexapods, originally planning on mounting small cameras on them and doing some fancy image processing stuff to have them drive around semi-autonomously. Most of my previous walkers had four legs, but we wanted to try something more ambitious and complex, if only for the sake of complexity being cool. I liked the look of the old http://www.lynxmotion.com Hexapod II walkers, which are pretty cool kits and have gone through a few revisions since initially introduced.

The completed robot looked like this when we were done.

As you can see, the design is modeled on the Lynxmotion robot. Using a copy of the free beta version of Rhino3D, I drew up a design:


This is the leg design, version 1. Kind of clunky, and you'll note it doesn't use servo motors.

This is the second version of the leg, the one we finally decided to use. Smaller, lighter, simpler.
This is the second version of the leg, the one we finally decided to use. Smaller, lighter, simpler.

Amazingly, despite the great hard drive crash of '99, I still have the exact plans on my computer. For those interested, the original DXF file and some images are available for download.

Most robots are made of plastic, aluminum, or steel. I was leaning towards using PVC or ABS plastic plate for our robots, as Lynxmotion had built their robots out of that, and I had already experienced how simple it was to cut and shape plastics with hand tools. However, we were able to procure some 10 x 22 inch sheets of 1/8 or 3/16 inch thick bi-directionally woven carbon fiber plates that had been compressed and baked in an autoclave for maximum strength. The carbon fiber plate was simply too cool not to use.

The hexwalker sits on my desk.
The hexwalker sits on my desk.

When we started to work with the material, however, a minor problem came up: CARBON FIBER IS EXTREMELY HARD TO CUT!

We had several failed attempts at cutting it, starting with a hacksaw. The blade was dulled frighteningly fast. So, we bought a band saw with large and small toothed bands. Another mistake. The carbon fiber dulled the cutting teeth, and the band saw gave us curved cuts that weren't acceptable. Even worse we weren't even a quarter of the way done with all the parts.

So, we pulled out the big guns: a table saw with a diamond abrasive wheel. Finally, we could cut the stuff like butter. The only drawback is that it has a huge cerf -- the width of the cutting tool, which determines the amount of material that is turned to dust during the cut -- and we were forced to wear dust masks to keep from breathing in the carbon fibers.

A different perspective of parts on the table.
A different perspective of parts on the table.

As for drill bits, we dulled a few regular bits before we started using 8 or 10 titanium nitrate (nitride?) tipped drill bits, which lasted us much longer, but still became dull after being used for an hour or two. And were there a lot of holes to be drilled! By last count, we drilled 552 holes in our robots...not including mistakes.

The best way to round the edges turned out to be the simplest way: hand files. Mark liked using the 25,000 RPM grinder, which cut really fast, but for the final shaping the files were best.

All the parts required to build two robots, laid out on the table. You can see that we already assembled a single leg.
All the parts required to build two robots, laid out on the table. You can see that we already assembled a single leg.

Once we figured out how to cut, drill, and tap (thread) the carbon fiber plate, construction went along more smoothly. It was a fairly simple matter to copy the printout onto some cardboard... cut out the cardboard template... use a pen to mark the size of the part on the carbon fiber plate... cut the plate using the table saw... carefully mark the points to be drilled... drill the holes...use a high-speed grinder to do the rough rounding of the edges... file the part so the edges are more perfectly round... use a tapping set to thread some of the holes drilled... wash the part...and put it on a clean table for assembly.

A single leg, held up against some drawn out plans for another project.
A single leg, fully assembled.

Yay! It stands!
Yay! It stands!

It took a surprisingly long time to put the two robots together after all the parts had been made, probably on the order of 6 hours or so. I mean, you spend a couple of DAYS making all the parts, and you figure you can bolt it together in an hour or so, but that's simple not the case.

There were a couple design changes along the way, the most important being that the frame supports are lighter than planned, and are modular, allowing us to reassemble it as a 4 legged walker with a little effort and time.

The quadruped form of the robot.
The quadruped form of the robot.

Our workbench in the foreground.
Our workbench in the foreground.

I controlled my robot with a 68HC11 microcontroller acquired from the Seattle Robotics Society. It was surprising to me how much tuning was required to get the thing to walk evenly. It really required a lot of effort and time to get all the servos properly calibrated for accurate position control.

Here's my robot at my desk, with the 68HC11 mounted on it and tiny power tether coming off the back.
Here's my robot at my desk, with the 68HC11 mounted on it and tiny power tether coming off the back.

Basically, to control all the servos simultaneously, I had to send a 0.5 to 1.5 mS pulse to the servos every 20-40mS, a more difficult feat then it sounds, since the number is going to be different for each of the servos!!! Similar, but different because of manufacturing errors (there are lots, believe me!).

This is the 68HC11 BotBoard, a nice kit that I got for cheap at the Seattle Robotics Society.
This is the 68HC11 BotBoard, a nice kit that I got for cheap at the Seattle Robotics Society.

My humble programming station. That's a _really_ old box, but it works fine for programming stuff via the parallel port.
My humble programming station, circa 1998. That's a _really_ old box, but it works fine for programming stuff via the parallel port.

Because it was so much fun just making the thing walk by pressing buttons on the controller (A small box with momentary switches running into the input port of the 68HC11) that I didn't ever get around to making it autonomous or adding any sensors to the robot until years later. Finally, in 2004 I finally added an ultrasonic distance sensor and extra servo to the front of the robot to provide a primitive forward looking sonar, as well as some infrared and bump sensors.

Actual terrain handling abilities of the robot are fairly poor. Carpet, hard surfaces, and small obstacles it can handle without a problem, but anything higher than 1.5" is nigh impassible. I have heard from someone at SRS that my robot performed better than the Lynxmotion robots (at least their old model), but I am skeptical of their estimate.

All in all, a fun project however.

Side view
Side view

Angle View
Angle View

Another angle view
Another angle view

Carbon 3-Motor

This is undoubtedly the best performing walker I have ever built. Fast, strong, lightweight, and much lower power consumption than the hex walker above. I think it is interesting and an important result that the simpler something is, often times the more effective.

This robot is not strictly complete in the photos because I ripped everything but the motor drivers out (you can still see them hot-glued in place for ruggedness), planning to rebuild the brain with more intelligence using Mark Tilden's little gold bicores (last two photos).

It's a three motor walker, so there are two drive motors and a waist motor. As photo #4 shows, it can step over quite high objects. I couldn't quite get enough traction on the feet to get it to walk over objects its own height, but it was very close. I think if I put better feet on it and changed the leg shape, it could probably do it.

The robot has a high center of gravity, so to keep the robot from falling over, there are travel stops to restrict the range of motion of the legs and keep the center of gravity in a safe region.

Also, it has two stickers that Jeff made for me, which state "Let it bring peace to all nations", and also "Robo Air". I think they speak for themselves.

Let it bring peace to all nations.
Let it bring peace to all nations.

Side view
Side view

Robo Air
Robo Air

When robots attack!
When robots attack!

Mark Tilden's cores
Mark Tilden's cores

Surface mount cores
Surface mount cores

The Bulldog

I actually didn't name this robot. I brought it down to Los Alamos and put it in a sumo contest with another guy's robot as sort of a joke (the other robot was much larger). It actually did pretty well, thanks to it's low center of gravity, although the other robot won the informal contest. Anyway, a guy who was filming it was narrating and called it "the Bulldog", so I kept the name since it seemed appropriate.

You're probably thinking it doesn't walk very well with the motors 90 degrees to each other, but it actually manages pretty well for a two motor walker (3 and 5 motor walkers are much smoother). This was originally a three motor walker, see below.

8 batteries means it can go for a couple hours before wearing out, so this robot was demoed a lot when nothing else worked. Also, it can actually walk upside down just as well as right side up, as the fourth photo shows.

Front left view
Front left view

Front right view
Front right view

Closeup
Closeup

It can walk upside down!
It can walk upside down!

Ye Old Bulldog

This was the precursor to the above robot. It had three motors, and walked reasonably well. However, I quickly stripped one of the motor gearboxes, and I didn't have any replacements, so I rebuilt the robot into the two motor version shown above. All I did was replace the front half of the robot with legs, it was a quick fix.

Sorry for the extremely blurry photos, I didn't know how to take photos in high school and I all I had was a disposable camera. These photos are the only evidence that remains of this robot.

Front left view
Front left view

Front right view
Front right view

Aluminum 5-motor walker

This was a lighter, better working version of the following robot. Like many of my robots, it never was given a final brain, only prototyped on protoboard. I absolutely hate debugging PCB's that don't work after soldering them together, so many of my high school robots only had functioning brains implemented on rapid prototyping boards.

Aluminum walker top view
Aluminum walker top view

Aluminum walker front view
Aluminum walker front view

Alimunim walker side view
Aluminum walker side view

Alimunim walker side view
Aluminum walker side view

Old aluminum 5-motor walker

This robot was a mechanical test. I was originally going to place large solar panels on top of the flat area, but it was simply too heavy, so it was rebuilt into the above robot.

Old aluminum walker side view
Old aluminum walker side view

Ed Bot

This very simple robot was built with my friend Ed, who successfully made solaroller, and wanted an easy first legged robot. He has a near identical robot, modeled after this one.

If you were wondering, the battery and circuitry normally goes on the robot between the motors...This is the "eviscerated entrails" look.

Ed Bot
Ed Bot

More Ed Bot
More Ed Bot

Still Ed Bot
Still Ed Bot

Daddy Long Legs

Another robot built with a friend, this used some pretty strong gearmotors and an interesting way of hooking up the motors to the frame. The frame is angle aluminum stock, and the motors are connected using a flat piece of aluminum wrapped around the motor and screwed into the frame, acting like a clamp. Works nicely. A good first robot.

Daddy long legs #1
Daddy long legs #1

Daddy long legs #2
Daddy long legs #2

Gift Bot

This was a gift for a friend. However, I thought it was boring because it had no sensors, so I never actually gave it to him. What a good friend I am.

Low profile
Low profile

Angle
Angle

Side
Side

Erector Walker

This ancient beast was a fun little thing I built with my friend's erector set. I love the claws on the front. One motor doesn't allow for turning, of course, but a mechanical leg system like that worked pretty well, all things considered.

Side view
Side view

Angle view
Angle view

My Miller Walker

For those familiar with BEAM Robotics, this was built following the original "Miller Tutorial" to building a simple microcore and two legged walker. It was my first legged walker, and one of the more cleanly designed ones.

The last photo is just silly, but it could actually sort of walk forward like that if you threw the battery out the back of the robot to counterbalance. It kind of looked like a dog begging for a treat by walking on its hind legs.

Note the two bump sensors I added to the robot to make it backup and turn (last two photographs only). It couldn't turn that sharp, having only two motors and strange mechanics, but it IS possible if you vary the timing carefully and add a few extra neurons.

Angle shot
Angle shot

The bottom
The bottom

Side view
Side view

With battery and sensors
With battery and sensors

Walking on the hind legs
Walking on the hind legs

Altoid Walker

This was a mechanical test robot I made after meeting Mark Tilden in Los Alamos. He showed me and some other people how to make really great springy legs for lightweight robots using catheter tubing and copper rod. He also suggested that an ideal form factor for a demonstration robot would be one he could put in an Altoids can, since you could carry it everywhere.

I never hooked up any electronics to this guy, because it was obvious that the center of gravity was back too far that it could never work anyway. Steeper than about 20 degrees and it tumbles backwards. I abandoned this type of walking configuration, but solarbotics sells a walker kit of this form, although much larger in size.

Altoid walker closeup
Altoid walker closeup

Altoid walker and altoids box
Altoid walker and altoids box

Pipe Bot

Here is an experiment in PVC pipe. I was trying to make a robust outdoor robot, so I made it out of PVC and was going to place rechargeable batteries, electronics, etc inside the tubes, and then seal it off.

Although an interesting idea, PVC is much too heavy. Thinner wall tubing would help, as would some strong, waterproof waist motor system. I abandoned this robot when I realized it was too heavy to walk with the motors I had planned.

Too heavy
Too heavy

Side view
Side view

Top view
Top view

PIC Programmer

Not a robot, but a useful little programmer I made that I have used a few times. It's the "No-Parts PIC Programmer" that I saw in some electronics magazine a few years ago. Simple, cheap, and it works. Note the wonderful workmanship *cough*.

The pic programmer.
The pic programmer.

Note the fine workmanship.
Note the fine workmanship.

Plasti Dip Example

Also not a robot, but a neat trick to waterproof and ruggedize a free-form circuit. You can get this stuff at a local hardware store for dipping tools and such. Smells awful, but when it cures it feels like a soft plastic or rubber. Here I have used it on a very small photovore as protection for the circuitry.

Plasti dip covering parts
Plasti dip covering parts

Crawly

Here's an interesting robot. If you are as big a robotics geek as I am, when you saw Star Wars: Episode 1, the best part of the movie were those kick-ass rolling robots with the shields. I loved the way they rolled around and then unfolded like some giant rolly-polly bug, so I tried to build one.

What worked: Rolling up. What didn't work: Unrolling (it always unrolls upside down!). It also never really inched along like I had hoped, although it thrashed around like a dying worm very nicely and sort of moved forward. The mechanical links between sections had too much friction, they need a redesign. There were two wires running the length of the robot that were supposed to be tensioned to let the robot make arcing turns, but they were either too lose, or too tight, it always interfered with other movements. Maybe elastic or springs would work better? In any event, the cables kept snapping.

For the curious, those copper plates are actually cut away sections of three metal toilet bowl floats. Naturally, wash them first before cutting apart. The copper is very easy to solder to, but corrodes very fast. If I had to do it over again I would add some sort of clear coat protection to keep them shiny. I do like how the ribbed edges give a nice artistic shell effect to this robot.

The worm
The worm

The underside
The underside

Curled up like a pillbug.
Curled up like a pillbug.

Solar Plant

This was just for fun. It would spin the little metal thing on the bottom (originally bright and shiny) when exposed to sunlight. I call it the solar plant, since it's more like a robotic plant than an animal. Hey, its got suction cup roots! Now I just need to make the other robots feed off of it, and I'll have a robotic ecosystem...

Sadly, after several years of being stuck to the OUTSIDE of my room's window, the motor finally corroded and this plant has stopped spinning.

Looks like an alien baby
Looks like an alien baby

The legs don't do anything
The legs don't do anything

Solar Buzzer

A bit of a joke. I found some pager vibrator motors were too strangely shaped to be used in a robot, so I made a noisemaker out of them instead. It's lightweight, and after about 10-50 minutes of charging, depending on indoor lighting, it will make a loud buzzing sound for 15 seconds or so if placed on a flat surface. It's just long enough that you forget completely about it, and then BZZZZT! Surprise!

It's important that this guy be placed on a FLAT surface, since otherwise it tends to vibrate off the shelf and fall to the floor. This has happened several times. Luckily, the robot hasn't been damaged yet.

Top view
Top view

King Blowhard

This was my (incomplete) entry for the Seattle Robotics Society firefighting robot contest. It's interesting because I tried to enter this contest without using a microcontroller, which is an interesting (read: nigh impossible) challenge. Unfortunately, I didn't know much about circuit design at that time, and the little state machines and sensors that I had rigged together did not work. The wheels were also slightly crooked, and I had difficulty getting IR sensors on top to detect the wall reliably (hence the bump skirt at the bottom).

The strategy of this robot is rather crazy, although creative. It only turns clockwise, powering only one motor. When it hits the wall or the IR sensors trigger, it switches to powering the other motor. This means it sort of tumbles along the wall, following the wall.

On top there were two infrared sensors placed at the focus point of flashlight reflectors (not in the photos, sorry). When they detected the candle's light signature, the robot's behavior would change and it would charge straight at the candle, powering the big squirrel cage fan on top and blowing the candle out. Hence the name "King Blowhard".

All that explanation is strictly theoretical, however, since the robot never got past the "crash into wall" phase and switch to the other motor phase. It was too big and cumbersome, and I started only two weeks before the competition deadline, meaning I didn't have nearly enough time to finish the project.

King Blowhard front
King Blowhard front

King Blowhard angle
King Blowhard angle