HomeCoverTECH NEWSFrom House Brew to Hasbro

From House Brew to Hasbro

Twenty years in the past, simply after the dot com meltdown, I used to be working as a advisor and had numerous free time on my fingers. I began indulging in my passion of constructing robots, and subsequently met up with different like-minded individuals on the Silicon Valley House Brew Robotics Membership. One of many first individuals I met on the membership was Bob Allen, an electronics engineer who had a machine store in his storage and will make superb mechanical issues.

Bob had constructed some fight robots prior to now, and we had been each enthusiastic about working collectively on extra attention-grabbing and sophisticated robotic tasks the place we might use all three robotics disciplines: mechanical, electrical, and software program engineering. Again then, my mechanical engineering abilities had been severely missing. Robotic-suitable mechanical elements had been exhausting to return by, however fight robotic elements had been extra available as a consequence of their recognition on TV on the time.

I had a pile of those elements and was keen to place one thing collectively. I keep in mind one afternoon going over to Bob’s home to debate the potential for doing a collaborative robotic venture utilizing this elements pile. His first remark was, “are you loopy? You’re going to destroy your own home if you happen to construct it with that stuff!” He then confirmed me a video of a robotic that might stability on two wheels, and I assumed it was superb. Plus, it was small and possibly wouldn’t destroy my home. Bob and I puzzled if we might construct one thing comparable. It couldn’t be that tough, proper?

Months later, in 2002, Bob and I lastly had a set of electronics put collectively, with the correct motors found out, and we had been able to take a crack at making it stability and drive round. Then got here the exhausting half. Constructing the robotic was one factor, however the software program problem of constructing it autonomously stability on two wheels was monumental. I feel it took a superb three months simply to get it to not fall over, and lots of extra months to make it drive round. The primary robotic was named “Bender” after the lovable beer-drinking robotic from Futurama. It fell down a lot that it made sense to call it after a robotic that was consuming on a regular basis.

A small two wheeled robot balancing with exposed electronics
Bender the Balancing Robotic, inbuilt 2002.OLogic

I didn’t know a lot about management principle after I began this venture, however I realized quick. A balancing robotic is basically an upside-down implementation of the basic inverted pendulum drawback. I realized the exhausting means making a balancing robotic that works effectively requires the next gadgets:

  • Excessive-torque geared motors with encoder suggestions
  • An IMU (a mixture of a gyro, accelerometer, and compass)
  • A microcontroller-sized CPU

The massive software program problem for a balancing robotic is that you simply want a management loop the place you learn the IMU, take away all of the drift and bias, after which transfer the wheels of the robotic within the appropriate route to maintain them beneath the robotic so it doesn’t fall down. It seems that the pace of this management loop just isn’t terribly crucial. On all my early makes an attempt, about 30 Hz was sufficient to maintain the robotic steady. It nonetheless sounds fairly straightforward as I’m describing it, however when you dig into the main points, it will get difficult. In 2021, the IMU models you may purchase off the shelf had been fairly cheap and dependable. In 2002, this was not the case, and I needed to roll my very own by taking every sensor individually and fusing them collectively in software program.

I purchased a brand new two axis accelerometer, however it was fairly noisy, and had linearization points. What this implies is if you happen to tip the accelerometer on its edge, and slowly rotate it by means of 360 levels, it is best to have the ability to draw an ideal circle from the information popping out. The information from this accelerometer regarded extra like a backyard squash than a circle. Then, I discovered some gyros at a surplus electronics warehouse—single axis piezoelectric units initially made for camcorders and pc mice. They weren’t broadly used, however for some motive the excess retailer had an enormous field of them.

The gyro is a very powerful measurement machine that you simply use for retaining the robotic upright. It offers you the speed of rotational acceleration because the robotic begins to fall. The issue is, if you happen to simply put a gyro on a non-moving floor, like a desk and begin taking readings from it, the zero level will slowly begin to drift. Compensating for this drift is the exhausting half: how are you going to measure a sensor when zero is consistently shifting?

The reply lies in an algorithm from statistics and management referred to as a Kalman filter. You’re taking your sensor measurements, run them by means of a Kalman Filter, and you may produce estimates of the place you suppose the true measurement really is. My math abilities had been fairly rusty from faculty, and after I dug into find out how to implement one in all these, I actually had to return to fundamentals and study what to do. It took fairly a bit of labor to get to the purpose the place I might construct a easy Kalman filter algorithm that I might use to fuse collectively my janky gyro and accelerometer. However as soon as I had it working, I might see on the bench that when my little IMU board was not really shifting, the output of my filter would keep steady, and when it was shifting, the measurements would even transfer within the appropriate route!

A small two wheeled robot with exposed electronics, camera, and sonar sensor
Flexo, inbuilt 2003.OLogic

Subsequent, I wanted a management principle technique to maintain the robotic balanced. There are numerous other ways to go, however I wished one thing easy. My mind harm after determining the entire Kalman filter factor, so I went with the only resolution, which is a proportional–integral–by-product (PID) controller. It seems {that a} PID controller all by itself is normally simply barely able to retaining the robotic balanced, however it doesn’t essentially cope with all the issues of driving round, carrying an offset load, or all the problems associated to bumping into one thing. I realized I had so as to add numerous additional phrases and software program filters to my PID equations, however I started to realize nice outcomes.

It was lastly time to make a smaller, improved model of Bender that might do extra issues. One factor I realized from all my work was that greater balancing robots are normally simpler—since they transfer slower, the management system doesn’t should be as exacting in making minor corrections. However making a very tiny balancing robotic turns into tougher and tougher the smaller you go. Within the TV present Futurama, Bender has a brother named Flexo. So, it made sense to call the brand new robotic after him.

Since I now had a PID algorithm ok to do issues like stability a robotic with an offset load on prime of it and drive up and down ramps, I might simply mount a shifting, pan-tilt-camera rig to the highest with none concern of the movement of the digital camera knocking the robotic over. I used a shade monitoring digital camera tuned to trace the colour purple, in order that if you happen to put a purple object in entrance of the robotic, it will attempt to transfer in direction of you. As soon as it obtained shut sufficient, it will simply stand there and stare at you. We took Flexo to the Homebrew Robotics Membership as soon as he was working, and there was a child there carrying a purple sweatshirt the robotic actually took a liking to!

As soon as we knew find out how to do the fundamentals, we constructed an off-road model of Flexo referred to as Large Purple, and one other model referred to as Fats Albert that might stand up by itself from laying on the ground.

Fats Albert had a novel mechanism and software program implementation that allowed for transitioning from the ground to standing after which again to the ground once more. We utilized for a number of patents on the entire setup and have stored them energetic through the years, however we nonetheless haven’t found out what to do with them—you probably have an thought, let me know!

Bob thought we’d have the ability to discover some corporations who is likely to be enthusiastic about paying for our experience in growing these sorts of balancing robots. We purchased a sales space on the first RoboNexus commerce present held in San Jose in 2004. On the primary day, we had been standing in our sales space with Flexo displaying off his balancing act, and who ought to come down the aisle driving a Segway, however Dean Kaman himself. He rode up, and stopped for a second, and mentioned jokingly, “Wow! Take a look at that! I’ve by no means seen something like that!”

On the second day, we met a bunch of enthusiastic engineers from Hasbro. They had been in search of extra attention-grabbing robotic toys they might add to their line, in addition to some assist with extra superior sensors in toy growth. We had been invited out to Hasbro company headquarters in Rhode Island to present a presentation to the engineering staff about what we might do. Bob requested if we would have liked to arrange a proper presentation, however Hasbro informed us to simply present up and produce some robots. Once we arrived, we had been shuttled into an infinite convention room that regarded prefer it might seat 100 individuals, and there have been 20 individuals sitting in there ready to listen to the presentation that we didn’t have! We thought we’d utterly blown it, however we simply turned on the robots and Bob began speaking. Quickly, the group warmed up, obtained out of their seats, and began coming to take a look at the robots up shut, and all of it labored out after that. Bob and I began a robotics consultancy referred to as OLogic Inc. in 2005 with Hasbro as our first buyer.

A man stands next to a statue of Mr. Potato Head
Ted Larson at Hasbro HQ in Rhode Island.OLogic

Hasbro thought that our balancing robotic thought might turn out to be a fantastic client product, so that they made the funding to begin turning Flexo into one thing that might be offered as a toy. Step one was to eliminate the excess elements and combine consumer-grade elements and motors. There have been two robots made: one was a full 3D printed prototype that was fragile as glass, and the opposite was simply the drive system with our current electronics mounted on it.

Left image show
On left, a prototype drive practice for AMP. On proper, the primary AMP prototype, referred to as Glass Man, inbuilt 2006.OLogic

We broke each of those prototypes extra occasions than I can keep in mind. Each time it broke, I’d name up Bob, and he would come over and we might work out a intelligent option to glue it again collectively. After tons of trial and error we finally had one thing that labored, after which it was time to make the actual product. After a couple of journeys to China and assembly some unbelievable manufacturing specialists, and a bunch extra work on our half, we lastly had the actual product able to go.

The product was to be named A.M.P., for “Automated Music Character,” or AMPbot for brief. It was a balancing robotic, however it additionally had a reasonably candy boom-box inside it, so it was a cell speaker system too. It got here with software program so it might dance alongside to the music being performed.

At this level it was 2008, and AMPbot was scheduled for launch in early 2009, however that was to not be. The monetary disaster of 2008 modified the priorities for all of the toy corporations, and after solely a couple of dozen pre-production prototypes had been made, AMPbot was cancelled by Hasbro.

By 2010, we had been wishing there have been methods we might revive the AMPbot venture. We obtained permission from Hasbro to take it to market on our personal if we might discover sufficient funding to make it occur. We reinvigorated the inventory of restricted prototypes we had constructed, and took AMPbot to the Demo Convention in 2010 and tried to launch it as its personal venture. We discovered numerous robotic fans who wished him of their houses. I imply, who doesn’t need a cool music taking part in robotic?

A group of AMP robots surround a black table holding drinks and playing cards
What occurs whenever you depart AMP’s alone in a single day within the workplace.OLogic

We attended a number of commerce reveals to try to search out curiosity for making AMP. Right here I’m describing AMP at CES in 2012:

To maintain OLogic going whereas we tried to determine find out how to get AMP out the door, we spent our time selecting up robotics consulting jobs. We had been in a position to get a one-off venture to construct some little balancing Android Robots for Google, and tour them round to occasions like Google IO, Google Sandbox, and Google Zeitgeist in London. Bob and I had numerous enjoyable—it felt like a travelling circus, bringing our robots to all of those occasions.

Three green two wheeled balancing robots that look like the Google Android logo
Balancing Android robots constructed for Google.OLogic

A variety of two wheeled balancing robots in strange shapes and colors
Ideas for a wide range of completely different balancing robots.OLogic

There’s each a cheerful and unhappy ending to this story about balancing robots. I made many associates alongside the way in which and began a cool robotics consultancy that’s thriving right now, however it appears that evidently AMPbot was by no means meant to be. Bob Allen has since retired from OLogic, however he can nonetheless be seen every now and then at a House Brew Robotics Membership assembly, as his love of robots lives on. The prototypes of AMPbot nonetheless adorn our foyer at OLogic, however we’ve since moved on to different tasks, and for now my balancing robotic saga is over.

Ted Larson is CEO and Co-Founding father of OLogic, Inc., an embedded methods analysis and growth firm with a concentrate on robotic purposes and applied sciences. Ted has over 30 years of expertise designing and constructing industrial software program purposes. His present curiosity areas are robotics, synthetic intelligence, and embedded methods in client {and professional} digital units.

From Your Web site Articles

Associated Articles Across the Net



Please enter your comment!
Please enter your name here

Most Popular