Happy NRW to all of you robot builders everywhere!
Interestingly no Florida events
I saw one just now. But a lot of events seem to be misclassified (looking in WI turned up an event in VA). A lot of these seem very kid oriented, which I guess makes sense.
Yes there is one “Florida Event” which is Northern California First Lego League (NoCalFLL). We can tell there’s no AI in the Robotics Week Event Search Algorithm.
I’m reading the book “The Promise Of Artificial Intelligence”, only to find that the AI that I have been mentally toying with for Carl is now called GOFAI - Good Old Fashioned AI (or First-wave AI). I have observed that many Ph.D’s seem to never move on from their thesis; I see that I have never moved beyond my first paper accepted to a combined International Joint Conference On Artificial Intelligence / American Association for Artificial Intelligence “What Does That Have To Do With The Price Of Eggs (In Russia): Text Processing With Conceptual Graphs” 1988.
Since when does anyone expect computers to actually do what we want from them?
Today I saw a billboard advertising, (what I think was), a robotics/AI conference somewhere in Moscow. Unfortunately I still don’t read Russian that quickly and 80 kmh is a bit faster than I can read and not wreck the car, so I didn’t get all the details.
Coming back I tried to find it again so I could take a picture of it, but I didn’t find it, it was probably one of those electronic billboards and I don’t remember which one it was.
At least the Russians still know how much eggs cost. . .
AFAIK, we can’t even get PEOPLE to “actually do what we want”, or even something as simple as behaving in a rational and reasonable matter!
Any chance of seeing that paper? It sounds interesting!
I have a book titled Expert Systems, written back in 1984 where they talk about an expert system, (MYCIN), developed by Stanford University that ran on a bleeding-edge DEC-20.
Reading their prophecies of massively parallel computing and the “gigantic leaps in AI” that would happen my the mid-90’s - essentially predicting quantum computers on everyone’s desktop - is hysterical!
And that’s a bad thing?
If you’re the one who is the foremost authority on cardiac surgery and invented the heart transplant - you’re suddenly going to throw all that into the trash and become a podiatrist?
If you have a good idea, a firm grasp on that idea, and your sights set on a specific star, why bounce around?
There are things I’m good at and systems engineering is one of them, especially regarding getting things to boot into unusual configurations - my first success being a GE mini-mainframe in 1975/76 - and I have continued studying that for years.
Robotics is fun, but figuring out how to turn the hardware and software into a pretzel and make it do strange and unusual things is my favorite part.
You have plans for Dave and Carl that might be considered the “Drosophila School” of AI - but what of that? Mendelian genetics and “The Drosophila School” are no less valid today than they were in the 1800’s, (Gregor Mendel) and the 1930’s, (Drosophila). These techniques still win people Nobel Prizes - and they represent essential foundational skills needed for everything else geneticists do.
This is work you have every right to be proud of!
I was able to see the value of having a bunch of publications on my resume, but for some reason I did not envision any value to keeping copies of my pubs. (Or perhaps I never wanted to be reminded of how simplistic my understanding of every topic turned out to be after publication. I have never been good about researching topics before I start thinking up solutions.)
At that time, I was tasked with writing a system to evaluate text of intel reports from Soviet soldiers in bars in the context of known brigade operational procedures to give advance warning of possible plans. The system used inferencing on report text classified by limited domain context graphs to fill “possible plan forms” with evaluation of the hypothesized plans in another limited domain context graph to assign confidences.
Of course the paper was a strictly unclassified form which made the technology content abstracted to be of no use to anyone but as a trophy on my resume.
I also have learned that my nature is to try to do everything from scratch, which tends to “re-find” all the domain’s problems without discovering the lessons learned by others that continued studying the topics I briefly countenanced as a debutante.
I, on the other hand, don’t throw away anything.
“bootfix” and “initfix” - my two utilities to enable SpartaDos to boot double-density on the, then, brand new Atari double-density drive is still around in various Atari archives and I still have a hard-copy printout of “Capsule Test”, (in 8080 assembler), that I wrote in the late 70’s.
It’s interesting to go back and review some of these old projects - the tricks and techniques used to make underpowered hardware and software leap through hoops can be valid today as well.
And no. Just because “hindsight is 20-20” doesn’t mean that foresight has to be 20-20 too. If that were true, you’d be winning the 20 bazillion dollar power-ball and running your robots on your private island in the Mediterranean!
And, I say again, that’s a bad thing?
There’s a story I saw on YouTube about some college kid who was late to a math class one day - he missed the first 15 minutes or so - and came into class just as the professor had finished one topic and was moving on to the next.
He quickly copied down the mathematical formulae on the blackboard - which he assumed to be the homework for the next class - listened to the lecture, and went back to his dorm.
He started work on the “homework assignment” and discovered that it was unusually hard - but he figured that this was his “penance” for coming to class late, he missed the explanation - so he’d just have to figure it out for himself.
He struggled and struggled and struggled, often on the verge of giving up - but he eventually solved it and submitted his assignment.
That next week he got called into a conference with his professor - in the Dean’s office no less! Oh &^%$#!!! I’m in trouble for handing it in late and coming to class late - again!
He gets to the Dean’s office and there’s the dean, his professor, a couple of other professors he recognizes, and some other people he had never seen before.
He was asked about his homework - did he get help? (Oh no! They think I plagiarized it!) They then asked him to work it through in front of them - which he did.
It turns out that the formulae on the blackboard in class that day wasn’t a homework assignment, it was an example of a classical “unsolvable problem” in mathematics - and this second-year student had solved it when the best minds in mathematics had tried and failed!
“Doing it again from scratch” isn’t a bad thing - especially if you figure out something that others have tried and failed to do. And you never know until you try.
Yes for mere mortals that reach retirement age. I used to relish the bleeding edge. Now I have to avoid relish and bleeding.
Even so, there’s a lot of work that still needs to be done back of that bleeding edge - and a lot of it is useful and important.
As for me, I tend to avoid sharp edges because otherwise I end up bleeding all over the place!
I found this refrigerator magnet in Bulgaria years ago - I’m tempted to make it my signature.
To mark NRW 2022 iRobot just announced an educational robot that runs ROS2 (as well as other options)
The Create 3 is also the base for the upcoming Turtlebot4.
Yes, the announcement put the nail in the coffin of my “Create3 Beta Tester” aspiration. Back in October of last year I responded to their request for Create3 hardware beta testers, and while waiting to hear I built up a “ROS2-Ubuntu20.04 Raspberry Pi Mobile Controller” to test their Create3 ROS2 API and Gazebo sim. I found a few issues and had a few suggestions accepted.
It was additional experience using ROS2, but I did not learn anything new about ROS2, nor anything that was useful in my GoPiGo3 ROS2 effort. Actually it burned me that here was a $300 robot that came with bumpers, a speaker, sensors, wall following, a dock and docking intrinsics, and particularly designed to interface and power a Pi4. (It lacks “Grove” sensor family support built-in that the GoPiGo3 features, and a RPi, and the great GoPiGo3 projects/examples library.)
In their defense, they did end up offering a 20% purchase discount for my “interest”, but I’m not going to bite.
And that would bring it down to a discounted price of ONLY $3,000?
I wonder how many GoPiGo3’s with distance sensor and Pi-3 you could buy for the cost of one of these beasties? A fleet, with Carl and Dave as the two carrier-class robots leading a flock of minions?
$300. . . Oooh, that’s bad.
And their junior version:
$200. . .
That GoPiGo-3 is looking a bit long in the tooth. . .
$250. . .
I have to admit that, though the GoPiGo wins hands-down as far as “cute” and possibly “expandable” (but I really don’t know that for sure), there’s no denying that if I were going to plank-down cash for a robot today, the iRobot Create-3 would be one heck of a contender - perhaps to the point of purchase.
The only thing needed (IMHO), to make this THE “Killer Robot” would be if they included the blockly and python programming available in their Root models and had a library of lessons and examples.
If it had all of that, this robot would absolutely CRUSH! - There wouldn’t be a robot within light-years of it.
IMHO, the only “bitch part” is the price. I can’t see a school, (except maybe Oyster Bay Cove, Port Washington, or South Hampton’s school districts), that could afford to populate a classroom with a bunch of robots at three benjamins a pop. Then again, I am sure the folks at iRobot are willing to negotiate a “school district” price.
 When I was shopping for a 'bot back in 2018, one of the things that REALLY ATTRACTED me to Dexter Industries in general, (and the GoPiGo series of robots in particular), was the forums. It was a large, vibrant, group of users with both engineering and the Top Brass heavily invested in it.
Though @cleoqc makes a brave front and works hard, (and I respect the living hell outta her), it’s not the same. Except for the occasional student who’s outta their league, there are only three, count-'em, three, non-DI users that are here more often than once a year.
M/R’s decision to move support away from the forum into a more “closed source” environment is really chilling.
If I were someone with cash to plank down on a 'bot - perhaps as an institutional buy - I would have serious second-thoughts about the GoPiGo 'bots after seeing how “intimate” the forums are.
@mitch.kremm - are you listening?
It reminds me of a line, (part of a song), from 1776, where John Adams says:
“Is anybody there? Does anybody care? Does anybody see what I see?”
What say ye?
$240 … which could look like 3000 roofies I guess.
Not bad for a self docking, wall following, left/front/right bumper, six near distance sensors, IMU, wheel drop sensors, front cliff sensor, floor motion sensor, speaker, buttons, lights, a 26Whr Li-Ion battery, with USB-C power and networking, bluetooth and WiFi networking/access point to ROS2 controller running all the necessary ROS2 mobile nodes (sensors, motors, intrinsic behaviors) and separate regulated accessory power.
The GoPiGo3 wins BIG on expandability, comes with a Raspberry Pi, unlimited programming, examples and API in six languages, …
The GoPiGo3 is still a strong and possibly the strongest educational robot today (outside of the ROS education market that Turtlebot has cornered).
Create3 is a half there ROS robot.
Not that I want to “diss” the GoPiGo, but how so?
It’s expandable, but then again, it needs to be.
With the exception of, (maybe), a pan-and-tilt, the Create-3 already has everything you, (I), put into a GoPiGo to make it usable. The sensors that we add are already there - the bumpers, the distance sensor, floor sensors, speaker, even a perforated top to mount a minion on! It even includes a dock and specialized docking sensors - and I am assuming that docking behaviors are built-in.
The one big advantage I see to the GoPiGo is the accessible Raspberry Pi interface pins on the PCB. If they were to re-design to use something other than SPI-1 as the primary interface, and implement true CE SPI support, then that would open up a world of expandability - everyone depending on SPI-1 makes things difficult.
One other thing is, as you said, the vast library of routines. One additional thing is that, though you can use it with ROS, you are not required to use it with ROS - which (IMHO) makes it accessible to a larger audience - including those, (like me), who are frankly scared away from ROS as a robotics environment.
IMHO ROS is like Python - the very few things “it gets right” are offset by the very large part that “it gets wrong” - but that’s an entirely separate thread.
What makes it “half there”? I’m no expert on ROS, but what’s missing?
There is learning with robots, learning about robotics, and there is learning about ROS.
The GoPiGo3 is strongest in learning with robots, can be used for learning about robotics directly, and is weak for learning about ROS.
The GoPiGo3 allows learning about electricity, analog and digital concepts, from buttons to colors. With I2C, PWM/Servo, Programmable Analog and Digital I/O, and serial interfaces built-in the avenues for learning are immense.
The Create3 does not have a general purpose processor included nor any sensor/effector interfaces built-in, so the Create3 is more narrowly for ROS learning, but since today’s ROS robots tend to rely heavily on either vision or LIDAR or both and the Create3 has no built-in interface for either, nor a processor to run the ROS nodes for additional sensors - I consider the Create3 only half there to being a ROS education robot. The Create2 was also weaker than the GoPiGo3 for learning with robots.
The GoPiGo3 fills more topics in the STEM curriculum than Create2 or Create3. Additionally with the “maker focus” in vogue these days, the GoPiGo3 is a better platform.