Why Do I Keep Finding What I Wish For In TurtleBot Land?

cyclicalobsessive · September 15, 2024, 5:01pm

I have a lot of “robot wishes”, most of which I don’t have time, knowledge, or enough desire to pursue making my wishes come true. Carnegie Mellon University has been implementing machine learning robots for many years, why should my robots require me to program every single function they need.

Some of my wishes over the years:

GoPiGo3 Simulator for trying Python programs, and ROS 2 GoPiGo3 Simulator for trying ROS 2 programs
Basic Simulation “world” of a square room with a couple obstacles
A machine learning program that I could give a GoPiGo3 robot a goal, and the program would run in a simulator on my desk to build a AI model that would solve the goal for my real GoPiGo3 robot.
- For example: Given a map of my house with the dock position and kitchen marked, find a path from the dock to the kitchen
- Example: Given Dave’s dock with an ArUco marker behind the dock, and various poses (position + heading) of Dave in the room with the dock/marker, build a model that drives Dave onto the dock successfully.

Turtlebot3 has had a simple to install robot simulation application for years.
Turtlebot3 has had simple to install ROS mapping for years.
Turtlebot3 has had simple to install ROS navigation for years.

In my quest to convert the Turtlebot3_Gazebo package for GoPiGo3, I kept seeing a bunch of files with “dqn” - there were dqn_stage1/2/3/4_world files, but I didn’t see anything that used them so I deleted them all.

Today, I wandered through the turtlebot3_machine_learning GitHub repo and discovered that dqn_stage2_world is exactly the basic square simulated world I was wishing for,

dqn_stage_2_world

and further they have a ROS 2 dqn_agent to run that does exactly what I was wishing for - “Deep Q Net” will learn to drive a Turtlebot3 robot to a goal. (Albeit the code is five ROS 2 versions stale.)

It seems like I might have saved myself a lot of headaches and time, by buying a Turtlebot3 instead of a second GoPiGo3 to create “ROSbot Dave”. (Or more probably had headaches at a somewhat higher functionality level.)

p.s. the really great video

jimrh · September 15, 2024, 5:40pm

I translate that to “Been there, done that, have an ENTIRE CLOSET full of those T-shirts. . .”

Didn’t you do that with the Create3? You expected the world and were sorely disappointed. Even the big-shot universities with the bottomless budgets couldn’t get that beastie to work and YOU are unsatisfied?

I think there was another one you received that you tossed back DOA too, right? Ultimately you discovered that it wasn’t all it was cracked up to be too.

You keep looking at someone else’s robot and saying “I shoulda’ got that one. . .” - imagining that robot “X”, (the one you didn’t buy), is going to suddenly be the silver bullet that makes all your dreams come true.

I hate to break the bad news to you but that ain’t the way it works. Even if you went out and spent the next five mortgage payments on a TurtleBot Extreme XL-5 with cameras up the wazoo and bumpers and laser rangefinders, and etc., etc., etc., you’re probably going to be disappointed.

As you mentioned so succinctly in a previous article, robotics is hard. Even when I think of my, (comparatively), simple projects, it’s not easy.

Getting multiple operating systems to play nice in the sandbox together. Finding a way to use an on-board display to communicate with the 'bot. Getting the GoPiGo libraries to work on Raspberry Pi O/S version “X”. Hoping, beyond hope, that my earlier projects, (like the joystick controlled robot), aren’t complete junk when I try to migrate that code to something more recent than President Reagan.

And you. . . . You want your robots to be self-aware, walk around the house chatting amiably with whomever it meets, pet the cat, fetch the newspaper, (and read it to you), and so on.

And you say MY projects are “beautiful unicorns”?

Maybe I don’t get it. Very likely so, but I think what you need is to rethink your goals.

As Dr. Laura Schlesinger said in The Proper Care and Feeding of Husbands, “You’ve forgotten to be grateful.”

You have accomplished much. You’ve accomplished more than I can ever hope to remember, and you’ve done it with what seems to be a practiced ease that continuously astonishes me.

And yet. . . . Like the woman at the well, you remain unsatisfied with what you’ve been given and what you’ve accomplished.

I don’t mean to criticize, but maybe you should begin to focus on your accomplishments and successes instead of your failures.

I remember a billboard that had an inspirational message:
“You haven’t failed, you just haven’t finished succeeding yet.”

You, my friend, have most certainly succeeded, far beyond what many dream of, including me.

Be thankful and remember that the goal isn’t a sentient robot, but to have fun trying.

cyclicalobsessive · September 15, 2024, 7:18pm

“Fun” isn’t all it’s cracked up to be. I had a choice: pay behavioral therapy quacks to control obsessive perfectionist thinking, or use the money to buy robots and complain they aren’t perfect.

FWIW, back then I decided to skip trying the Turtlebot3 because the Turtlebot 4 was coming out soon, and then decided to try the Create3 because I already had a LIDAR and an Oak-D-Lite. And yes, both robots fell short of my expectations. Turtlebot3, which I have not owned, does not have bumpers or a dock, but that software (and documentation) … I can’t help but be envious.

Another few months and I will know if the GoPiGo3 with Turtlebot3 software will enable my dreams.

(I saw a disquieting post on the ROS Reddit yesterday. A guy running a Raspberry Pi 5 / ESP32 controlled ROS 2 robot was complaining about the same “clogged pipes behavior” that I saw on the Raspberry Pi5 / Create3 ROS 2 robot. He mentioned something about micro-ROS though, so it may be that his architecture was more like the Create3 than like GoPi5Go-Dave where I have all ROS stuff on the Pi5. )

jimrh · September 16, 2024, 6:11am

My only argument with this is that you end up filling your life with a lot of negative energy.

This plays hob with your endocrine system, releasing loads of stressors into your bloodstream, and (ultimately) negatively affecting your health.

I, (who have my own mental health issues to fight with), am fully aware of the dangers of stressing over things I can’t control.

Sure, you need a goal and you’ve set the bar unbelievably high. But rather than getting wound up that Dave/Carl haven’t achieved “X”, enjoy what they have achieved and renew your efforts towards the next goal.

You’ll be happier and healthier for it.

Idea:
I remember from years ago that there was a particular sub-group in the ham radio world that tried to make the best ham rigs possible using the fewest parts and the tiniest batteries.

They even had competitions to see how far they could get on their “Altoid-tin” sets with some people achieving incredible distance records, even when compared to the “big” rigs.

In a way, you’re trying to do the same thing.

Maybe Dave/Carl aren’t huge multi-kilowatt transmitters with banks of liquid-cooled power MOSFETs as the final, but there’s still a particular joy in getting a smaller set to do “big” things.

And the GoPiGo isn’t really in the “Altoid-tin” class. It has some pretty impressive chops and there’s a lot it can do.

I’d say that this glass is more than half full.

What say ye?

cyclicalobsessive · September 16, 2024, 11:45am

Yes, that’s QRPp, under 1W. Smaller the rig, larger the antenna has to be.
I’m … QRP on 3 watts:

jimrh · September 17, 2024, 8:24am

Maybe you want this beastie?

It has wall following, object avoidance, mapping (?), voice, even eyes and it cleans too!

God only knows how much it costs.