AlexTheProgrammer - building a Real Robot for Real Life for everyone's house!

@alextheprogrammer

And the AI for it would be my primary focus ...

And that is the hard part 🙂  It all started with Shakey. I still remember seeing it on a tv science piece narrated by Walter Cronkite.

@robotbuilder love this video! It's old but the concepts are still very relevant! 

@alextheprogrammer 

And those same concepts for planning and pathfinding (which is planning) to achieve some goal are used to give AI to the NPC (non playing characters) in video games.

In one sense these characters are simulated "robots" that have goals and make plans to achieve those goals and have reflex actions and so on.  I spent some time coding simple simulated worlds and their simulated agents and it gave me a good feel for how to write AI for a robot.

My other interest has been machine vision but I like to program at the low level,  more fun then ready made solutions from OpenCV as I like to understand how things work all the way down to the lowest level.

A robot identifying the position of an orange on a tree is the first step but actually planning how to pluck the orange and responding to force feedback as it does so is something else again.

I wanted to build a robot something like Loki but I don't think it will ever happen now. 

http://dshinsel.com/loki/

@robotbuilder 

Posted by: @robotbuilder

A robot identifying the position of an orange on a tree is the first step but actually planning how to pluck the orange and responding to force feedback as it does so is something else again.

Here you're approaching a very interesting topic and that's what being ignored by a lot of robot builders - the actuator (grabber). There're many different ways of building them but the best one, IMHO, is the one that resembles a human hand as much as possible in both shape and ability to receive tactile feedback. My robot will have both of those. I watched numerous videos about different projects that companies build around the hand, some are more complex than others.

Some are really cool like the flexible electric muscles ones but the muscle is really hard to produce at home it costs millions to buy. Though I think, some of them can be manufactured and I saw the videos about that. This is the best example of that btw:

But I also found a very simple example for inspiration here in terms of mechanics - a kids toy with fingers, ability to expand and contract the palm, fingers, and thumb:

I also didn't find a good enough way of measuring grip and pressure gauges aren't gonna work due to price and shape so instead I have something completely different in mind - capacitive skin that would help me with both grippinness (the outer layer is gonna be PTG plastic) and an ability to register how hard and sufficient that grip is by having two conductive layers within it that you can think of a capacitor with variable distance between the terminals. 

Btw, is it your website http://dshinsel.com/loki/ form the footer? Those look really cool! How did you build the physical parts (enclosures, frames, etc)? Did you 3D print them?  

@alextheprogrammer 

That is NOT my website. As I wrote I wanted to build something like Loki it wasn't something I had built myself.

I have seen the hydraulic cyborg hand.

Many devices such as these can do a lot of thing if controlled by the human visual system. A lot of those utube robots doing stuff are human operated for AI even if they are using advanced mechanics.  I had a toy owi arm that I could use to pick up blocks easily.  However using a webcam on its elbow and trying to write code to do the same thing not so easy.

There are lots of examples online but the problem is cost. Only with mass production can that cost be reduced 

Have you seen the James Bruton robot videos?

@robotbuilder 

Just discovered the channel and watched da few relevant videos - very useful! The force controlled robotic grabber and arm are the best ones: 

Yeah, he spends a ton of money of all of the parts and building but I guess all of that is from his sponsors (printers, filament, components, etc). 

It's pricey to build on your own and if I wanna outside funding (aka a proper startup), then I will need to be doing this full time. 

@alextheprogrammer 

I think this should probably be moved to the other robots section?

I think he can provide a lot of useful information on building and programming a robot.

For a domestic robot which might be useful for picking up objects and cleaning benches I think this arm designed would be the easiest to build and program.  He doesn't seem to have done anything since this last video.

@alextheprogrammer 

The human hand evolved the way it did for very good reasons. The example above is really very limited in the things it could manipulate. The human environment was designed to work with human hands and that includes any hand held tools the robot might like to use. There is also the option of being able to attach different types of hands on the working end of the arm that has the three basic moves, two rotations at the wrist and an open and close action.

The human hand would be difficult to match let alone the sensory/motor systems that controls it. We can "see" objects with our hands when we touch them. With visual feedback we can use mechanical hands such the the extensions we might use to pick up something from the floor or from a high shelf.

Visual feedback is the only way you can compensate for limited touch and pressure feedback systems. Until the vision problem is solved or an artificial skin is developed covered in touch, temperature and pressure sensors, machines that manipulate the environment will be very limited. The hand extension aids are an example of what can be done with the human visual system and the force feedback to the hand via the handle.  This is why the or the prosthetic hands used by amputees are useful.

So I suggest you think about the tasks you want the machine to carry out and then work out how that can be achieved with the crude mechanical systems (compared with animal bodies) that we can actually build and control.

There is a lot of hype around robotics and indeed AI in general.  It is nothing like human intelligence. Yes AI in control of the nukes and some inane goals to achieve may decide to eliminate mankind but it will not be because it has become sentient.

@robotbuilder Actually, this robot in terms of base and the raising platform is almost exactly what I'm building! 

And he's using the same Intel RealSense D435i camera that I have! And he's also using ROS! Looks like I really need to watch the entire series of videos about this robot! 

@alextheprogrammer 

He was a toy designer by trade. He has certainly been productive.

His work on the really useful robot seems to have come to a halt?

https://www.youtube.com/user/jamesbruton/videos

People get all excited about building a robot but then hit a brick wall when they find out that even with 2.5 million dollars of advanced hardware it is harder than it looks when it comes to useful AI solutions.

"The idea of a robot helper that could navigate a human-centric environment and interact with it in the same way we do was, of course, attractive.

But the problem proved, and still proves, harder than anyone guessed. Even the latest humanoid robots fail spectacularly at the most ordinary tasks that humans do without thinking. Asimo, which operated in a sort of semi-pre-programmed manner, was far behind even these limited capabilities."

https://techcrunch.com/2018/06/28/honda-reportedly-retires-the-iconic-asimo/

@robotbuilder I noticed that after hi kinda finished the robot's physical design and build, it all came to a halt, disappeared. The same happened with DB1 too! 

I think one of the reasons for that might be the fact that different ppl have different skillsets and are good in certain parts of building a robot: some are great mechanical engineers, some are great electrical / electronics ones, and other, myself including, are the best with software but not so great in the two disciplines above. To me, building that physical body of the robot would be just the beginning. James said in one of the videos that he's not a software engineer and that part is the most challenging to him. 

As for myself, I at least wanna try and hit some kind of wall myself. Ideally, it should be past the mechanical and electrical phases. I have spent quite a bit of time refreshing my electronics knowledge from schooling from over 10 years ago using LinkedIn learning tutorials - I had a lot of advanced electronics classes back then but I didn't appreciate them back then and the technology was pretty dated and Arduinos weren't even mentioned. I also learned Fusion 360 as these days it seems to be the most popular CAD for mechatronics, more popular then AutoCad from my schooling days and 3D Max that I know pretty well but it's not really a CAD at all but more an animation software. I got myself a 3D printer and already printed quite a bit of parts. That also had some learning curve to it. 

Another challenging thing about building a robot is costly components. Some are really costly like Lidar, others are not so costly but the costs add up. I recently started buying stuff from AliExpress for that reason - I'm ok to wait for a month but pay $100 for what normally runs to $1000. 

But once I'm done with the basics (effectively, the Really Useful Robot), I have a ton of ideas to try that I didn't see anyone else doing. 

@alextheprogrammer 

@robotbuilder I noticed that after he kinda finished the robot's physical design and build, it all came to a halt, disappeared. The same happened with DB1 too! I think one of the reasons for that might be the fact that different ppl have different skillsets  ...

It is more than a skill set. Even highly professional and successful programmers are not writing AI programs that can make robots do anything approaching human performance.

Another challenging thing about building a robot is costly components.

Exactly. So it may be a lot cheaper to buy a second hand or even a new electric wheel chair for a robot base. Economy of mass production. And purchase one of the commercial robot arms they now have which makes James Bruton's robot arm look like the weak toy arm it is.  If you have good ideas and programming skills the robot companies would be glad to have you try them out. Recently I saw a news item on a robot arm that packs avocados and a quick search I found it here.

https://startup.network/startups/426402.html

That is where the money will be and the perfect environment for testing ideas. I see they offer renting the robot arm (and software) to allay up front costs.

One of my interests was robotic visual systems (at a hobby level) which only requires a webcam. However it was just an interest as I never worked in the field. I used to manage a family citrus property and knew enough about electronics, AI and programming to believe I could have programmed a robotic arm with a camera to pick oranges. Back then it was too expensive and labour was cheap so it wasn't economically viable to spend any money on it.

Toy robots can be fun and teach something about electronics and programming hardware and maybe even be used to show "proof of concept" for some ideas like a visually guided robot system.  However when you want a practical robot it needs the precision, speed and strength to carry out some specialized task.

Posted by: @alextheprogrammer

I noticed that after hi kinda finished the robot's physical design and build, it all came to a halt, disappeared. The same happened with DB1 too! 

Alex,

I noticed the same thing when I started reading this site in 2021.  Our host worked out all the mechanical and electronic aspects for the ‘Real Robot’ but when it came to the control program progress came to a standstill.

For me that’s part of the appeal of working with robots, bringing together those three areas of a build.  Electronics is the tough one for me.  I like to think of myself as a systems analyst when it comes to the electronics.  I read enough to know what I need then I look for some vendor that has already manufactured the breakout board or whatever. I recently switched from using Arduino boards to the Raspberry Pi Pico and it’s been a positive experience.  Their C++/SDK is great.

I’ve never used AliExpress.  I’m in the US and I buy from Pololu Electronics, Sparkfun and Digi-Key.  It does add up and availability has become an issue recently but I prefer to give my $ to stateside companies. 

Good luck with your project.

Tom

@thrandell @robotbuilder It looks like a lot of things that have already been invented are already available out of the box as ROS packages. I bought that RPLidar for $62 on AliExpress and it arrives this month. It looks like combining Bill's ideas from DB1 + James Bruton's setup for navigation plus some of my ideas for the end effector that I'm already modeling in Fusion 360 might result in something. I have already made a fully 3D printed base for the DB1. It looks like I'd just need to publish the command velocity topic from that nav unit to receive it on some single board computer e.g. Jetson Nano that I already have along with RPLidar to achieve the navigation part and then switch to the end effector and AI. Will keep you posted here what the outcome is gonna be!