Home Built Qrio?

[frank]

New to the forum, and new to this form of robotics.

If you can build a robo-one type robot from off the shelf parts similar to a robotis bioloid, then why can we not build something "smarter" like the Sony Qrio?

There are three major parts to a robot like this:

Mechanics- Getting the correct range of motion. It's already been done, so it's just a matter of showing the "formula" so we can all see it.

Computer hardware- what kind of available electronics can we purchase to hold the information?

Programming- Once it is written there should be much of a need to change it, just add to it. (like adding facial recognition later down the road).

Here is an idea: Why don't we work together and source out parts and pieces to build a working frame that has the range of motion the same as a Qrio. After all it's just servo's and joints. It's the fluid motion that we are not seeing in home built models and this can be overcome. We should focus on an opensource way of building this. Maybe we can call it a robotic version of Linux?

Just a thought.....

[limor]

Hi frank and welcome.

The main problem is not really the range of motion. If you take any of the humanoids on this site, they have enough range of motion to look human if there were actuated properly.
The main problem is control. Qurio has allegedly several cpus running at hundreds of mhz and many many sensors. The humanoids here have a microcontroler that runs at less than 20mhz.
Motion fluidity using open loop control is very difficult because the robot is hardly "aware" of what's going on around him.

The way to resolve this is to add more sensors and more CPU power and update the servos positions at least 50 times/sec based on whatever the sensors are telling and whatever interpretation is given to the sensor readings.
A gumstix attached to the microconroller can take you a long way in CPU cycles/sec of on-board brainpower.

[Chemical]

Probably everyone here could make something like a qrio.... if it wasn't for a little thing called money

With alot of work you could buy the parts, build, mold, program something very similar to QRIO with amazing technolodgy, but it costs money, alot of it. And thats why you don't see anything like QRIO in homebuilt, the same reason school kids make Go Karts and not sports cars.

[heke]

I think we haven't seen yet what Kondo-robots, Bioloid and others can really do. Same thing with Commodore 64 long time ago: at first it was only a blue screen (a good thing at the time), but in the end we saw jaw-dropping demos.

Of course our simple robots can not reach the skills of Qrio ever. But little by little we can see more smooth motions when the programmers brush up their programming skills and when they learn more about their simple robots.

A dual-core laptop is nowadays sold in every super-market. Sooner or later a Qrio-style robot will be sold similar way. But time for that is not yet.

[RamDragon]

I doubt it'll be any time soon that we see a Qrio style robot for sale on Wal-Mart shelves. Hell, since Sony killed their whole robot program, it'll be even longer.

Still, the autonomous functions it had makes you wonder what you could do with a few hundred a year to toy around with. Wireless, self-charging dock, learning abilities, audio commands, visual recognition... it's easy to see why the dream. Still, most of it's software, and I'm not sure that networked controllers isn't the way to go to achieve the smooth fluid motion, which dictates a larger frame anyway.

I mean, it's all fine to want to build within the Robo-1 guidelines, but I'm really waiting for someone to build a 20 foot tall robot to go push over the Neogentronyx heap. I'd do it myself, but I misplaced my $100,000 bill.

[JonHylands]

I think its doable (building something like Qrio in a RoboOne size), but you need to offload the serious processing to a PC. Put embedded controllers onboard to handle high-frequency motion control, and feed that over a high-bandwidth Wifi link from a PC. Send wireless video from the robot to the PC, and run something like OpenCV or RoboRealm or Mavis or whatever for vision processing.

- Jon

[Humanoido]

I agree that we're seeing just the beginning (actually second beginning) of humanoid robotics. I had looked at scaling up several of my robots and encountered several obstacles - money, parts, and a small lightweight power source.

Already we are seeing some improvements in software using the existing humanoid platforms, making the motions more smooth. One of my priorities is getting my robots to walk smoother and make them more human-like.

Processing power is another thing. I've waited 50 years, and still, the processors are, well, not powerful enough. I'm talking about handling intelligence and all the humanoid functions. At least we can use small microcontrollers and connect them together on our current budgets. I think the Federal Government should give us free grant money to continue our humanoid research and projects.

Evolution... have patience, we'll get there.
Humanoido

[wthierry]

Im rather new to all this, but wouldnt the easiest way to build a QRIO be to use one of the existing biped robots add some sensors and then link to a computer via 802.11 which processes and tells the robot what to do and how to react. This way your robot can be controlled by a fast machine but have low overhead on the robot. This is my first post, so please dont flame me if im being too simplistic.

[Humanoido]

You could probably get a hold of a Qrio and duplicate it.

humanoido

[Layton]

I'm fairly good at 3d modeling and I'm considering making a 3d version of QRIO that could then be printed using rapid-prototyping printers. My problem is that if I am going to build it in 3d I need to first look at a base robot like the KHR-1HV and see how it can be modified to better fit the proportions of QRIO. I'm not even talking about things like voice recognition or smooth motions. I just want to build a small robot that can walk and move around with a sleek exterior similar to QRIOs. Maybe include some LEDs in the head and chest and a small camera so that it can output images/video to an external pc. But I am a complete n00b at the robotics stuff so that will have to wait awhile.

In the meantime, maybe I will get to work modeling QRIO. He is a complex and smartly-engineered little guy, with lots of interconnecting parts and details. But I think its doable, I've seen a decent version of him on one of those pay-to-download 3d model file sites, but since I can do 3D I see no point in paying the $150 charge for it. Whenever I get around to completing it I will share the data. Once printed, I could even create molds so that others could have copies of the parts. A little silver paint, and...

[Robo1]

Alright layton

Have a look at this this. The only problem you will have is how do you drive these joints. its all well and good having a 3d model of the robot but have you thought about how you will run the cables through the joints (slip ring). How you will drive these joints so that you get 3DOF around the same point (pulls). those are the difficult part.

just my thoughts. To help you out theres a model of a RN1 and KHR knocking around this site and over at tribotix theres files for the most servos.

Good luck Bren

[Layton]

Hey Bren, looks like you were onto something but the posts date from last year. Whatever happened with your build? Was it just too difficult to do or was it more a factor of time? It looked like a cool custom bot in your render, a little like ZMP's Nuvo.

As far as the servo cables go I don't think there is much issue, assuming the parts are hollow enough there should be room for them. I come from a game art background and not an engineering one so I'm more concerned about how I'll manage to get all the parts to fit together nicely and attach to the servos. But with an accurate 3d model of the base parts and enough elbow grease it shouldn't be impossible (though perhaps improbable).

On the plus side, if I can't get things to work at least I'd have the option of printing a cool little QRIO figure model.

[Robo1]

I'm still working on the bot if you I took it to the UK robo-one meet the offer week. working on adding pressure sensors for the feet to try out some ZMP measurements.

Bren

[Layton]

Ok so I started with a QRIO schematic to sort out the number and placement of servos. I figure 28 servos total (the hands will use 1 servo each and have a basic gripping ability).

I plan on using the RB2000 as a base and will use higher torque servos in the shoulders and legs (the RB2000 kit I'm looking at comes with 19 servos, so the other 9 I will need to order as a pack, might as well get the more powerful ones).

The nice thing about the RB2000 is it has a speaker and 16 ch LED support, which should be enough to do the lights in his eyes, ears, and chest without any multiplexing.

I haven't started the modeling process yet, but based on my rough measurements the robot will be about 40cm tall (or 2:3 scale of the actual QRIO).

One concern I have is that the shoulder rotations, if limited to 180 degrees, won't offer the same fluidity as the real QRIO. I'm still keen to go ahead with the project, but ideally servos with 270 degrees would be used for the shoulders to give him a nice broad range of motion in them (if you watch videos of QRIO getting up from a fall, both on his back or on his stomach, you will see about 270 degrees of rotation there). 180 degrees is fine for the rest of the joints though.

So if anyone knows anything about 270 degree digital servos, post about it please. Oh, and Bren if you feel like it post up some photos of your custom bot. I'd love to check him out.

[Robo1]

Hi Have a look at here. Haven't uploaded any more pictures for this as I'm in the process of installing a screen for the head.

Just a thought about using the RB2000, can you get the servo positions from the servos I remember this being the problem with the original RB1800. would love to have a play with the QRIO files once you have them up and running in CAD could you post them in STEP or offer files.

Bren