Thymio II Educational Robot

[simplesi]

Wow
I really like the concepts behind this little device

https://aseba.wikidot.com/en:thymio

I particularly like the examples showing as a jack of all trades

https://aseba.wikidot.com/en:thymioexamples

I haven't looked into the innards yet but I feel an RPi based clone project coming on

Simon

[Ravenous]

Haven't looked at any detail but electronics wise it seems to be based on a single PIC processor. So a lot less processing power than a raspi, though it has a lot more sensors - IR reflective things for example. (I think the raspi would need a lot of analogue inputs adding to use the same arrangement.)

Can't tell what the motors are but they look like the fairly simple geared hobby motors a few of the hobby shops do. They seem to run from H-bridges made from discrete mosfet transistors or something.

I am guessing a raspi equivalent would need to be quite a bit bigger, for a bigger battery. But I've only glanced at it so far.

(Incidentally to save anyone hunting around I found a price, in France, 140-something euros, which sounds pretty reasonable to me for a finished article with that much sensing.)

[Ravenous]

They're well ahead of us - this example has the Thymio robot, with a Raspberry Pi as the brains, and a raspi camera too (and some Lego):

https://aseba.wikidot.com/en:thymioexplorer

The Thymio alread has a library allowing the raspi to control it. They think of everything!

[simplesi]

Even more wow

This implies 128 CHF == £86 delivered

If the central processing hardware is cheap then not worth re-designing around an RPi and just stick with their concept of using the RPi as a wireless programable controller (with a special version of ScratchGPIO of course

However, it may be worthwhile replacing their controller directly and just using same form factor/sensor concepts.

I need to work out how to get one in my hands for a bit of reverse engineering (I mean assessment for learning in schools)

Simon

[Ravenous]

It looks like it has a USB interface and the raspi controls it by that.

The whole thing is well integrated; probably not worth trying to rip bits off from the schematics. For example I think all the infra-red collision sensors are analogue and need an ADC channel each (most PIC controllers have simple ones coming out of their ears.)

By the way for robot fans the main guy behind this has been doing swarm robotics for years:
http://stephane.magnenat.net/articles.html
Lots of unique little robot designs on there (several too small for a raspi). Also some of the more recent papers are about using the current robot with younger children. Academic research oriented but still probably interesting.

EDIT: the 2004 paper is one I first saw years ago - little one-armed robots, several can join together to bridge large gaps. No idea how much it would have cost in those days, but I'd have liked like a small army of those!

[mondada]

Hi Guys,

I am the leader of the Thymio II project. Some comments:
- we have hundreds of Thymios in primary schools and baccalaureate schools, and we hope to provide more and more educational material with Thymio II. None is connected with RPi until now, but some experiments are in progress
- the controller costs few dollars, not sure it make sense to remove it.
- yes, all is controlled through USB, there is already a software package to control all sensors and actuators, exchange events, and even connect to ROS for the more advanced roboticists.
- everything is open hardware and open software, feel free to have a look !

[simplesi]

Hi
I'm the author of ScratchGPIO

I am very keen to get a low cost wireless Scratch controlled robot into our schools.

Yours is by far and away the best candidate in terms of form factor and sensors and value for money that I've seen so far.

But I think you could be right in that it would be better to design a case for a Raspberry Pi to include a battery and just mate the two together.

I estimate the cost of provding a Raspberry cased add-on to your Robot to be about £40 (Rpi, case, SD card, WiFI) not including battery.

If there was just one case that included everthing then we could save £5-10? And then save £5-10 on your controller??

I think I need to build a non cased prototype with the same features as yours and see if the overall savings would be worthwhile.

But I need to persuade one of my schools to buy one of yours to test out

I think you have built a very very very good device and I thank you for making it open source

regards
Simon

[mondada]

Making a specific case for both Thymio and raspberry Pi is of course possible, but is a big investment (molds for plastic injection) and makes sense only if there is a large sale behind. Do you feel that there is a real market behind?

[Ravenous]

I doubt there is a large enough market for Raspi - specific changes to the Thymio. As we've seen in the past for the Raspi, adding a special version for a different market tends to add a lot to the finished cost.

People who are interested can - as in the Examples page - add a raspi (or other small computer) with its own battery, especially if a few Lego blocks are glued underneath in the right place to lock the assembly onto the Thymio. Other sensors like distance measurement, magnetometers, etc. could be added this way too if necessary.

Also some people would want the raspi to be physically mounted on a top deck, where the thymio's touch pads are. Other people would want the raspi mounted below with a heavier battery, and the Thymio still driving the whole machine by Lego gears attached to its wheels (there are examples of this in the Examples too.) So there would be several competing mechanical designs to pick from...

The more advanced hobbyists can often get access to a 3D printer these days too.

[simplesi]

I want to see a market for £50 -> £100 educational robots programmed using a block programming language like Scratch (preferably in Scratch)

Currently a Raspberry PI based robot will cost £60 but would have to be built up and no-where near as versatile and child friendly/robust as the Thymio

I can see that a lot of thinking and development work has gone into the design and features for the Thymio and I think it might be worth making a Raspberry Pi based version of it.

An alternative could be to use the Thymio directly but with a bluetooth module and to remotely program it using a Scratch intermediate program on a PC/Laptop.

A Raspberry Pi joint case might be the cheapest /easiest solution as it could still have the top sensors.

Simon

[Ravenous]

A Raspberry Pi joint case might be the cheapest /easiest solution as it could still have the top sensors.

Probably more to it than just making an extended top - the capacitive touch switches, LEDs and so on would need extending from the main circuit board to reach the raised top.

The cost of the optional top would be very significant - designing and making injection moulds is a big one-off cost apparently (I'm no expert).

Also an integrated battery pack to support the raspi would need to be designed - and possibly another USB port for the raspi, etc.

In addition Thymio's visual version of its language looks really neat - I had a look at the tutorial over the weekend, without actually trying it I think it's pretty well designed even for quite young kids. (Dare I say, I found it easier to understand it than Scratch having skimmed very briefly through both languages. Not that a mid-forties bloke with two degrees is their target market! )

[simplesi]

Yes - the visual programming enviroment is easier than Scratch and would be useful for younger pupils.

But I'd like a bit more for older pupils and since Scratch is what they will be learning to use for normal Computing - it makes sense to use it for Robotics as well.

The Bluetooth option could be a good compromise but I've not experienced trying to use mutiple Bluetooth (I envisage about 15 bots in use simultaneously) devices in a classroom before so I don't know how that would work in practice.

I definitely need to get one into my hands to play with

Simon

[mikerr]

The more advanced hobbyists can often get access to a 3D printer these days too.

Puts hand up - yep, 3d printer owner here - can be had for < £500 now.

Just put the design up on somewhere like thingiverse.com and we can all print our own.

[mondada]

I want to see a market for £50 -> £100 educational robots programmed using a block programming language like Scratch (preferably in Scratch)

There is an ongoing discussion on porting Scratch on Thymio, and ongoing developments to have a wireless Thymio....

Yes - the visual programming enviroment is easier than Scratch and would be useful for younger pupils.

But I'd like a bit more for older pupils and since Scratch is what they will be learning to use for normal Computing - it makes sense to use it for Robotics as well.

Scratch could be indeed a nice intermediate programming environment between VPL for pupils and the text programming environment of Aseba Studio.

Just put the design up on somewhere like thingiverse.com and we can all print our own.

The full 3D model of Thymio is available at https://aseba.wikidot.com/thymiohardwaresource
First parts for 3D printing are under https://tinkercad.com/users/ewJZs0SwLJA-thymio3d
Example of printed wheels are under https://aseba.wikidot.com/it:thymio3dresults

[simplesi]

There is an ongoing discussion on porting Scratch on Thymio, and ongoing developments to have a wireless Thymio....

Scratch is already there on the RPi itself and/or Scratch is also being used on PC/Mac machines to control an Arduino based robots via Bluetooth

http://www.youtube.com/watch?v=e-yL9zIyY90

So simple addition of a low-cost Bluetooth module to the Thymio that could be used to pass commands could make it very quickly available to be wirelessly controlled via Scratch

I MUST get one of my schools to buy one

Simon

[simplesi]

Just ordered one

Can't wait to try it out

@mondada - It will show up as an order from Joanna Walters

Simon

[mondada]

Enjoy

[meltwater]

Regarding bluetooth.
If the controller has serial pins, then a bluetooth module would be easy to add for around £7. The firmware would just need to respond to serial commands, as it would to USB commands.

Alternatively, a small Atmel chip may well be able to sit in the middle to convert serial commands to suitable USB commands (not quite as easy though). Although, the IR link may be a simpler method, wouldn't take much for a device to take serial output from a bluetooth module and convert it to the IR signals expected by the robot.

Q: Would the controller board be available to purchase separately?

[mondada]

Regarding bluetooth.
If the controller has serial pins, then a bluetooth module would be easy to add for around £7. The firmware would just need to respond to serial commands, as it would to USB commands.

£7 is extremely expensive, for this product we need to be within the few $ for this type of functionality. And certifications are a real and costly problem....

Regarding bluetooth.
Although, the IR link may be a simpler method, wouldn't take much for a device to take serial output from a bluetooth module and convert it to the IR signals expected by the robot.

IR comm among robots is already implemented and included in the firmware, but the bandwidth is much lower!

Q: Would the controller board be available to purchase separately?

No, sorry. Would probably cost (in small qualities because of the market) the same price as the robot

[meltwater]

That is a prebuilt bluetooth serial module, so cost price would probably be much lower if built independently. However, the certifications would be an issue (which I think is why the RPi did well to avoid built in wifi and bluetooth - despite broadcom being market leaders in them).

Depending on if you directly wire into the IR line (I guess it is IrDa - although didn't spot the IR transmitter line in the hardware), the data rate could be increased if not going through the IR sensor. Although it could be made as clip on module instead. For basic control, the data rate might not be a huge issue.

If one set of the serial pin connections were available (appears to be 4 uarts on the PIC), then it would just be a serial bluetooth module that is required (within the firmware I imagine it would be possible to send the data to the same handler for the IrDa). Standard Bluetooth modules, sourced and sold separately, would probably avoid the certifications. Bluetooth would add support for control via Android as well as the Rpi and wireless control for PCs too. For apple support you'd need BLE modules (which I've not looked at so far - they don't seem to be as easy to get and in Apple tradition, not as open support as Android).

Hopefully if the market expands, then boards will be made available too, as they would be very useful (not least by presenting a standard platform for controlling things). Would make a nice RPi addon.

[simplesi]

@mondada
Is it in the post yet?

On the plane to Manchester tonight?

Simon

[mondada]

That is a prebuilt bluetooth serial module, so cost price would probably be much lower if built independently. However, the certifications would be an issue (which I think is why the RPi did well to avoid built in wifi and bluetooth - despite broadcom being market leaders in them).

Depending on if you directly wire into the IR line (I guess it is IrDa - although didn't spot the IR transmitter line in the hardware), the data rate could be increased if not going through the IR sensor. Although it could be made as clip on module instead. For basic control, the data rate might not be a huge issue.

If one set of the serial pin connections were available (appears to be 4 uarts on the PIC), then it would just be a serial bluetooth module that is required (within the firmware I imagine it would be possible to send the data to the same handler for the IrDa). Standard Bluetooth modules, sourced and sold separately, would probably avoid the certifications. Bluetooth would add support for control via Android as well as the Rpi and wireless control for PCs too. For apple support you'd need BLE modules (which I've not looked at so far - they don't seem to be as easy to get and in Apple tradition, not as open support as Android).

Hopefully if the market expands, then boards will be made available too, as they would be very useful (not least by presenting a standard platform for controlling things). Would make a nice RPi addon.

BT is expensive and complex: either you include a full module and this costs a lot, or you need to certify and this is complex and costy. We are studying this in detail since long time and have already a working solution.

The IR solution we have is not mainly for communication, is based on the existing IR sensors and can allow to exchange data at very low bitrate, but detect the source orientation and have an idea of the distance.

Concerning the board you are right, we will see how this evolves

[meltwater]

Yep there are lots of ways to do it, each with + & -. Cost is always a killer, since as soon as you price it out of the market there is little point adding great features. So totally understand.

The IR stuff sounds interesting, they are quite useful in that sense for proximity and detection. Via analogue inputs quite a rich source of information. Leveraging extra information out of basic sensors is a neat way to make the most of the hardware.

[Ravenous]

The IR stuff sounds interesting, they are quite useful in that sense for proximity and detection. Via analogue inputs quite a rich source of information. Leveraging extra information out of basic sensors is a neat way to make the most of the hardware.

Don't forget they are also used for low bandwidth communication between other bots - classic swarmbot stuff - maybe play search party or something like that...

https://aseba.wikidot.com/en:thymioapi see "Local communication"

[simplesi]

Its here!

https://vine.co/v/MuIIizJLObr