New Controller for KHR2/KHR1-HV/Manoi

[shsan]

This week-end here in japan a company named Ixs presented a new board.

http://www.ixs.co.jp/products/controlle ... ard-j.html

I was in Akihabara by chance and learned that there was a presentation today too (not only yesterday) so I went to check out what it was all about at the Kondo robot spot.

Well the board is really interesting and might solve a lot of issues we have with the RCB3.
8 Analog channels, 24 PWM, (some digital channels also available too but not directly a bit of soldering is needed).
Linux on board with a 200MHz ARM7 with 32MB of ram and USB support.

That would solve all the issues we have with the slow feedback I guess. Since quite a lot of the "Intelligence" could be put on the board itself.

Since Ixs is the main provider of KHRs oversea, I would not be surprised to see Robosavvy selling that board

It's definitively something to follow closely, I still have to find out exactly what kind of software it's coming with (well except the Linux). I know they provide a library to access the servos etc..
But I am afraid my japanese is not good enough to get a lot more information.

(And I was "ambushed" by a local TV crew too, lol I hate beeing filmed

shsan.

[limor]

looks impressive

Did they show off any special performance of the KHR/Manoi thanks to the board?

[shsan]

Not really. The board was mounted on the back of the AT01 and we talked mainly about technical aspects.

Also what do you mean by performance?

There was no live "demo" while I was there. And honestly I was more interested in the details than looking at new moves.

But the other board that they have (same site RM-EYE product) is targeted at Soccer since it's a bit like a CMU board to do color recognition.

I think the main improvements will probably be in autonomous behaviour.
With now enough power to do quite complex things, I would not be surprised to see someone hook a webcam (possible driver problems here) and do some nice shape recognition or something like that.

Also having now 8 analog input (and possibility for even more if you hack it a bit with some multiplexers), you can imagine great progress.
Putting 3 or 4 Infrared sensors around the robot plus the 2 gyros and the accelerometers would certainly enable people to do quite a lot in terms of collision prevision etc..
Or you can easily imagine putting pressure sensors on the feet to enable better steps climbing etc..
Or all of the above and more.

For RoboOne competition, you can imagine starting to see complex "semi-autonomous" motions, (not that it was not possible before) but I think it will now be open to more people. Since the hardware part is done.

Also the kind of scenario or motion that you can make with this board is now almost unconstrained.

32MB of RAM... and a USB port. Just plug a 4GB stick and you have memory to do a scenario that would last 10 years

And since it's a linux and gcc open platform the possibility are like for the bioloid kit only limited by what people can program.

I guess we will have to wait and see what we(roboteers) can come up with.

mmm I typed to much for a simple question

[limor]

I agree with what you say about the possibilities expanding by having more computation power onboard the robot and more sensory input.

My question was if they actually showed something "amazing" that could not have been done before with a more simple controler board.

I've been playing a lot with Gumstix and robostix on the bioloid for a while which gives 400mhz to the robot. indeed it opens up a new way of dealing with the humanoid -- updating the direction and power of the motors 100times/sec based on loads of sensors.
that's where the future of hobby humanoids is heading.