View Full Version : [Discussion] To brain, or not to brain

01-02-2010, 09:52 PM
So I've been going back and forth debating myself, and want to know what everyone else thinks.

My previous approach has been to stick a computer in everything, rovers, walkers, doesn't matter. heck I've been having dreams about a PC based humanoid. but when it comes down to it, I'm probably not doing much of anything that really requires a PC. Sure I have some lofty goals and all, but how much of that ever becomes reality given that my walkers can go at most an hour before their batteries are depleted.

Then there are some really cool micro controllers out there. The CM-510 for example...The SSC-32 coupled with the Bot Board II, or even by itself just to name a few...Heck people are even using cell phones now. All of these to various degrees are capable of doing almost anything that I have managed to do using a full on computer. Not to say that I am actually pushing the envelope there.

And then there are a series of wireless communication methods, which seam to be pushing us closer and closer to the point where we can take the brains...the big brains out of our computers, and just leave something that can relay data about what the robot is doing and feeling back to the mother that it no longer carries on its back. Is this even possible?

Matt posted a series of recordings of a lecture he put together. In which he talked about standardization in hobby robotics, being the next step, the same way standardization revolutionized computers. (There is a part of me that wonders if we aren't there all ready.) Maybe separating the brains from the robot are what we need in order to really take that next step. ( I hope I'm not misquoting...correct me if I am)

Who knows.


01-02-2010, 10:26 PM
There's two pieces to that puzzle. The requirement for realtime on the bot (such as for motor control), and there's no way to escape that. Then there's the computationally expensive stuff which can be routed off the bot. Two very different pieces of the puzzle.

01-02-2010, 10:26 PM
If you're building a walker, you really need the control logic to run on the robot locally. So, for example, with Roz, I could have decided to leave the Arbotix onboard, with NUKE running onboard IK. I could then use a PC with an Xbee to run the "brain", and have it transmit very simple commands to a simple program running on the Arbotix to control it by telling it to "move forwards at this speed", or to "turn at this rate".

In fact, I had something exactly like that running a couple weeks ago, so I could control Roz using a joystick. I wrote a simple little script in Squeak Smalltalk to read in the joystick info from the Windows API, and then sent commands to the Arbotix over the Xbee wireless serial link.

- Jon

01-03-2010, 06:31 PM
I agree with both Adrenalynn and JonHylands, some level of local control is needed. I'm a big fan of trying to get an autonomous bot that can surrender (some) control to a bas station, or relay data etc. to the station.
But I also believe a bot should be capable of doing something without always checking in with mama, even if it is just wandering around and getting in trouble.
Right now, the longer ranged radio systems require more power than the short-range ones (logical, eh?) Why use Bluetooth when you can get WiFi for the same power consumption? Currently, we can't, but wave bye-bye to the short-range stuff (at least for robots) when they can bring the power use way down.

What we need is a microcontroller capable of handling whatever local tasks are required, with some overhead for 'independent action', and cheap easy-to-use telemetry to connect with base. Well, to some extent, we have that now. Of course, we all want more power, more capability, at less cost.

Now, standards...
Ya gotta luv standards, there's so many to choose from...