Re: BT in robotics

At 09:03 AM 6/3/2005, Bruce Hopkins wrote:
> 2. How else (other than Bluetooth) could you send data from your
> cellphone to your robot? IR isn't an ideal solution, since a lot of
> phones don't support it.

Use the data capability of the phone to set up their equivalent of an IP 
connection.

On my TMobile phone this is painfully slow, under 100 kbits at best, often 
less than 50 kbits, even when literally next door to a cell. I did some 
tests with a borrowed PCMCIA data modem at a TMobile sales shop, and that 
convinced me not to buy it. Plus the charge per kbyte is still outrageous.

> 3. I created the JB-22 kit, so yes I use it. I use it alot.  It's a
> very cheap kit that allows anybody to get started with Bluetooth
> development.
>
> 4. When I get a chance, I'll try to do some real tests for you
> regarding pure data tx times. The 2-3 second time frame that I gave
> you is the time it takes for my java class to startup, get the cached
> connection parameters, create the connection, create a session (which
> is optional), read a file from the hard drive, send the data, close
> the connection and session, and return. In 2-3 seconds, in my opinion,
> that's pretty impressive. And of course, I used the JB-22 kit for all
> this.

So where I have always hit a brick wall is getting some pluggable BT module 
with a SPI or serial interface to add to an embedded system. Just like 
we've done with Maxstream and Linx RF modems. These don't seem to exist for 
BT and I cannot understand why. These need to be under $100 each to be 
practical.

The Maxstream modems we use do 38 kbits with an actual range of a city 
block (1/8 mile) in our downtown office. They are under $100. Others have 
range of 1 km or better for under $200. And you can really buy them and 
they are easy to use.

> 5. Bluetooth does has a 7-node limit for piconets. However, you can
> overcome this limitation by chaining your piconets together, which is
> called a scatternet.

So does each piconet have a different channel then? How would this work in 
a swarm of 20 robots for example, all on the same stage or in the same lab?

> 6. Bluetooth has lot's advantages (trust me, I can go on and on) for
> instance, you can run TCP/IP over the Bluetooth protocol. Let that
> sink in for a second. That means that you can have all your wireless
> Bluetooth devices communicate to each other with an IP address, and
> you'll have the full TCP/IP stack to play with. Your networked
> application doesn't even have to know that all the nodes are wireless
> and doesn't have to care about communicating to the nodes using the
> Bluetooth semantics.

We'd like to run JXTA on top of everything as the "nuetralizing" app layer. 
It runs on any wireless platform and any wired TCP/IP platform.

> I'm also a Java fanatic, and Psinaptic makes a Jini implementation for
> the CSR BlueCore2 Bluetooth chips. That means that you can have your
> Bluetooth devices form a Bluetooth network, and then overlay it with a
> Jini network. Very, very cool.
>
> http://www.psinaptic.com/blue_core.jsp

Yes I know the folks at Psinaptic through our work with Dallas TINI, and 
JavaOne over the years.

Will you be at JavaOne? I'm there all week and giving TS-1464 on Thu.

> 7. Yes, we definitely need to collaborate, because I can see some
> really cool applications coming on the Bluetooth side of things.


OK then 2 questions

1) have you looked at running JXTA on top of your dev kits? Then you have 
seamless integration with JXTA on wired and other wireless node. Any JXTA 
node can discover and use services from any other node. It's designed to 
scale into the billions of nodes. So robots all over the planet could 
communicate and share data and services with any other JXTA nodes which 
could be PCs, PDAs or cell phones.

2) could your dev kit be adapted to a non-USB interface for embedded system 
use? Or is there another BT module which could be used? So far everyone 
I've seen and tried to buy is a dead end: either a $gazillion license 
needed, $5000 dev kit needed, etc. It's been very disheartening.

(On the other hand several under-$500 Zigbee kits are available today - we 
have the Freescale one, as are many other good RF modem dev kits such as 
Maxstream). Why do BY vendors make this so hard? Maybe you know the reason 
for such high barriers to entry. Is there any hope for this changing?

Regards

Bruce