Slashdot Mirror

This is great and all, I just don't see why this is front page news. I would consider ARM-DSP hardware with Linux support mainstream rather than a bold step taken by TI. A random grab from sponsored adds on google:
http://www.sandorlabs.com/
http://www.compulab.co.il/
http://www.plexxa.com/
http://www.atmel.com/products/AT91/
http://www.xbow.com/
http://www.lynuxworks.com/

All products seem to support Linux on ARM/XScale and (at least) some in combination with a DSP.
Sure, Texas instruments is a heavyweight in the embedded world, but is this just another clueless ScuttleMonkey post or did I miss something?

Mesh networking is nothing like cellular networks. In mesh networks each node participates in the routing process, relaying packets intended for other receivers. Of course this is not the case with traditional cellular networks where routing takes place only in the based stations and the core network connecting the base stations. Furthermore, mesh (also called ad-hoc) networks are self-configurable and self-healing. You throw the nodes over an area and they themselves discover who their neighbors are, discover routes to other nodes in the network using distributed on demand or proactive routing protocols, and if a link fails they can automatically reconfigure their routing tables. Since nodes relay packets destined to other nodes, the range of the covered area can increase with the number of appropriately located devices, unlike cellular networks where the range is solely determined by the base station (BS) and phones' antenna transmission range. There is virtually no single point of failure as is the BS for traditional cell networks. Issues like hand-over are usually handled by the mesh network gateways. The defining steps on the subject were done by researchers at UC berkeley and xbow http://www.xbow.com/Products/Wireless_Sensor_Netwo rks.htm. Another pioneering company in the field of mesh networking is Embernet www.embernet.com, these guys developed commercial h/w and s/w for this purpose more than 3 years ago.

I think I saw the same thing on a Discovery Channel special. I believe it might have been these guys http://www.xbow.com/ but I could only find information on it in a newletter http://www.xbow.com/General_info/Info_pdf_files/Xb owNewsletter_Q1-05.pdf

I think I saw the same thing on a Discovery Channel special. I believe it might have been these guys http://www.xbow.com/ but I could only find information on it in a newletter http://www.xbow.com/General_info/Info_pdf_files/Xb owNewsletter_Q1-05.pdf

Currently I'm working with Crossbow motes. Some are the size of about 4 AA batteries stacked 2x2. Others are the footprint of a quarter and half an inch tall. Crossbow seems to have the best prices on development kits and individual motes. Ember, Smart Dust, etc. are much more expensive.

Sensor Network research has been going on for quite some time. CENS isn't the only place doing the research. Some of the original work actually came from SCADDS from USC/ISI, with some of the same people now at CENS. All the UCs are very involved in Sensor Networks, Berkeley for example was the orginator of the current most popular hardware, the motes, now manufactured by Crossbow. Intel also makes their own version of the motes, though they are not sold comertially yet.

Most of the hardware runs in a specific OS called TinyOS, which is open source. Other hardware, like the Stargates (also from Crossbow) run an arm version of debian.

You can find lots of neat info about Sensor Networks from the specialized conferences like Sensys, IPSN, etc.

Most of the projects done with Sensor Networks have been geared towards the biological monitoring fields. However, the funding comes in from NSF as well as DARPA so sometimes it is security/military based.

It's an interesting world out there...

cl

Check out the MIT Cricket Indoor Location System. http://cricket.csail.mit.edu/.

It is commercially available from Crossbox Technologies http://www.xbow.com/Products/productsdetails.aspx? sid=116

I've got a number of zigbee products in my lab for sensor net research. You can get them...

http://www.xbow.com/Products/productsdetails.aspx? sid=101
http://www.moteiv.com/

Wrong. Buy a small INS here. There are standard units that contain three accelerometers and three rate gyros (one for each axis), which is what you need. They're getting smaller; 1 cubic inch units with all six sensors are available, and a single-chip version has been prototyped.

Most serious robotics projects today have one of these. They're not good enough for navigation by themselves, but they can provide attitude info just fine.

The basic way you figure out "down" is by using the accelerometers for the long-term component and the rate gyros for short-term corrections. You do lose accurate "down" if you go round and round in a circle for a while. Some aircraft artificial horizons have that problem.

In addition to research-grade and hobbyist software and hardware, some companies are beginning to sell more user-friendly stuff. Sensicast Systems in particular offers a couple of software products for Crossbow MICA2 sensor boards. Of particular interest is the Development System software which allows a user with Xbow nodes to configure them and aggregate data from them into a database -- great for people just wanted to try to get their own sensor network running. They also have their own "H900" wireless sensor net system that automatically does temperature and humidity sensing or pressure sensing.

I saw Berkeley and Intel also present on this technology at O'Reilly's Emerging Technology Conference in 2003. The presentation synopsis is here, although the presentation sadly is not:

http://conferences.oreillynet.com/cs/et2003/view/e _sess/3797

They are doing amazing sci-fi type stuff with their Motes already, it was a pretty amazing presentation, touching on swarm behavior, conspiracy theories, technical deployment issues, and just plain good-old fun hackery. The wired article really should have mentioned that serious hobbyists can purchase a mote starter kit and other stuff here:

http://www.xbow.com/Products/Wireless_Sensor_Netwo rks.htm

Note that there is a classroom starter kit. I would think this sort of stuff would get high-schoolers really excited about science. A great stocking stuffer for your local high-school (although at $1,000 or more maybe a little out of my budget).

Imagine a Beowulf cluster of..... never mind. :)

I'm not sure if anything short of Pentium/Athlon would be state of the art enough, but I'd recommend the Atmel AVR. Firstly, there are lots of demo/examples with it and existing software archives exist.
This processor is used in some smart-sensor applications where you have distributed sensors.
Here's a 1998 EDN mag review and some simm circuit boards which make project computers.

Since I'm sure most of /. is more interested in coding a 1 square inch sensor than protecting a 300 foot tree, here's some programming background on the little bastards (which I work with on a daily basis, as part of a sensor network research group in a VA university).

- the architecture

The motes run 4MHz or 8MHz processors, with built in memory. The amount of memory varies across mote models (currently Rene, Rene2, Mica, Mica2, Mica2Dot, and SmartDust) but we're talking 16KB to 128KB of program memory, 4KB to 16KB of data memory, and 4Kb to 8KB EEPROM for permanent storage. They have a short range radio capable of I believe 10kbps, and use an active message model to provide what we know as "ports", so that you can direct a message to a specific handler based on its message type. The packet sizes top out at 36 bytes. The motes are powered by two AA batteries, which can last a surprisingly long time if the radio is put to sleep. Your main means for debugging: 3 LEDs ... you can begin to imagine the headaches I face on a daily basis.

- the bridge

When deployed, most motes are programmed with routing protocols to autonomously establish networks, which are used for data aggregation and getting sensor readings around. The network is rooted at a basestation, a "powerful" PC without the restricted computation, communication and power limitations of a mote. This way any complex processing is offloaded to the PC, and the motes don't waste battery power doing stuff the PC can do instead. So what bridges this mote network to a PC? Well, it's a programming board. You plug a mote directly into the thing, and you hook up a db-25 to your parallel port, and a db-9 to your serial port. The parallel port is used to program the mote's instruction memory, and the serial port is used to receive messages sent by the mote to the PC. The mote that's hooked up to the programming board is loaded with code to translate RF packets to UART, and vice versa.

- sensing

Motes are equipped with 10-bit resolution ADC sensors which can read light and temperature. Other sensor boards can be hooked up to motes to read vibration, acceleration, and a bunch of other stuff. The motes commonly read their sensors, stuff the data in a packet, and send it along to the basestation for processing. That's the generic application model, at least.

- security

The main part of our research deals directly with implementing security in the sensor networks. This is far from easy, since you can't even store a public/private key in the mote's limited memory, let alone do anything with it. The protocols used are complex, involving securely distributing keys, efficient authentication protocols, and all this in 16KB of program memory (on Rene2s) INCLUDING the operating system! Just remember that the point isn't to stop a mote from being compromised, it's to realize it's compromised and drop it from the network. There are supposed to be thousands of motes in the network after all, so dropping a bunch won't hurt.

---

Here's hoping that background will help avoid the mass privacy paranoia that we /. readers love so much. At the time of this writing, motes aren't small enough or cheap ($250) enough to produce en masse, nor are they tiny enough to go unnoticed (remember the 2 AA batteries?). Yes, there are exceptions, but 1 square inch are the smallest production versions I know of (Mica2dots). And until they stop running on batteries, their biggest hindrance is their short lifetime, so they currently can't be constantly monitoring anything for months on end.

Aside: Take a look at the Spec. It could change that whole last paragraph. :)

As for the military surveillance stuff, that's what motes are ultimately designed for, to be dropped on

These tests were down with MICA sensor motes which can be purchased from Crossbow Technology.

These motes run TinyOS, which was developed at UC Berkeley.

More information about TinyOS:

• TinyOS FAQ
• TinyOS support page
• TinyOS SourceForge page
• nesC Compiler SourceForge page
• MICA sensor mote hardware designs

Yes, my job does involve programming for these motes. I have four of them on my desk acting as an ad-hoc wireless sensor network now.

These tests were down with MICA sensor motes which can be purchased from Crossbow Technology.

These motes run TinyOS, which was developed at UC Berkeley.

More information about TinyOS:

• TinyOS FAQ
• TinyOS support page
• TinyOS SourceForge page
• nesC Compiler SourceForge page
• MICA sensor mote hardware designs

Yes, my job does involve programming for these motes. I have four of them on my desk acting as an ad-hoc wireless sensor network now.

• So far it's just hype, but it looks like they're going to show at E3 next spring, so we'll get to see it then.
• This thing probably isn't any smarter than an Aibo. It seems to have about the same sensor suite, and about the same claims of intelligence. But if you've ever played with an Aibo for any length of time, you realize how pathetic its software is. It has trouble locating its own red ball, and it can't deal with a table edge.
• I suspect that it doesn't really balance, either. It has huge feet, and is probably a statically stable walker, like a windup toy. The description says "The robot's innovative mechanical design will allow it to get back into walking position even in the event of a stumble effortlessly". This probably means a canned righting sequence, like an Aibo. But if they really put in the accelerometers and gyros needed to do it right, that's a major advance. (You can buy that stuff now, but it's a bit too big. But see a newer one. Millions of rate gyros are going into auto anti-skid systems.) If it has those, it's reprogrammable, and the motors have enough power that running is possible, I'll buy one for my own work.