Sensor Area Networks (SANs)

[davidbrady]

Folks,

For the time being I'm going to set configuring WaGu aside so I can get back to our Sensor Area Network.

I'd like your opinion on a few things.

First, I think scattering arduinos throughout the bus may not be the best solution. Too many arduinos creates too many points of failure. Plus I'm not convinced the arduino with it's associated sensor hardware is up to the commercial/industrial environments our bus SAN requires. Also, the arduino doesn't seem to have the computing power required forcing us to eventually hang them off a Rasberry Pi or some other full purpose high-power computing environment.

To solve some of these issues I started to look around at what the robotics world and the automated factory world is using.

I stumbled across an interesting architecture pioneered by a gentleman named Saul Greenberg from the University of Calgary, Wiki.

His architecture may suit our needs better. Basically he takes a bunch of sensors, aggregates them, and connects them to a computing platform via USB. The computing platform can be any type of machine running a host of real operating systems. It can even be a Rasberry Pi. The rPi can be located anywhere and is powerful enough to communicate via the full TCP/IP stack with any other host, local or remote. This means ostensibly you could control your bus with your iPad (over iOS) from your homebase living room sofa, many miles away.

The idea would be a set of sensors implemented using Phidgets connected to a rPi communicating wirelessly to a GUI somewhere (iPhone, iPad, Android, Laptop, Dashtop).

Check out these links: Phidgets, Robotshop, DSP Robotics, Linux Support.

[DOSZORROS]

David:

I am far from an electrical genius (spell dumb) but I am curious where you are headed with this and what your ultimate goal is? What systems do you want to monitor or control?

Thanks in advance

[davidbrady]

(05-04-2013 18:36)DOSZORROS Wrote:  David:

I am far from an electrical genius (spell dumb) but I am curious where you are headed with this and what your ultimate goal is? What systems do you want to monitor or control?

Thanks in advance

Hi George,

Hopefully these devices won't require genius! I really don't want to reinvent the wheel, and being basically lazy, I'm scouring around looking for off the shelf components that will be close to turn-key (I just turn the key and it works). I really don't want to spend a lot of time writing code, been there done that. But I'd love to have at my fingertips a cheap, compact, general purpose single board computer, preferably running Linux (most likely the Debian distribution) to monitor a host of things. Anything really, from S60 fan speed, to pressure differences, to cameras, to overall machine health. I'd still like to experiment with auxiliary air tanks on my steer axle suspension, but I need to measure accelerations in a controlled way for changes to be meaningful. So I'm thinking a few 3-axis accelerometers and maybe vibration sensors, but then I need to do a fast fourier transform to isolate frequencies. Playing around really. That's all. Who knows if it'll turn into anything for the general population.

The trouble (or fun) right now is that this industry has exploded so it's kinda hard to settle on one family of single board computers (SBC). Today I discovered the PhidgetSCB3, which runs Debian and which is very much geared towards the robotics and sensor world. It's an architecture that's more specific with a host of USB connected sensors. The prices are higher the Arduino, BeagleBone, and Raspberry Pi but what they've done with USB connected sensors is compelling. I very much like the idea of using USB (which is shielded twisted pair) to supply power and extract data from sensors. Need to do more research though...

[Arcticdude]

I thought USB was only good for 8 feet. Beyond that, you needed boosters every so often.

[DOSZORROS]

David:

I realize the design version will be on your coach. I would like the Beta version. NO!, I do not want the design version.

I wish I would have had you as one of my college professors (I might have learned something).

[davidbrady]

George,

Part of the fun is my son wants to do some things to control his model railroad, and we have two ponds on the property so R/C boats might be in our future. Maybe we'll configure a water-gun turret! If I can find something simple enough that my 7 year old can do it, then I've succeeded!

John,

It looks like 5m is max for USB 2.0, but there are tricks available to go further (USB to Cat5 extenders); although, the remote end may need a separate power supply. Plus, the idea is that the microcontroller would wirelessly communicate with a laptop on the dash. USB from sensor to controller, wireless from controller to laptop.

[DonB]

(05-04-2013 21:16)Arcticdude Wrote:  I thought USB was only good for 8 feet. Beyond that, you needed boosters every so often.

The design standard for USB is 5 meters. Signal boosting isn't the issue, it is one of timing.

I routinely use 10-foot USB extensions on things like my tethered Droid and the receiver for my wireless keyboard/mouse without problems, but a lot more would probably be asking for trouble.

[davidbrady]

Then depending upon the sensor needed, there can be an additional 12' of sensor cable.

Sensor > 12' cable > Phidget I/O Board > 16' USB Cable > Phidget SBC3 (single board computer) > Wireless > Dashtop Laptop

Star Topology.

[pgchin]

FWIW the auto industry is all over this with many different buss and sensor based configurations and implementations, a zillion computers and networks, all different....... IBM did the code for the Chevy electric car deal..... millions of k-locks of code..... I know the division GM(general manager)....... suggestion for you....maybe a call to one of your retired pals in the auto industry could yield you to someone or something you could use / borrow/ misappropriate/buy,etc..... they no longer use but would be great for your use case!

[davidbrady]

The development in this area is unbelievable Pete. Single board and single chip computers are defining a new industry: ubiquitous computing. It's here, it's now, and it's cheap. Just recently I stumbled across the Phidget lineup; we've already spoken about Arduino, Beaglebone, and Raspberry. I'm a kid in a candy store!

It's hard to argue with the popularity of Arduino and rPi. This alone makes it difficult to leave these platforms out of any WaGu design. The connectivity of these devices has also exploded with the development of radio transmitters and networking protocols. Today I ordered two Arduino UNO's and three XBEE S2 ZigBee Wireless Shields. This is all that's needed to build a mesh of connected Arduinos.

What I'd like to be able to do is grab a node (Arduino with XBEE), place it near the trouble area, wire up and connect a sensor, power on with maybe a battery pack. The node will then automatically join the wireless mesh network and establish a route back to my laptop via an XBEE ZigBee router. From the laptop I'll configure gauges to display sensor data. Unbelievable! The past few years has made all this possible. Everything is available, off the shelf, and cheap.

So, first thing that I plan to do is set up a 2 node mesh: one edge device and one router, both XBEE ZigBee enabled. Parts are on the way!