Monday, March 10, 2008

Mesh Networks on Transportation: will it work?

A monthly magazine for technical types, Baseline, has a long article about me and the idea of using mesh networks in the transportation realm.

The article is well researched and interviews people across the industry. A sad omission is the lack of named attribution to Andrew Blumberg, my colleague who has done all of the work on privacy protection. The article also puts the cost of the in-car boxes too high by a factor of two.

My favorite is the closing paragraph:

“I see Robin as one of the global thought leaders of transportation technology,” Villa says.

Allies in the open-source community are hopeful.

“It’s completely doable with the technology that’s available today,” says Sascha Meinrath, research director for the Wireless Future Program at The New America Foundation, a Washington public policy institute and think tank, a leading expert on community wireless networks and a member of the Meadow Networks board.

“There’s pretty much no scalability limit and no throughput limit. We’re 80, 90 percent of the way there. It’s just a matter of finding a municipality, a company, a patron willing to fund this.”


Hey, it's my blog and I'll blog it if I want to :)

15 comments:

Unknown said...

If you've got all these cars driving around with WiFi radios in them, why not put the WiFi radios in the GPS navigators that everyone's buying for $150 now. If these navigation computers then exchanged recent (anonymized) travel time history using an open wireless P2P protocol, each car could pretty quickly build up a picture of where the congestion along their planned route is, and avoid it. Because this data could be exchanged peer-to-peer, it would not need to be managed by any government or industry-operated ITS systems.
Just a thought.

Anonymous said...

Hmmm. Don't be surprised if TomTom introduces floating vehicle data soon in their top-end units, using GPRS. But what's their interest in including an open P2P radio? For them it's the traffic data that has all the value, they can send it back to each user.

Unknown said...

Well, if we're talking about combining this with WiFi congestion pricing, then we're assuming the WiFi radio is already there. But maybe not in the same box as the navigation system.

You are correct, companies like TomTom would probably not have an inherent motivation to include WiFi radios. They'd rather provide a service that they can charge for, like Dash Express, which is already doing this for $12.99/month. But I'm not sure there are a lot of people willing to pay #12.99 per month, myself included. A P2P solution would not have a monthly cost. If enough people were asking for it, they could be motivated to add the feature. Anyway, I think that we will shortly be seeing GPS chipsets that have WiFI built in, so the P2P radio may already be there anyway.

But turn the question around and take it back to congestion pricing -- as a driver, what would be my motivation to put a box in my car that charges me when I drive around? Why would I vote for politician who pushed such a plan? I think the motorist would have to get something out of it. If they could buy a device for $200 that was both their congestion pass and a GPS navigator that could accurately route them around congestion, I think you might get some interest.

Jeremy Gore said...

Hi Robin, saw your TEDtalk and really liked the goLoco idea. I just signed up.

Mesh networks - really all sorts of distributed p2p networks - are smart, robust, and have so little opportunity for big profit that few will pursue them. Maybe what it needs is a kick in the pants from the right heavyweight looking to get into online communications on the cheap, someone more interested in services on open standards than in maintaining hardware... like say, Google.

This could be a killer opportunity for their Android platform. As I understand it, Google also owns some very interesting patents on transportation routing. They're also rumored to be in the bidding for the 700mhz spectrum auction. Also, Apple might be interested for very similar reasons, except they ARE interested in the hardware side - their iPhone and iPod touch already being excellent mobile communications platforms. TMobile could also be an interesting player - their mobile network is relatively poor, which has led them to branch out in interesting directions, such as their hotspot@home option that lets you make calls via wi-fi for free.

Edward Vielmetti said...

I can imagine a setup built out of existing parts that would do a mesh network in a car - start with the Berlin Freifunk's BATMAN routing code, stick it in a $100 router, wire it up to 12v and you're pretty close to done.

This gets even more interesting if the mobile router also fits in your backpack, or if it's code running on a cell phone. Power consumption will be the limiting factor is my best guess.

Tom Cheesewright said...

Power does seem to be the limiting factor when you want to stick something in a backpack as I've outlined in this little sketch. However for vehicles it makes perfect sense. People already manufacture 12v Linux-based routers for cars for the security and law enforcement industries, so should just be a software issue to start mucking around with it - no need for bodging or expensive custom hardware.

Edward Vielmetti said...

If you build the portable hardware on a laptop, you could get a good expensive high quality laptop battery - this laptop I'm typing on has a 4700mAh battery, so figure out your power consumption and divide based on what your power needs are.

Ashwin said...

Why Wi-fi? Why not just a smartphone. If I remember right, a company called Waze from Israel has a solution that does dynamic rerouting based on peer's navigation. Cheaper than wifi and really dynamic and fool-proof on a gprs connection

Unknown said...

Not everybody has a smartphone with a data plan. WiFi would work without a monthly fee.

Unknown said...

Ash: why not just smartphones? Smartphones are great, but they will only give you location and acceleration. There are a whole bunch of other car data points that are interesting: fuel levels, engine checks, emissions, ignition enable, door unlock. An open platform would enable so much more. Here is the link to a 2 min interview on that topic:
http://networkmusings.blogspot.com/2009/09/creating-open-in-vehicle-technology.html

John Gilmore said...

Mobile mesh networks are a great idea -- but they don't work yet.

Check with One Laptop Per Child. They discovered that when they put 30 kids in a room, the mesh network breaks down. It fills up with control messages and there's no room for actual communication among the users. Their new XO-1.5 laptops don't have mesh, just ordinary WiFi.

The idea of mesh is fine -- but there is not a working, usable, ready implementation that scales up to one classroom, let alone an entire school or a highway or a nationwide or worldwide grid. The reason the Internet was adopted worldwide is because it could actually scale up without breaking. Mesh isn't there. The 802.11 mesh works on 5 nodes at a time, breaks between 10 and 30. If you want mesh as a worldwide tool, track and fund the research to make it scale up.

I know Freifunk has bigger networks - but they aren't mobile, and they themselves admit that their network won't scale to thousands, tens of thousands, or hundreds of thousands of nodes.

John Gilmore said...

By the way, the idea of cars forming mesh networks (and then sharing their music collections with each other) came from Cory Doctorow's book _Eastern Standard Tribe_ (http://craphound.com/est/). In this blog post, http://craphound.com/?p=403 , Cory credits the idea to "his pal and former business partner John Henson".

Please tell people where the idea came from, as you publicize it.

Unknown said...

He may have come up with the idea of sharing music between cars this way, but ad-hoc vehicle-to-vehicle networks have been discussed for many years before that book came out...
http://www.cs.ucla.edu/~kclee/RoutingBookChapterKLULMario.pdf

Unknown said...

In response to John Gilmore's comments:

The last thing that the OLPC experience proves is that WiFi mesh networks don't work.
What it definitely proves, is that if you have the "when you have a hammer, every problem looks like a nail" mentality, you are not going to solve many problems.
For the record, what fills the spectrum when you stick a large number of XOs in a small space, are multicast packets from the application stack, not mesh control messages.
Now the real question is, why would one need to use a mesh in a environment like that (instead of a plain access point) and as one of the designers of the XO mesh, I don't really know the answer...
I definitely never had that scenario in mind when designing OLPC's mesh stack.

As far as to why XO 1.5 doesn't have the embedded mesh stack of the original XO, it has to do a lot more with the lack of resources to port the mesh code to the firmware of the new WiFi chip and a lot less with whether the technology worked or not...

Unknown said...

The idea of car meshing is older than WiFi. I do remember people talking about using ad-hoc networking of cars to collect, disseminate and propagate traffic information in the 90s when I first joined the Media Lab.