Friday, May 30, 2008

P2P begets C2C

The next step in automotive control is coming, and a sign of it shows up in the product of a small Mt. View company named Dash. Dash has been selling on Amazon.com since the end of March. Their first product, a GPS unit that has C2C (Car to Car) communication capacity. The C2C is actually buffered through the company's central servers which aggregate location and speed data and send that back in terms of traffic updates. This is more immediate and accurate than getting feedback filtered through municipal traffic sensors. Getting info from the municipal sensors is called "C2I", or Car to Infrastructure. That's part-and-parcel of what the US calls "Vehicle Safety Communication", or VSC, which is nowhere as cool as "C2C".

I spoke with a company representative last year before the product went public. She said that the C2C wouldn't be part of the initial product, but would be built-in and once there was a critical mass a software update could be delivered to allow C2C and/or C2I functionality to be rolled out.

Though the Dash has some high hurdles to overcome in that there needs to be a critical mass of C2C-enabled cars for a lot of it to work to its potential. Less so for C2I--much like a "Flash Pass" it could be sold by municipalities as a premium commuter service and increase ROI for C2I's infrastructure-side improvements, to build a cycle of improvements.

How you get to that critical mass is beyond my ken, but Dash told me that their
exit strategy was to sell this to a car manufacturer, or, if they had interest from more than one, to license it to many, and become like EDS. Do that and hey'll have a big enough network. Of course, On-Star and the european consortium "EURCAR" is working on a similar agenda, and they've chosen to call their initiatve C2C. They're cleary the cool kids on the block (which includes Audi, BMW, Damlier, VW, Fiat, and Renault--notice the absence of The Big 3 and Nippon Co's?)

One of the first things to work out is the spectrum on which all this will happen, right now they're talking about the 802.xx technology in the 5.8Ghz, at the high end of that. But of course it has to be unfettered, not-noisy, and relatively secure (more potential for mischief than hacking an ATC center).

The thing about this initiative and the potential hinted at by The Dash is that I believe the future of accident avoidance and really the next phase in automotive control (now that we have ABS, ESP,PASV, TCS, lane-change notice and auto-parking) in C2C is direct communication to alert surrounding cars to lane changes, braking, erratic steering as well as general traffic speed for congestion alert. Enabled with nearest-neighbor information, the car can know in advance of its position when another car is planning to change lanes or is slowing--without the driver having to notice lights, guage deceleration rates, move their foot, etc.

If MB has a brake system that will buy you a 1/4 sec. by "pre-positioning" its brake pads when you lift off the throttle immediately, imagine what gains you could get by electronic inter-car communcation. Take the slow wetware out of the loop. Um, that's you and me, pal. First enabled for freeway/highway travel and filtering down to urban traffic, there are difficult but not insurmountable problems to solve, but it is clearly imaginable at this point. For rural and suburban roads the challenges are harder. We'll leave that for now in the hands of the DARPA Challenge folks. Handling anything, anywhere, anytime from behind the wheel is currently a job for the slower but much more adaptable wetware between our ears.

If we invest in some relatively small additions to C2I infrastructure in urban settings, the car could know well in advance when lights will change, speed limits, etc. Add that up and you really have the makings of great efficiencies in auto traffic and not just safety.




[above image from the Renault presentation "Avoidance and Action in a C2C Network". ]

Coming Into an Intersection Near You
Imagine for a second that all cars were equipped with C2C. Or, in the closer future, imagine that a majority do. You're overlooking a major traffic intersection (don't know why, but I'm thinking of DeAnza Blvd. and Stevens Creek Blvd. Let the Apple employees be the lab rats). It's about 9am on a weekday and traffic is pretty heavy. At some moment, without drama, the C2I detects that
all cars within 500yds of the intersection are C2C equipped [unequipped cars can be "ratted out" to the intersection by their equipped neighboring vehicles. "I can see you, but I can't hear you!"]. At a synergistic moment, the traffic lights begin to blink all their colors in all directions, alerting drivers to the upcoming thrill ride. At the edges, all pedestrians and cyclists are signalled to stop. Inside the cars, the C2C system alerts each driver, "Automatic intersection ahead, yield control to me in 5, 4, 3, 2, 1....I have control of your car. Control will be returned to you shortly. Listen for the countdown."

The car can take on the very simple task of steering a little on approach to keep in-lane stability. This is much simpler than the auto-parking mechanism of Lexus and others. Carnegie Mellon and Stanford have been running cars around their campus on autonomous control for years now--well-defined lanes, low speeds, no traffic. Lane change attempts are suppressed. Remember, that all the cars have "immediate radius destination" signals being sent out, so early moves into turn lanes and lane adjustments for speed (slow to the right, enforced by the cars themselves! Heaven!) have already been made. Then every car gets locked into a steady state on approach to keep the real time planning as simple as possible, Drivers basically don't have control over their pedals.

In all directions at the same time and all the cars just go through the intersection without any drama.

The cars interleave themselves with as little slowing as possible; the system creating the following distances necessary to get all the cars through the intersection in a basket weave. Little wasted gas, low chances of accident. It is Atari Frogger on a grand scale and everyone makes it across.
As non-equipped cars approach the intersection, it goes back into human-interface mode, the signals go back to normal, the audience of appreciative pedestrians whistle their awe...and maybe the whole thing happens again in a few minutes.

It occurs to me that perhaps motorcycles will be the most difficult to bring into this fold. But that's an off-the-cuff observation and would need more thought. This would be fun software to work on, bordering on "AI". Certainly some "hive think" and "emergent behavior" aspects of it.

When could this happen? Within 20 years, easily. Within my lifetime, certainly.

Other resources.


Wednesday, May 21, 2008

Acronymicon

I don't think I made it to be the first on the web with this, but the new Indiana Jones movie, "Indiana Jones and the Cave of the Crystal Skull" is a mouthful. So I condensed it to the acronym: IJatCotCS. Google it. Anyone besides me? I have to say though, that IJATCOTCS is either an acronym for a joint DEA/FBI operation, or a new JAVA/CSS conversion API. Or something from the Pentagon. Yep. That's it. It's a Pentagon acronym. Trebly-embedded. Which reminds me. Is it really true that "JAVA" stands for "Just Another Vacuous Acronym"? If it doesn't it should.

Tuesday, May 20, 2008

Dancing With the Stars, courtesy of Jib-Jab

You remember "This Land is Your Land" from the '04 Presidential election. Do you think those guys over at Jib-Jab have been asleep since then? They haven't and here's the proof. Nominated for one of the best Flash applications on the 'net.


Thursday, May 1, 2008

Smart NAS<->USB Copies

I was looking at NAS solutions again today, as 1TB drives external get to my arbitrary but magic pricepoint of <$0.25/GB, and found this nice LaCie item.
What surprised me was buried in the comment by the second reviewer. He wrote:
"Also I had another USB external drive that I wanted to copy the data from to the NAS so I connected it directly to the NAS and mapped both drives on my laptop (which is wireless) to do a copy and it ran super slow, for some reason it did not just copy from drive to drive but it was transferring across my wireless."
Wow.
This is why we need basic computer "science" (really, a practicum that isn't focused on using apps) as a part of the general education. More English education needed to combat run-on sentences is also indicated in this case--but I digress.

Looking deeper into his situation and confusion, wouldn't it be cool to have devices smart enough to figure out what this user wanted (and expected) to do? I.e., the computer looks at the source and destination targets and sees that they're on the same IP and so hands them (maybe the source drive) a signal to check for routing optimizations, which might include looking for all USB hosts and clients for some matching target credential.

Or in this particular case, couldn't the NAS figure out what is going on, since it is receiving both the source and target destinations and spoof the client computer into thinking that the transfer is beginning, continuing and then ending as long as it takes to make the USB xfer? The user gets all the "flying folder" feedback (or spinning beachball for you Mac'ers) and for the appropriate time whilst the USB does its thing.

If Cisco and Comcast can spoof to interrupt Torrent copies, it can't be that hard to sniff this out. And for that matter, if the NAS kernel is Linux (as it surely is), couldn't you stuff some iptables rules that would handle this "feature"? Allowing this could be as easy as a feature in the web interface that, on detecting a USB drive, enables a checkbox that says, "allow disk-to-disk copies between by networked clients"? Defaulted to "on", of course. You might want the files to fly through your CPU/OS stack, because you might have antivirus checks to apply, Google desktop, or other benevolent (or, yes, malevolent) spyware to apply. Hmm... off by default then?

Surely I'm not the first to have thought of this (sm).