http://hardware.slashdot.org/hardware/06/08/05/0533259.shtml scroll to the bottom for great comments.
We are grateful to http://www.accutracking.com for supplying us with the extra data we needed to program our maps in the fashion we wished.
All hitchhikers have stickers from http://www.fineartregistry.com/ on the back side, for documentation and as a deterrent to stealing the hitchhikers.
Special thanks to:
Safari kiosk mode provided by Saft (http://haoli.dnsalias.com/Saft).
For questions or comments, please contact Julie Newdoll: exhibits _at_ YLEM.org or 650 591 7999
How does the tracking system work?
We are using Motorola i415 cell phones from Boost Mobile. These are prepaid phones; $50 gets you a phone, including the first $10 of calls or data; no contract is required. (We bought most of ours from EBay, which saved us a little money.) Despite their low cost, these phones are packed with features. The most important from our standpoint, of course, is GPS. They have GPS capability that can determine the phone's location even when not connected to the cellular network, and that can be queried from a Java program (unlike some phones which have A-GPS but cannot be accessed from a program, and may even only be used for emergency calls). They use the Nextel network for voice and data. We send GPS data several times an hour. Fortunately, data access on Boost Mobile is a mere 20 cents a day, so the initial prepaid $10 is good for 50 days of tracking. (Another handy feature is that the account can be boosted using any web browser over the net, independent of the phone.)
If we're unfortunate enough to have a phone stolen, it won't do the thief much good: it will be able to make calls for a few tens of minutes at most, and then can't have its account boosted without a password. The phone itself is worth perhaps $30 on EBay; a new SIM card costs ~$20.
The phones are loaded with other goodies too: they can run MIDP Java applications which can be installed using a data cable (more on that later), they can be used as cellular connections to the Internet (for 20c a day!), have the usual memo/calendar/phonebook apps, and a neat navigation app, Telenav, which costs $1.50 a route. Finally, the Boost web site is written in the plainest English one can imagine for a telco (compare the fine print that litters the major carriers' web sites), not surprising given the demographic Boost is marketing to.
Powering the phones
The standard i415 battery lasts between one and two days when the phone is running the tracking software; clearly not enough for an extended hitchhike. Phones come with AC adapters, and, optionally, a car charger. Both of these supply 5-6V to the phone. So, we simply connected a large sealed 6V lead-acid battery to each phone. We used the PowerSonic PS-6360 (36Ah) from Rage Battery in two hikers (Noyce and Terman), and the Tempest TR50-6-FR (50Ah) from Batteryspec.com in the other three, as our supply of the PowerSonics dried up. In hindsight I wish we'd used the 50Ah battery in the other two, especially as they have much further to travel. It's important to connect the lead-acid battery only when the phone's internal battery is fully charged, otherwise too much charging current flows and the phone shuts down.
Shutting off the backlight
The phone consumes ~90mA at 6V with the backlight off, rising to nearly 300mA with the light on. Although the i415 has an option to shut off the backlight, this is ignored when an external power source is connected, and the light stays on all the time. This is not good. By pure chance I found that a particular key sequence would turn the light off even when an external source was connected. The light comes back on if anything causes an alert on the phone: a text message or phone call comes in, a key is pressed, power is interrupted, etc. So, we set the phones up to forward all calls to our house phone, and not accept any incoming text messages. (Even though we have not given the phone numbers out to anyone, we found that every once in a while Boost Mobile would make a promotional call to a phone.) We're hoping we get 3 weeks or more operation.
(For the record, the light goes off when you select run in background from the Accutracking application's menu.)
There is one outstanding issue which we could not resolve and may scupper one or more of the hikers: the phones are known to occasionally pop up a dialog that says something like Satellite data is outdated, Update at 800.569.2532 to buy; tracking stops until this dialog is dismissed. This problem is known to Accutracking and appears to originate from Nextel or Boost.
Getting the tracking data using Accutracking
We use the Java tracking application from Accutracking.com. The application itself can be downloaded for free and installed on the phone. Then it has to be registered with Accutracking, which costs $5.99 a month per phone. Once registered, the phone application is set up with a unique ID. It can then send location data back to Accutracking.com, which can be viewed on a map on their web site.
There are various different options that can be configured on the phone: how often to send data (we send data every 20 minutes), the maximum number of points to cache on the phone when it cannot call home, whether to use cell tower locations in the absence of a GPS signal; see the Accutracking site for full details.
Installing the application requires a USB data cable and various items of software (Windows PC only) available from Motorola. If you get a starter kit it includes a USB cable as well as both AC and car chargers, but in any case they are all readily obtainable from various sources (cellphoneshop.net is cheap and fast).
We also tried the free application from Mologogo, but it did more than we needed it not only reports a GPS location to their server (which can't be easily downloaded), but also attempts to download a map to the phone showing your location (and any nearby friends' also using the system). In addition to the map being unnecessary for our intended use, we found that when it could not get a map it would sometimes hang.
Making the maps
Accutracking provides interactive maps showing locations and tracks, but we decided to build custom maps for our web pages. The map data and processing is done on a server at sonic.net (a truly excellent ISP).
The first step is to get the raw data from Accutracking, which provides a facility on its web site for users to download a .csv file containing latitude, longitude, time, etc. A little script using curl logs in to their site hourly and grabs the last hour's data, and adds it to the complete set we maintain on the sonic server.
We wanted to indicate on our maps the difference between a location obtained via GPS and those which represent cell towers, but Accutracking's .csv did not contain this. A quick email to their support address resulted in them adding this within a couple of days thanks, Accutracking!
Having got the data we then massage it into an XML file for use with Google Maps. Duplicate points are eliminated, as are GPS points very close to each other (most likely due to noise in the signal); this helps reduce map clutter, and speed up the map display.
One unexpected problem arose when the first two hikers were put on a delivery truck. The tracked locations bounced around in an implausible manner: the truck would be motoring down a freeway leaving a nice trail, and then suddenly would be whipped back to an earlier location. Eventually I realized that this only happened when we received cached locations (i.e., the cell phone could not reach a cell tower and saved the location until it could be sent later). The good support folks at Accutracking have confirmed that when a cell tower location is cached the phone's internal clock is used to timestamp the datum; otherwise the cell tower clock or GPS time (based on UTC) is used. The whiplash was due to the cached locations being marked with time from a different time zone (presumably local time). Luckily our received data distinguishes cached locations from non-cached, and cell tower locations from GPS locations, and so I can apply a correction when converting the raw data to XML. As of this writing (Jul 22, 2006) a single correction appears to fix all the points, but I don't know if that will hold as the hikers cross into different time zones. It appears to, as Noyce has gone from Eastern to Central already, but I can't say for sure I confess I don't really understand what's happening. I hope I don't have to factor in location, as I have been unable to locate a programmatic method for converting latitude and longitude to time zone.
Phones: $50 each (less on EBay) including $10 of phone/data charges (good for 50 days of tracking at 20c/day)
How well does it work?
Julie and I have had one of these units in our cars for some weeks. They can get GPS signals fairly reliably, even though the phone is foamed into a plastic case (just as on the hikers) and left somewhere in the back of our vehicles, not necessarily right by a window. However, during the journeys of two of the hikers by freight truck to their initial places of abandonment, they very rarely got GPS signals, but did consistently report cell tower locations (see maps for Noyce and Terman).
Mario says, if you want to know how this was done, download the source. Jim Pallas comments, "Of course, I don't know how you feel about it, but I'm sure glad you married Mario. His maps are everything(!)"