The new version will display my phone's location in human terms; actually in Larry's terms. We have software that learns places that are significant to Larry's phone (like home, office, pool) and will use these landmarks rather than raw gps data.

The points are gps or cell tower readings of Larry's outdoor activities over the past day (or more).

Please note that Larry reserves the right to display incorrect information. You get what you pay for. For example, it may show that Larry is in his office (go check his web page and read the message on the phone screen)) when he is really watching a movie, or that he left his office at 6:30 to head home but really stayed there another 15 min and only claimed to have hit lots of traffic on the way.

Privacy is the right to lie about yourself.

To see just the last recorded location of Larry's phone, click here.
To see the places that Larry's phone likes to "hang out" click here and choose a cluster size. This is the first step in convering GPS locations to human-understandable locations. That data is from the summer.

To get back to Larry Rudolph's home page click here or go to http://people.csail.mit.edu/rudolph.

Very soon, the location data will be made publically available; if you want access to it now, send him email. (There is now a database of over 300 cell towers in the Boston area and their gps coordinates -- well the gps coordinates of the phone where it heard the cell tower. I am happy to make this data available.) Everyone talks about location-aware computing, and experiments are usually contrived. Much of Larry's data is "dirty" -- there were bugs in the code that used to to report the wrong location when it was indoors. It sometimes takes 5 min for the GPS device to get a valid fix. Often the data shows Larry arriving at his office in the morning and then leaving the football field in the evening before arriving at home. It rarely shows him actually leaving his office building. In the morning it appears that he leaves from the BU bridge over the Charles river, but rarely shows him leaving from his home. Sometimes, the gps unit is not charged and sometimes he forgets his phone at home. Real data is very flaky. The good news is that the data seems to be getting better over time.

Here is my poorly written python S60 code (there are some libraries, and I will add them as soon as I upload them to this server, here are some: newgps.py , newbt.py , btdis , BtDiscoverLocations.pys . ). This code does the following. Periodically, (every 15 or 180 seconds), it looks around for bluetooth discoverable devices and records any that it finds. It also records the current cell tower. If it finds my own bluetooth GPS device, it connects and tries to get a reading. Every 10 periods, it uploads this inforomation to my own web server. If my GPS device is found and gives a good reading (e.g. I am outside), then the time period is 15 sec, otherwise, it is 180 seconds. Actually, the code is more sophistocated -- now that I know my pattern of activity, I probe less frequently during nighttime, when the phone is idle; I was getting tired of recharging the phone every day.) All this code is for python on Symbian phones.

I also track bluetooth ids. All this information is processed and automatically corollated with my calendar. From calendar entries, I can figure out the bluetooth id of the phone of any visitor I meet whose name is in my calendar. If I have a meeting with Alice at 2:00 pm for an hour and I detect a new bluetooth id during that period, I can deduce that the bluetooth id belongs to Alice. (I keep this information private, sorry.) I do the same thing for GPS / Cell tower data. If my calendar indicates that a meeting is at a particular place, then I now know my own personal names for the location. Each night, I update my journal or diary. I get to see who I meet and where I have been. It is a great time to apply names to locations and bluetooth ids that have been unidentified.

I am preparing a paper that describes all the analysis needed to do these corrolations. It is easier than any of the sophistocated published results that just look at one sensor (gps, cell tower, wifi base station or bluetooth) to figure out interesting locations.