• read more:
• « older
• newer »
•  · 
• home
• archives
• about
• contact

Customizable mapping software

Herself asked yesterday if I could develop a website that gives taxi directions based on neighborhood landmarks which are often essential when navigating from a cab and making sure you are not getting taken for a ride.

This is an interesting idea which has been thought of before, if I am not mistaken. Somewhere I have a link to the sites in question, and if I find them I will link them. Sites such as Flickr and del.icio.us, which handle photos and links respectively, leave it up to the users to tag each image or link without any top down hierarchy. There are only so many ways you can tag science articles or self portraits. Both systems self-organize without a strict taxonomy overhead. Yet, with mapping, it would be easy to allow a strict taxonomy and user defined tagging. Country, state, county, city, and even neighborhood could form a basic taxonomy which user defined tagging could overlap.

The biggest obstacle, it would seem, would be the agreement of landmarks and features. First, you would need to seed a database with a bunch of people's idea of what landmarks actually are and what they look like. The Catholic Church on the corner, the Red Barn, etc all have addresses, or at the very least, a latitude/longitude that could be machine read. The physical address of the Catholic Church (1 Peter Way) would be linked to the user term, "Big Church on the Corner." Perhaps when we all have gps enabled phones with cameras we could email photos with gps data to a into the database and automate the process.

Once you have all of this data in the system, the problem becomes the issue of "bad data." Duplicate addresses, names, etc would plague the system. Not to mention the entries which would be joke entries (because it will happen). To combat this bad data, a system of flagging entries a la Craig's List and some sort of rating system would need to be implemented. Both these systems have a high chance of being gamed, so rating and flagging must be combined with IP and user records. Even then, the system will be gamed, so other ways to secure the data validity would need to be developed. Perhaps not every landmark would be eligible to be rated at the same time, or new landmarks would have a grace period (in terms of rating and history) before it is introduced into the main pool of landmarks.

Finally, a good AI algorithm would have to be written to take all of the landmarks, mapping data, and preferred routes and decide which is the best. I guess if we are going through all of this trouble, getting access to the road based traffic data (those inductive loops set in the pavement) to further add data points for the AI to use, along with historical trends would be useful. This would mean that the "most correct" route would change throughout the day and throughout the year based on vehicle throughput, closures, construction, etc. All of this data take together, you could make paths based on mileage, vehicle throughput, or by landmarks you designate.

Will this ever happen? Probably not. But parts are already here. We have Mapquest based on Navteq data, phone cameras (which have gps transmitters - but we can't access them), and computing power. We have already benefited from this research already - most major cities have central traffic centers, with traffic modeling and (primitive) AI making decisions. Yet, a commercially viable system we have described is not ready for prime time. Someone steal this idea and make it ready, just give me a free membership.

Comments

Post a comment

Name:

Email Address:

URL:
Remember Me? YesNo

Comments: (you may use HTML tags for style)