|
Kris - no disagreement on Navizon's software and method of collecting data. What we have found, however, are two important variables that do have a significant impact on the efficacy of the data you collect: (1) Hardware - devices are not all created equal and each has its own power, signal, transmission, reception, etc. properties. This is true not only among similar devices, but when you talk about an array of models, manufacturers, versions and then what individual users may do to modify them...well, you can see how the problem grows exponentially. When peer based systems rely on 'personal' equipment to collect data, you have lost quality control because of it. We learned these lessons after testing literally dozens of devices. (2) Process and methodology - Peer based systems also rely on individuals to create coverage areas. What you end up seeing when you analyze peer-based data are two very distinct patterns. One is randomness - in some areas motivate individuals will be extremely throrough and systematic while in others it will sporadic and chaotic. You simply can't control, or rely upon, the motivations of individuals. The other is human nature - and when it comes to travel, you see it in having a bias to always migrate to major thoroughfares. So if you rely on peer contributions using their 'everyday' movements, they will always seek the easiest and most direct routes. Which generally starting point=>side street=>main thoroughfare=>side street=>destination. Even fleets, like FedEx, UPS, USPS, taxis, police, etc. all follow these same systematic patterns. When you combine these two, the error factor they introduce into the underlying data simply can't be corrected through processing. If you are going to create a system that relies on a stable, predictable and reliable data source then you need to control all aspects of the data collection to ensure high quality output. Jed > ----------------------------------------------------------------------
|
_______________________________________________ Geowanking mailing list [email protected] http://lists.burri.to/mailman/listinfo/geowanking
