Mobile computing systems are playing a more and more important role in everyday life and applications based on both local and wide area distributed wireless networks are being launched all the time. One of the more challenging applications is real-time traffic navigation, and this thesis focuses on our experience in the design and implementation of a REal-Time Traffic Navigation System (RETINA).
There are a large number of challenging issues involved in designing a system that can support navigation requests and meet the arrival deadlines. These include the management of real-time traffic data, the execution of navigation requests and the monitoring of the best paths for these requests.
Focusing on minimizing the probability of missing arrival deadlines as well as reducing the overhead for monitoring the best path, a replicated directed graph approach for searching the best path in which the servers cache traffic data from other servers is proposed. Since the accuracy of this approach depends on the currency of the cached traffic data, a number of approaches for refreshing this data are explored and an adaptive push and pull technique is proposed.
RETINA was implemented using two basic components: database servers and mobile clients. The server program is implemented in VC++ 6.0 on Windows NT and the 2 client program is implemented using both VC++ 5.0 for Win CE 3.0 and VC++ 5.0 on Windows 2000.
The server process consists of multiple flags for different functions including servicing requests from mobile client and for system management. A SQL server is connected to the server program at each site and it maintains a database containing the road connections of a district in Hong Kong.
A simulated process is defined at the server to create new traffic data values randomly for the road segments maintained by the system. Each client machine, PC or handheld PC, is pre-loaded with a set of maps for the roads defined in the database servers. The client program supports search for best paths, as well as a number of location-dependent requests.
In addition, a number of simulation experiments were used to explore the effect of various parameters such as push threshold and traffic update rate on the effectiveness of the various proposed mechanisms such as adaptive push and pull. The results show that our algorithms do perform well for various operating environments.
Author: Hung Hon Tak, Dick
Source: City University of Hong Kong