| Summary: | 碩士 === 國立臺灣海洋大學 === 電機工程學系 === 98 === For people who frequently spend long hours behind the wheel, possibly stopping at many different places in and around metropolitan areas, automatic route planning is quickly becoming a necessity. Besides guiding people to their destination faster under normal conditions, route planning may also help reduce fuel consumption. However, most current route planning products, be it pure software running on a desktop computer or a dedicated navigator device assisted by GPS, do not update their suggested routes according to the latest traffic conditions. As a result, the generated path may no longer save people time and effort getting to their destination should there be unexpected traffic jam or any accident ahead. Including the latest traffic conditions as parameters requires that large amount of traffic related data be collected, assembled, analyzed, and filtered for various purposes before updated paths can be sent to the users. For an on-line service covering a metropolitan area, the computing power and storage capacity involved can be formidable, especially during the rush hour. This motivates us to pilot a cloud based solution to this interesting problem. A cloud based solution in this case has two obvious advantages. First, high degree of elasticity is a basic requirement for any cloud architecture offering commercial service. Automatic route planning with dynamic updates fits in a category where the amount of incoming service requests can vary greatly with time. Second, by building the solution on a cloud platform, a service provider can focus on making its service more useful, without the risk of capital investment on hardware facilities.
This research adopts the well-known Google App Engine as the cloud platform and A* as the backbone path planning algorithm. The algorithm has been modified into accepting a parameter associated with each intersection and representing the degree of traffic jam at that location. Preliminary experimental results have shown that the solution proposed in this thesis not only provides shortest paths on demand, but also re-routes the users in time when adverse road conditions occur. Our program succeeded in re-routing the user with a faster new path 238 times out of 264 simulated runs. The success rate is over 90 percent.
|
|---|
