Summary: | 碩士 === 國立中興大學 === 資訊科學研究所 === 91 === In recent years, the development of wireless communication environment has made it popular to use mobile devices to get information and exchange messages. Because of the limit energy in mobile devices, how to download information efficiently and to save the energy for mobile devices have become essential problems. Related studies used to insert additional information in the broadcast cycle for the purpose. Their solutions can be classified into two categories: predictive and non-predictive broadcast schemes. Non-predictive broadcast schemes, such as the hash schemes and index schemes, insert the hash function or index tree into a broadcast cycle, resulting in longer broadcast cycles and tuning time because of collision or tree traversal. On the other hand, predictive broadcast schemes can reduce tuning time effectively if their prediction errors are minimized or limited.
Existent predictive broadcast schemes include the slope prediction method and the linear-programming prediction method. Since both schemes have a different set of parameters for performance tuning, they are difficult to compare analytically. In this study, we propose a new broadcast scheme, namely least-square prediction method, to improve the prediction errors of both schemes, and conduct a simulation study on the three prediction-based broadcast schemes. The performance metrics include index space, probe wait time, prediction error, tuning time, and access time. Experimental results show that the least-square prediction method can effectively reduce the prediction error, improving 83.24% of the prediction error in the linear-programming prediction method and 93.40% of that in the slope prediction method. The linear-programming prediction method has the shortest tuning time, while the slope prediction method needs the least index space for the broadcast cycle.
|