透過您的圖書館登入
IP:3.133.86.172
  • 學位論文

增加分散式高斯賽德爾繞送演算法之平行度於無線感測網路

To Increase Parallelism Of Distributed Gauss-Seidel Routing Algorithm For Wireless Sensor Networks

指導教授 : 柯仁松
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


在高密度 (massive-dense) 的無線感測網路 (Wireless Sensor Network,WSN) 中,繞送路徑可以視為連續的線段,我們可以使用宏觀參數來重新描繪 WSN, 定義出一個地理空間 (geographic space) 的宏觀模型取代傳統的圖論模型,進而 避免大量節點所產生的繁瑣細節。在宏觀模型下,我們可以把負載平衡的繞送問 題描述成一組偏微分方程式(PDEs),經有限差分法 (finite different method,FDM) 轉換成線性聯立方程式後,繞送問題便轉換成解聯立方程式,再以高斯賽德爾迭 代法 (Gauss-Seidel iteration method,GSI) 來求解,其解即為 WSN 中的繞送方向。 然而在密度差距過大的 WSN 中,高斯賽德爾迭代法會發生無法收斂的情況, 而戴拉葛薩迭代法 (De la Garza Iteration method,DLGI) 可以解決這個問題,但 是需要耗費的相當可觀回合數,因此我們提出一個平行度更高的演算法 Color-Grouping Method,將相同回合中能夠更新的格子點,根據其距離的特性區 分出來,再將這些格子點分成同組,依組別順序來進行迭代運算,可以大幅減少 GSI 及 DLGI 收斂所需的回合數。

並列摘要


In massive-dense Wireless Sensor Network (WSN), routing paths can be considered as continuous lines. We could re-describe WSN by macroscopic parameters, defining macroscopic models of geographic space to replace traditional graphic models, and then it can avoid the complicated details produced by large number of nodes. In macroscopic models, the routing problem of load balance can described as a sort of partial differential equations (PDEs), after transforming it into linear equations by use of finite difference method (FDM). The routing problem now is transformed into solve this equations, and we can solve it by Gauss-Seidel iteration method (GSI) which solutions is routing direction in WSN. However, GSI cannot converge in some special WSN which node density varies significantly. De la Garza Iteration method (DLGI) can solve this problem, but it needs considerable rounds. Thus we design Color-Grouping Method to increase parallelism of these algorithms, which method group grid points which updates in same round according to property of distance, and this method will do iteration in grouping order, it can substantially reduce the number of convergence rounds of GSI and DLGI.

參考文獻


recursive data dissemination protocol for wireless sensor networks, UCLA
sensor networks, cornell university library, 2011.
Consumption in Large Scale Wireless Sensor Network, 2011 Third International
[1] R.F. Erbacher, S. Hutchinson, Distributed Sensor Objects for Intrusion Detection
Systems, in: Proceedings of 2012 Ninth International Conference on Information

延伸閱讀