無線細胞網路中之行動模型與位置追蹤

Abstract

行動模型可說是個人通訊服務的本質所在，它可以被用來評估行動管理系統的效能，而行動管理正是無線多媒體網路的成功關鍵。在本論文中，我們提出了一個經過一般化的隨機行動模型，此模型可以很簡潔且精確地描述出具有多樣里程、直走取向的個人移動軌跡。其中我們打破移動軌跡關於網路佈局的連結方向限制，使得個人移動軌跡不會因為網路佈局的過份簡化而被扭曲地不切實際。 根據此行動模型的前提，我們針對以距離為準的位置追蹤法，提出了一個向前預測的策略，使得在不增加額外電話傳呼的狀況下，降低了註冊位置的更新頻率。首先我們在一維的網路佈局上，利用一個遞迴演算法，尋找最長的註冊位置更新循環，而決定出最佳的個人註冊位置。之後，我們利用一維網路佈局上的最佳解，去預測二維網路佈局上的最佳解。我們利用電腦模擬去做分析，發現我們所提出的方法對具有不同移動頻率及里程大小的個人都可以有效且近乎最佳地降低位置追蹤的花費。同時分析中也指出，對於追蹤移動里程大的個人位置而言，傳統的隨機行動模型較適用於基地台訊號涵蓋範圍大的無線網路，而在基地台訊號涵蓋範圍小的微細胞無線網路中，追蹤移動里程大的個人位置的花費往往被低估。 除了追蹤未建立通話頻道的行動電話外，本論文中也探討了位置追蹤對於已建立通話頻道的行動電話之效用。未來想要容納更多的無線多媒體網路使用者，基地台勢必會愈來愈密集，為了有更好的服務品質，網路管理者通常以預約頻寬的方面，使得跨越機基地台的通話中斷機率降低。我們試著在其中加入無線定位系統，讓網路端週期性地知道通話中行動電話之精確位置，期使網路管理者行使頻寬預約的效果能加大。然而實驗結果顯示，精確預測行動電話之移動軌跡，並不足以使跨越基地台通話中斷的機率降低，而主要的因素是在於預留頻寬之總量大小。

Mobility model is essential in Personal Communication Services (PCS). It can help to evaluate claims that related to the performance of mobility management subsystem, which is the key to success of wireless multimedia networks. In this dissertation, a generalized random walk process named normal walk is proposed to model the motion of PCS user such that random movement with multi-scale straight-oriented characteristics can be represented briefly and specifically. By liberating the motion from the inlet/outlet constraints of mobility graph, the trajectory of normal walk is never distorted even if network layout is over-simplified. Based on the normal walk mobility model, this dissertation presents a lookahead strategy for distance-based location tracking so the rate of location update can be reduced without incurring extra terminal paging costs. For linear mobility graphs, the optimal registered cell is found by an iterative algorithm so the average update cycle length is maximized. For planar mobility graphs, the author employs the results from linear cases to determine the eligible registered cell. Performance gain is evaluated by using Monte Carlo simulation for mobiles with different degrees and scales of mobility. Analysis shows that the tracking cost for mobile users with large mobility scales in microcellular networks, costs which are usually underestimated by the traditional random walk model, can be effectively reduced. More than tracking the idle mobile, the dissertation also considers that for mobile user with call in progress. In the next generation mobile networks, cell size will become smaller in order to accommodate high user density and multimedia traffic. To improve the call-level quality of service, the author tried to employ the wireless geolocation function along with the concept of shadow cluster to survive the handoff call. With the instantaneous position reporting, the traffic dynamics can be captured by the network so the dynamic bandwidth reservation are expected to be more effective for active mobile subject to handoff. However, the experiment results are totally different from the intuition. The key factor to prevent handoff call from dropping is not a precise prediction on user's mobility, but the amount of reserved bandwidth. Such a fallacy is enlightened by a simulation study.