- 會議論文
無線感測網路應用於多重定位點之研究
The Research of Wireless Sensor Network for Multiple Locating Points
摘要
近年來隨著無線網路的發達，因此發展出許多不同無線網路的型態，而無線感測網路(Wireless Sensor Network, WSN)更是近年來無線網路的新技術。目前在無線感測網路上最熱門的應用為定位技術，但是無線感測網路在定位技術方面仍擁有許多問題點必須去克服。其中影響定位準確性最大的問題則是我們目前環境所造成的。以往室內定位的方式，都是在晶片上撰寫數學程式，而達到定位功能。本文中提出一種新的定位方式，藉由接收到訊號強度值，透過類神經網路來訓練及模擬來建構定位方式。本文採用Freescale-13213無線感測網路節點模組來針對目前實驗環境來探討在此環境當中，多重定位點對我們室內定位精確性測試；透過多個定位基地台節點擺設的實驗模式，並結合類神經網路來模擬感測節點在固定座標定位上的精確性，藉此尋找到感測節點在固定座標上較佳的定位精確性以及較佳實驗環境。我們提出倒傳遞網路與徑向基網路這兩種不同網路架構運用在多重定位點當中。我們測試訊號時發現，訊號的強度會因為目前現場的環境而有所漂移，經由實驗方式我們發現倒傳遞網路所訓練出來的模擬命中率，不如徑向基網路訓練出來模擬命中率來的高，倒傳遞網路在偏移量分別為1%、3%、5%時，命中率分別為85%、60%、40%。徑向基網路在偏移量分別為1%、3%、5%時，命中率分別為95%、90%、85%。
並列摘要
Along with wireless network developed in recent years，it had developed many different types. As Wireless Sensor Network (WSN), had became a new focus of wireless network technology. The way to do indoor location in the past is writing program on chip to work on location. In this paper, we mentioned one new method. Constructing the way of location, through accepting the Received Signal Strength and using Neural Networks to train and simulate. We use Freescale-13213 Wireless Sensor Networks module to treat in this environment, the effect of Multiple Locating Point on the accuracy of Indoor Locating. By the way of setting many Local Base Stations and combined Neural Networks theory to approximate the accuracy of Sensor node on fixed Location. Then find better accuracy of sensor node on fixed Location, and the better experimental environment. The article mentioned the applications in Multiple Locating Point of Back-propagation network and Radial basis network. And we discovered the strength of Received Signal changes by environment. Through experiment method we found the approximated accuracy of Back-propagation network didn't higher than Radial basis network. The deviation of Back-propagation network is 1%, 3%, 5% respectively, and the percentage of hits is 85%, 60%, 40% respectively. The deviation of Radial basis network is 1%, 3%, 5% respectively, and the percentage of hits is 95%, 90%, 85% respectively.