| 摘要: | 在這百家爭鳴的物聯網(IOT)時代，大量的裝置需要連接到網際網路，小到智慧家庭的應用，大到智慧城市及工業4.0。看似無限大的商機，但實際上缺乏統一標準。網際網路工程任務組(IETF)之下的6LoWPAN工作小組定義了IPv6和IEEE 802.15.4間的介面，使大量無線裝置可透過IPv6直接連網。ROLL工作小組訂定了RPL協議，使得基於IPv6的低功耗的無線網路就此形成。它不但低成本、低能耗、適合大量佈署且與現有網際網路協議相容，還有明確的標準可以依循。
RPL協議利用“目標函數”定義組網行為，每個節點都必需依照路徑成本來選定父節點，形成一個最低成本的網路。傳統無線網路，對於路徑成本的計算，使用的是最簡單的“累加法”。延著路徑把每一跳的成本加起來，即是到達根節點的成本。但是累加後的度量值，無法真實反映出路徑的實際成本。舉例來說，總成本為6的路徑，可由不同的路徑集合組成，它可以是{2, 2, 2}、{5, 1}或是{3, 3}等…。假設路徑成本與通訊的距離成正比，那麼在集合中包含較大數值，對於通訊品質則是不利的。但是累加後的度量值，無法分辨父節點的路徑中是否包含長跳距。使得子節點頻繁更換父節點，造成控制訊息在網路內大量傳送、封包流失變的更加嚴重。
本文將以統計學的觀點來篩選父節點的路徑成本，使用近似平均值與標準差的概念，來分析路徑集合，考量計算標準差的時間複雜度較高，改以平均絶對離差來實現，並將成本度量多播到相鄰節點，使得鄰近節點都有足夠的資訊篩選父節點。經模擬後證明，相較於官方定義的兩個目標函數，在高密度節點佈署下，有著更低的封包流失率、略低的時延及微幅增加的吞吐量，對於整體無線網路可靠度來說是一大進步。
;In this era of the Internet of Things, a large number of devices need to be connected to the Internet, from smart home applications to smart city and industries 4.0. It seems to be an infinite business opportunity, but in reality, there is no uniform standard. The 6LoWPAN working group under the Internet Engineering Task Force (IETF) defines an interface between IPv6 and IEEE 802.15.4, enabling a large number of wireless devices to connect directly over IPv6. The ROLL Working Group defines the RPL protocol to make a low-power wireless network based on IPv6. Not only is its low cost, low power consumption, but it is also suitable for large deployments and is compatible with existing Internet protocols. Further, there are already standards to follow.
The RPL protocol uses the "objective function" to define networking behavior. Each node must select the parent node according to the path cost to form a network with the lowest cost. Traditional wireless networks use the simplest "accumulation" for path cost calculations. Adding the cost of each hop on the path is the total cost of reaching the root node. However, the accumulated metrics do not truly reflect the actual cost of the path. For example, the total cost is 6 paths. It can consist of a different set of paths, which can be {2, 2, 2}, {5, 1} or {3, 3}, etc. Assume that the path cost is proportional to the distance of the communication. If you include a higher path cost segment in the set, it will have an adverse effect on communication quality. The accumulated metric cannot distinguish whether the parent node′s path contains long hops. The parent node is frequently replaced with the child node, causing the control message to be transmitted in a large amount in the network, and the packet loss is more serious.
This article will determine the path cost of the parent node from a statistical point of view, using the concept of approximate mean and standard deviation to analyze the path set. Considering the high time complexity of calculating the standard deviation, the mean absolute deviation is used instead, and the cost metric is multicast to neighbor nodes so that the neighbor nodes have enough information to decide the parent node. After the simulation, it is proved that compared with the officially defined two objective functions, under the deployment of high-density nodes, there is a lower packet loss rate, a slightly lower delay, and a slightly increased throughput. It is a big step forward for overall wireless network reliability. |
