Research On Some Key Technologies For Wireless Sensor Networks In Underground Mining Environment

As a new technology with cross-disciplinary contribution, wireless sensor network (WSN) has widely application in industry, agriculture, military, environmental monitoring, rescue operation, etc. Meanwhile, ultra wide band (UWB) has emerged as a promising communication due to its abilities on low power consumption, high data rate, anti-multipath, anti-interference, accurate ranging, localization, etc. Therefore, WSN and UWB could be naturally integrated as UWB sensor networks for some special environments like underground mining applications. Considering the features of WSN and UWB, the thesis mainly researches on the application of wireless sensor networks in underground mining environment, specifically on the related topics of modeling, ranging, localization, routing and medium access control using cross-layer design method.Time hopping UWB (TH-UWB) is one type of impulse radio UWB systems with high performace-price ratio. The randomly generated TH code could resolve multiuser interference in underground mining. According to the characteristics of underground TH-UWB signal propogation, we modify the indoor modeling by referring the channel models of IEEE802.15.4a, and propose an indoor alike underground UWB path loss model. Compared with other models, the UWB path loss model takes into accout the influence of different shading on varied areas and locations so that it facilitates the estimation of received signal strength indication (RSSI). Using the propagation model, maximum likelihood estimation (MLE) based RSSI ranging is designed to avoid the error of random ranging and improve the stability of underground ranging. In addition, because of the multiple access function in TH-UWB, the RSSI ranging integrates the multi user access into MLE phase, and improves the ranging throughput in wireless sensor networks. The Cramer-Rao lower bound (CRLB) is derived and then used to verify RSSI raning algorithm in simulations.On the basis of underground channel model and ranging information, a distance related localization algorithm is proposed using multidimensional scaling (MDS) idea, which is called non-metric MDS and maximum likelihood estimation (NMDS-MLE). Non-metric MDS only requires a monotonic relationship between the proximity of objects and the Euclidean distances. NMDS-MLE establishs this kind of relationship between estimated distances and Euclidean distances. In addition, it uses MLE based RSSI ranging to improve the performance of localization in wireless sensor networks. NMDS-MLE is a low cost algorithm, which only requires basic distiance information but could derive the relative coordination and abosolute coordination of all nodes. It also decreases the demand of the number of beacon nodes for assisting localization. The performance of NMDS-MLE is validated using theoretic analysis, simulation and real testbed experiments.The location information could support the design of location based routing in mining environment. Underground communications present unique signal propagation characteristics due to the geographic and geological features, which in turn impact the underground multi-hop routing pattern. Our research is the first work to address modeling underground tunnels and designing underground tunnel routing in a holistic cross layer manner. For simplicity and without loss of generality, a cylindrical tunnel model with three dimensional coordinate is proposed, and then a hybrid signal propagation model is used to estimate the RSSI by switching between free space and two ray model. A holistic routing protocol based on the geographic features of tunnel, called BRIT (Bounce Routing in Tunnels), is designed for underground wireless sensor networks. BRIT uses forwarding speed as a routing metric for forwarding selection in multi-hop routing, by integrating location, distance, signal quality, data rate, queuing status, etc. In addition, BRIT provides a route suppression mechanism to avoid message flooding in route discovery, and a multi-path strategy to recover routing path. BRIT is evaluated in terms of network throughput, packet loss rate, stability and latency using simulations under random, lineup, bipartite and spiral deployments.The simultaneously channel access in communication range or forwarding data in multi-hop routing by multiple nodes will cause detrimental collision and interference in wireless networks. The scheduled channel access in wireless networks could create a conflict-free channel access schedule. However, it has inherent performance drawbacks due to the wastes of unused time slot assignments and the long delays between successive channel accesses of each node. The distinctiveness and orderliness of each symbol in the Latin square present very attractive features in the scheduling problems. A dynamic multiple access scheduling based on Latin squares (DyLS) is proposed in the medium access control (MAC) layer for channel access schedule problem in wireless sensor networks. DyLS modifies time division multiple access (TDMA) scheme, and random access scheme by generating deterministic schedules according to the Latin squares. Using on-demand graph coloring algorithm, every node involved in communication is assigned a unique color in its two-hop range, and then assigned a corresponding row index of Latin square. In peer-to-peer access, each column in the Latin square corresponds to backoff time; in multi-hop access, each column in the Latin square corresponds to the access priority. Simulations in fully-connected network and multi-hop network show the near-optimum performance in DyLS for channel access scheduling in comparison with NAMA, UxDMA and IEEE802.11DCF, respectively.