| dc.description.abstract | Wireless sensor networks (WSNs), which are formed by tiny and low-cost sensor nodes derived from the recent advances of science and technology, have been gradually paid more attention and now are being used for a variety of application areas, including environmental and habitat monitoring, health care, building/home automation, traffic control, etc. Among various WSN applications, the sensor readings from network nodes will be valuable to users for advance analyzing or processing if the correct time and location information at the moment is attached to them. Moreover, many collaborative tasks or algorithms need each sensor node to maintain a global clock and be aware of its current locations at the same time so that the tasks or algorithms can be run well. Thus, both time synchronization and localization for sensor nodes are essential and important services in WSNs. There have been many schemes proposed to realize these two services in traditional wireless networks. However, wireless sensor networks are rather different from traditional wireless networks due to restricted energy resource, large network scale and so on. Hence, new solutions for these two services must be proposed specially in WSNs.
To let all nodes maintain a global clock which is synchronized to that of a reference node, we design a time synchronization protocol for multi-hop WSNs. In the proposed protocol, unsynchronized node estimates the clock drift ratio and offset with the reference node to become synchronized. By periodical re-synchronization, the un-synchronization conditions such as nodes failures or topology change can be easily overcome. We implement our protocol in the Berkeley MICAz platform. The experimenting scenarios are 5-node and 18-node multi-hop topologies, and the re-synchronization periods are 30-second and 300-second. The experiment results show that the average synchronization errors of all nodes run with our protocol are ranged within several micro-seconds and are less than those of the previous protocol. Our proposed protocol uses lower communication overhead and establishes more robust synchronization situations for all nodes in the network. The synchronization accuracy required by different applications can be achieved by using different re-synchronization periods.
To obtain the node location information for mobile sensor networks, we also propose a localization scheme to improve the localization accuracy of previous work based on SMC (Sequential Monte Carlo). It operates under the assumption that a few part of sensor nodes know their positions. The valid samples to represent sensor nodes’ possible locations are filtered upon the location information from the location-aware nodes (anchor nodes) within two hops as well as that from the location-unaware nodes within one hop of which the locations have been initially estimated. In addition, we propose a moving direction predicting method to further enhance the accuracy of the location estimates. Simulation results show that our proposed localization algorithm performs better than other SMC-based algorithms in most network configurations with various mobility models and different moving speeds. Since the estimated location information of neighboring nodes unknown to their actual positions is utilized, our localization scheme can still work well for low anchor-density networks. Furthermore, each node’s number of samples in our proposed scheme is adapted according to the estimated sampling region at each time slot. Thus, the computation cost and memory occupation can be obviously decreased. To minimize the network traffic cost, we use a simple sectoring scheme to represent the possible located region of each location-unaware node. We also propose a novel moving direction constraint to refine more accurate samples for location estimate.
| en_US |