DePauw University, Computer Science, Greencastle, IN 46135
Wireless sensor networks (WSN's) are a cluster of wireless and often mobile devices, called nodes, which are set up within an environment to monitor its conditions and communicate with each other or a base station. While there is a simple hierarchical structure in most sensor networks, ad hoc sensor networks do not have a base station and the nodes must devise a system of communication among themselves. In order for sensor networks to be useful, they must be able to effectively communicate with one another. This requires that all the nodes in the network be synchronized in time. Without time synchronization, wireless nodes would be unable to communicate effectively with each other and important information would be sent and received in a haphazard manner. Networks would thus be unstable and prone to attacks from enemy nodes. In this work, we propose an improvement to the existing Timing-sync Protocol for Sensor Networks (TPSN). Our contribution is three fold. First, we propose the use of the energy efficient temperature compensated crystal oscillators (TCXO's) to compensate for clock drift due to temperature changes. Second, we show the need to account for network access time during time synchronization. We performed simulations on the Network Simulator Ns2 to compute the average access, send and receive times using standard Ns2 benchmarks. Our experimental results show that the access time is significant when compared to both the send time and receive time and should therefore be taken into consideration during time synchronization. Finally, we designed an algorithm for multi-tier time synchronization. Our algorithm allows nodes to synchronize with the lower level nodes in an efficient manner and prevents network congestion.
Traditionally, clock drift is corrected by having every pair of nodes exchange two messages with time stamps and computing an average of the time difference. This method can put a huge load on the resource constrained sensor nodes by increasing the traffic in the network. In addition, it can also drain the power resources of the nodes. We propose the use of the energy efficient temperature compensated crystal oscillators (TCXO's) to prevent clock drift due to temperature changes. These oscillators contain a temperature compensation circuit that senses temperature and accounts for the effect of the temperature change on the oscillator.
Currently, time synchronization protocols pay little or no attention to access time and either describe it as non-deterministic or zero. Since time synchronization protocols estimate the time difference between nodes by sending time-stamped messages, it is important to consider access time which constitutes a significant portion of the latency of the message. When nodes want to account for the difference in their timestamps, they must subtract the time it takes for the message to be sent (access time) from their actual timestamp. In current literature, access time is not included in synchronization, therefore nodes were not accurately synchronized. Using Ns2, we were able to obtain the access time, send time, and receive time using standard wireless benchmarks. Furthermore, we computed propagation time for distances ranging from 10 meters to 100 meters (the maximum wireless range of sensor nodes) and concluded that while most protocols pay more attention to propagation time, it is in fact negligible (at most 0.2 µs) compared to the access time which is in the order of a few milliseconds.
Lastly, we have developed a multi-hop synchronization algorithm for sensor networks. Our proposed algorithm has a discovery phase (similar to TPSN's discovery phase) where nodes are given levels depending on their proximity from the root node (level 0). Nodes within transmission range of the root node are assigned level 1, nodes within transmission range of level 1 are assigned level 2, etc. The root node broadcasts a message to synchronize all nodes at level 1. While level 1 nodes are being synchronized, the nodes in level 2 are communicating with their neighbors in the same level. From this information, the level 2 nodes can group together so that each group can receive a broadcast from a single node from level 1. By doing this, it is possible that some nodes in level 1 will never broadcast synchronization messages to nodes in level 2, which reduces power consumption in multi-hop synchronization.
[Abstract (DOC)]