This project is focused on the study of the energy hole problem in multi-hop wireless sensor networks.

· Overlapping neighboring Layers Scheme (ITNG08)
      In our previous work, a highly scalable and fault-tolerant network architecture, named as the Progressive Multi-hop Rotational Clustered (PMRC) structure is proposed, which is suitable for the construction of large-scale wireless sensor networks (WSN). In the PMRC structure, sensor nodes are partitioned into layers according to their distances to the sink node. A cluster is composed of the nodes located in one layer and the cluster head in the upper layer closer to the sink node. The cluster head is responsible for forwarding data to its upstream layer. As traffic load is concentrated on the cluster heads closer to the sink node, these cluster heads will deplete their power faster, which limit the lifetime of the whole network. This is referred as the energy hole problem of the PMRC-based WSN.
     This problem can be solved by overlapping neighboring layers (OL). By this way, more cluster head candidates are available for each layer and the intra-cluster communication energy can be reduced, which ultimately helps prolonging network life time. Through analysis and numeric results, the reasonable overlapped ranges are decided such that the energy consumption of among the cluster heads of different layers is balanced. Simulation results with the selected overlapped ranges confirm that overlapping neighboring layers balances the energy consumption among cluster heads of different layers and prolongs network life time.

· Minimum Overlapping Layers Scheme (CCNC09)
     The minimum overlapping layers (MOL) scheme is proposed to mitigate this problem and overcome the limitation imposed by the fixed overlap range of the OL scheme. We assume that a node is eligible to be elected as a cluster head only if its residual energy is higher than the pre-defined threshold. The initial overlap area between neighbor layers may be reduced to include only the initial cluster heads of the clusters in the downstream layer. When the residual energy of one initial cluster head falls below the threshold, it will be deliberately "pushed" to its downstream layer by increasing its layer number by one. The result is that the overlap between these two neighbor layers may grow a little larger towards the upstream layer direction on the next round of network formation. Consequently, throughout the network lifetime, minimum overlap between neighbor layers is kept and the cluster size is dynamically changed. A variant of MOL, the MOL with Initial Overlap (MOLIO) scheme is also proposed. The simulation results of the OL, MOL, and MOLIO schemes show that the MOL scheme significantly prolongs the network lifetime than the OL scheme for most transmission ranges and the MOLIO scheme achieves better results than the MOL scheme at larger transmission ranges.

· Load-Similar Node Distribution (FGNC09)
     As the traffic follows a many-to-one pattern, the network lifetime of PMRC-based WSNs using the MOL scheme is still limited by the number of sensor nodes in the initial first layer. As pointed in Wu¡¯s paper, node distribution strategy is rather important in balancing energy consumption in multi-hop static sensor networks with single sink node. Unfortunately, existing schemes for the energy hole problem in WSNs, assume that the network only lasts to the time when the first node is dead and the layer boundary is fixed. These schemes may not be suitable for MOL-enabled PMRC-based WSNs in which the layer boundary is dynamically changed during network lifetime.

Figure 1 Average load vs. d.

     In this work, we propose a load-similar node distribution strategy to solve the energy hole problem for the PMRC-based WSNs. First, the load analysis in the continuous space of the network is performed, which reflects the gradual change of the layer boundary in the MOL scheme (as shown in Fig. 1). Then based on the analysis, the load-similar node distribution strategy is proposed. Fig. 2 illustrates a sample load-similar distribution. Simulation results confirm the superiority of the proposed load-similar node distribution over the nonuniform node distribution and uniform node distribution strategies for MOL-enabled PMRC-based WSNs.

Figure 2 A sample load-similar distribution