CRP: Cluster Head Re-election Protocol for heterogeneous wireless sensor networks Ravi Tandon Sukumar Nandi Abstract—Sensor nodes in a network consume energy in a non-uniform manner. Designing clustering proto- cols that are heterogeneity aware is still an open issue in sensor networks. This work focuses on hierarchically clustered heterogeneous sensor networks. Sensor nodes organize themselves in self organized groups called clusters. Each cluster consists of a cluster head and some member nodes. The sensor network consists of nodes of two different energy levels. We show that current heterogeneity aware protocols (such as SEP [1], DEEC [2] etc.) are unable to distribute the usage of energy amongst the sensor nodes uniformly. We propose CRP, Cluster Re-election Protocol, that is heterogeneity aware and enhances network stability period (period before which the first node dies) over current protocols. We show, by simulation, that CRP improves network stability by reducing the variation in the residual energy of sensor nodes. Therefore, CRP is able to exploit the heterogeneity present in a sensor network in an efficient manner. Index Terms—Clustering algorithms, Wireless sensor networks. I. Introduction Wireless sensor networks are autonomous systems of tiny, battery powered sensor nodes with limited power and radio capabilities. These sensor nodes are widely deployed for various data collection applications such as health monitoring, military surveillance, target tracking etc. Limited battery life of these sensor nodes is one of the primary concerns. Recharging batteries is not pos- sible due to either a large scale network deployment or inaccessibility of the region in which they are deployed. Therefore, improving the lifetime of a sensor network becomes a major issue. Sensor network clustering and data aggregation are certain techniques that reduce energy usage and thereby enhance the lifetime of sensor networks. Clustering is a technique which partitions the sensor network into smaller groups called clusters. Each cluster consists of cluster head and member nodes. Member nodes collect data from their environment and send it to their respective cluster heads. Due to a high redundancy in data collected by proximally located sensor nodes, cluster heads compress data into a single message and send it to the Base Station. This technique is referred to as Data Aggregation. Clustering, therefore, results in unequal distribution of energy used by sensor nodes. Variation in distance of cluster heads from the Base Station, non uniform topological features of terrain and disparities in radio characteristics such as transmission losses lead to dif- ferences in energy usage amongst sensor nodes. Protocols such as Direct Transmission and Minimum Transmission Energy (MTE) also lead to disproportionate energy usage amongst sensor nodes. Direct Transmission protocols lead to depletion of energy in sensor nodes that are farther away from the Base Station. MTE protocols lead to a depletion of energy in sensor nodes that are proximal to the Base Station. Heterogeneity, therefore, is a natural consequence of network operation. Heterogeneity also finds application in certain techniques of energizing sensor nodes by introduction of high energy sensor nodes. This work considers heterogeneity of energy in sensor nodes. The sensor network is assumed to consist of two types of nodes – high energy nodes (advanced nodes) and low energy nodes (plain nodes). While some of the current het- erogeneity aware clustering protocols elect cluster heads stochastically (based on their initial residual energy e.g. SEP) others require global knowledge (e.g. LEACH-E), or are based on a scheme that predict residual energy within the sensor network (e.g. SEP-E). This study throws light up on the drawbacks of such schemes and proposes a protocol that elects cluster heads optimally. Cluster head election is based on the current residual energy and requires local knowledge of residual energy of sensor nodes. The protocol is referred to as Cluster Head Re- election Protocol (CRP). Rest of the paper is organized as follows. Section II outlines some of the earlier proposals that deal with the issue of energy efficient clustering in sensor networks. Section III describes the energy model, network model and parameters taken into consideration for analysis of CRP. Section IV describes the working of CRP protocol and the philosophy behind its design. Section V presents an analysis of time, space and message complexity of CRP. Section VI presents the simulation study and results. Section VII concludes this work with prospects for future research study. II. Related Work Significant research efforts have been devoted towards solving the problem of non-uniform energy consumption in sensor networks. Sensor network protocols such as [1]–[4] deal with the issue of heterogeneity in energy of sensor nodes. Such heterogeneity aware protocols elect sensor nodes with more residual energy as cluster heads with higher probability. Other protocols [5]–[12] rely on unequal