Published in IET Communications Received on 25th October 2011 Revised on 30th January 2012 doi: 10.1049/iet-com.2011.0784 In Special Issue on Energy Aware Wireless Network Protocols ISSN 1751-8628 Network lifetime maximising distributed forwarding strategies in Ad Hoc wireless sensor networks B. Panigrahi 1 S. De 1 B.S. Panda 2 J.-D. Lan Sun Luk 3 1 Electrical Engineering Department, IIT, Delhi, India 2 Mathematics Department, IIT, Delhi, India 3 Universite ´ de la Re ´ union – Laboratoire LE 2 P, Saint Denis, Re ´ union, France E-mail: swadesd@ee.iitd.ac.in Abstract: The authors propose three variants of distributed and stateless forwarding strategies for wireless sensor networks, namely greedy minimum energy consumption forwarding protocol (GMFP), lifetime maximising GMFP (LM-GMFP) and variance minimising GMFP (VAR-GMFP), which aim at maximising the network lifetime while achieving a high forwarding success rate. GMFP selects a forwarding node that minimises per-packet energy consumption while maximising the forwarding progress. LM-GMFP extends the GMFP algorithm by also taking into account the remaining energy at the prospective one-hop forwarding nodes. In VAR-GMFP, on the other hand, the packet is forwarded to the next node that ensures a locally high mean and low variance of nodal remaining energy. Through simple probabilistic analysis the authors prove the intuition behind the optimum forwarding node selection for network lifetime maximisation. They then model the lifetime maximisation of a sensor network as an optimisation problem and compare the practical protocol-dependent network lifetime with the theoretical upper bound. Through extensive simulations the author demonstrate that the proposed protocols outperform the existing energy-aware protocols in terms of network lifetime and end-to-end delay. 1 Introduction 1.1 Motivation Wireless sensor networks (WSNs) have gained significant importance in recent years with many application areas, such as transportation, environmental monitoring, health care, national security and structural monitoring. A key challenge in such networks is devising system architectures to realise distributed sensing, data forwarding and aggregation tasks, subject to hard system constraints, such as limited energy. Owing to the difficult environments and a large scale of deployment, recharging or replacing the sensor nodes’ batteries may not be feasible. Since there is a high cost associated with the network maintenance caused by frequent battery drainage, energy saving in a WSN to maximise its lifetime has drawn significant attention of the researchers. As communication range of the field nodes are much smaller compared to the sensing area, field sensor to the sink communication is generally based on multihop forwarding. Also, since the field nodes have limited memory and processing capabilities, distributed control forwarding becomes an obvious choice. Furthermore, if the sensing applications are delay-tolerant, stateless forwarding is practiced, where the nodes do not need to create and maintain routing tables. Typically in the distributed forwarding protocols, local neighbourhood information and the destination location are considered available in some form, for example, through geographical positioning system (GPS) (e.g. [1]) by other virtual localisation techniques (e.g. [2, 3]). In distributed forwarding, a best relay node is decided at a transmitter from its local neighbours based on various criteria, such as the amount of energy a relay would consume, remaining energy at a candidate node, distance progress toward the destination, link quality between the transmitter and receiver, receiver buffer size etc. A common constraint faced by any distributed forwarding strategy in WSN is the wireless channel error. Pure geographic greedy forwarding protocol variants (e.g. [4 –7]) minimise the source-to-destination hop count by choosing the forwarding nodes at each hop that are as close to the destination as possible. However, this approach may not be optimal in throughput and energy consumption because of more number of retransmissions caused by channel errors. Some other energy-aware routing protocols consider either transmitter energy consumption only [8, 9], or transmitter – receiver energy consumption without accounting the channel errors [10], or energy minimisation without allowing distributed control [11]. In a pure energy-aware forwarding, at every hop the node nearest to the transmitter is selected as the forwarder, as it offers the lowest average number of transmissions per successful packet forwarding. In addition, if the remaining energy of the neighbour node is also considered, then the node with the highest remaining energy will be selected among the nearest nodes [12]. No significant performance gain can be achieved if conversely the minimum energy consuming node is selected from the IET Commun., pp. 1–11 1 doi: 10.1049/iet-com.2011.0784 & The Institution of Engineering and Technology 2012 www.ietdl.org