ENERGY-AWARE ADAPTIVE LOW POWER LISTENING FOR SENSOR NETWORKS Raja Jurdak, Pierre Baldi and Cristina Videira Lopes School of Information and Computer Sciences California Institute for Telecommunications and Information Technology Cal-(IT) 2 University of California, Irvine, CA 92697 ABSTRACT Energy efficiency is a central issue for all wireless sensor network applications. Idle listening on the wireless channel constitutes a large portion of overall energy consumption in sensor networks. The need to reduce idle listening energy consumption has led recently to the design of BMAC [1], a MAC protocol for sensor networks which provides 8 low power listening modes and 8 corresponding transmit modes. Here, we propose a cross-layer mechanism called Energy Aware Adaptive Low Power Listening (EA-ALPL) that enables each individual sensor node running BMAC to set its own listening mode according to its duty cycle and its number of descendants in the routing tree. EA-ALPL also enables nodes to learn the listening modes of their neighbors and to choose the appropriate transmit mode in order to ensure correct packet reception. We show through deployment experiments that EA-ALPL yields overall power savings ranging between 16% and 55% depending on a node's logical topology position. 1. INTRODUCTION Advances in processor, memory, communication and sensing technology have fueled increased interest in sensor networks. Sensor networks have a wide range of military, civilian, and environmental applications. Regardless of the application, sensor networks will require unattended network operation for months or years, typically with limited battery resources at each node. Consequently, energy efficiency is a central issue in all sensor network applications. Many protocols have been designed to provide energy- efficient behavior at both the MAC layer [2] and the routing layer [3]. At the MAC layer, idle listening constitutes a large portion of power consumption because data is sent infrequently. This effect is even more pronounced in monitoring sensor networks [4]. Thus, energy-efficient MAC protocol proposals have focused on minimizing idle listening at sensor nodes [1,5]. The recent work by Pollastre et al. describes a new sensor network MAC protocol called BMAC [1], which aims at reducing idle listening at sensor nodes. BMAC proposes that each node wake up periodically to check for channel activity. The wake-up period is referred to as the check interval. BMAC defines 8 check intervals, and each check interval corresponds to one of BMAC's 8 listening modes. To ensure that all packets are heard by the nodes, packets are sent with a preamble whose reception time is longer than the check interval. BMAC therefore defines 8 different preamble lengths referred to as transmit modes. Additionally, Pollastre et al. analytically derive optimal listening modes based on the number of neighbors of a node. In their experiments, they determine the maximum neighborhood size in the network, and they set the optimal listening mode for that neighborhood size. The experimental results yield significant energy savings for BMAC over previous protocols. The other major strengths of BMAC are its modularity and flexibility. BMAC provides interfaces that are accessible to higher layer protocols and applications to set listening and transmit modes on a per-packet basis if needed. Pollastre et al. also suggest that using these interfaces to set listening and transmit modes according to additional information on the application and operation of a sensor network could produce further power savings for BMAC. Building on BMAC, our work proposes a cross-layer mechanism called Energy Aware Adaptive Low Power Listening (EA-ALPL) to better adapt to dynamic sensor network topologies and nonuniform energy consumption. In BMAC, setting a network-wide listening mode disregards the non-uniform and dynamic local states of individual nodes. EA-ALPL enables each sensor node to set its own listening mode according to its local state. Per node listening modes are more energy-efficient, but it is difficult to predict the state of each node prior to deployment. To address these challenges, EA-ALPL introduces the following novel contributions: 1. It enables each node to set its own listening mode based on its current state information. 2. It allows each node to dynamically learn the listening mode of its routing parent in order to locally set the appropriate transmit mode. 3. It proposes the dependence of listening mode on both topology-related information and duty cycle at a node. The dependence of the listening mode on a node's topology position ensures adaptability to a dynamic sensor network topology. EA-ALPL reduces idle listening at each node by selecting the optimal listening mode for the node's current number of descendants in the routing tree.