Sensor networks are transforming the way we deal with the physical world with applications ranging from enviromental monitoring to security and health-care. Each application has different requirements and constraints which have to be carefully designed for. However, three broad needs are common to most sensor networks: energy efficiency, speed of data collection and reliability, for example to node failures or malicious attacks. Our research program addresses these broad neds by developping fundamentally new techniques. We are currently validating these ideas in a testbed of Tinyos sensor nodes.
In a significant class of sensor-network applications, the identities of the reporting sensors constitute the bulk of the communicated data, whereas the message itself can be as small as a single bit. For instance, sensors used for monitoring whether and at what location an environmental hazardous situation may occur need to send their location and very few bits, albeit at very short time intervals, to convey as fast as possible the alarm information. We are interested in addressing the question of how should communication happen in such identity-aware sensor networks. To do so, we need to re-examine the traditional source-identity/message separation and propose a scheme for jointly encoding the two, through a specifically designed coding scheme. We have recently developed a communication method for identity-aware sensor networks and show it to be energy efficient, simple to implement, and gracefully adaptable to scenarios frequently encountered in sensor networks—for instance, node failures, or large numbers of nodes where only few are active during each reporting round. We are planning to extend these ideas to sensor networks that convey GSM (location) information, to infrastracture sensor networks for transportation networks.
In many situations, what one does not say conveys information, just as what one might say. This idea can be used in wireless sensor networks by utilizing silence to convey information and thereby reducing the energy used. We explored this in the context of function computation in wireless sensor networks. We develop provable protocols that allow information to be inferred by the fact that nodes remain silent. We thus use silence and time to minimize the amount of communication required to compute various functions. We investigate the time-complexity trade-off such protocols offer, in a communication-complexity framework. We focus our attention on symmetric functions (where the ordering of the variables do not matter), that include most statistical functions and functions of interest for sensor networks. For such functions we develop methods that are optimal (worst case and average communication complexity). We believe that these ideas are very promising for health-monitoring sensor networks (for example, on-the-body sensors), where minimizing the number of node transmissions can also reduce health risks, and this is a direction we plan to further pursue.
Multipath diversity techniques in sensor networks construct multiple paths from each node sensor to the sink, allowing more freedom in route selection, and thus achieving higher reliability; less intuitively, under certain circumstances, they even achieve lower energy consumption. Network coding allows to achieve the benefits of multipath diversity, namely, increased reliability, by opportunistically taking advantage of broadcasting, at a significant savings of the network resources. Our first proposed techniques are currently being deployed in a TinyOs prototype at EPFL. We plan to continue working both on the theoretical foundations as well as prototype implementations of these ideas, as we believe they can have a significant impact in future sensor networks.