Distributed parameter and state estimation in a network of sensors

A wireless sensor network (WSN) consists of spatially distributed autonomous devices using sensors connected via a wireless link. Sensors may be designed for pressure, temperature, sound, vibration, motion... Initially WSN were developed for military applications (battlefield surveillance). Now, many civilian applications (environment monitoring, home automation, traffic control) may take advantage of WSN, see, e.g., [KM04, Hae06]. Applications suggest many research topics, from the design of protocols for communication between sensors, localization problems, data compression and aggregation, security issues... All these problems are made more complicated by the constraints imposed on each node of the WSN, which usually has limited computing capabilities, communication capacity and a very restricted power consumption. Here, the application we consider is WSN for source tracking, which may be important when considering mobile phone localization and tracking, computer localization in an ad-hoc networks, co-localisation in a team of robots, speaker localization... Figure 1 illustrates a typical localization problem: a source represented by a circle moves in a field of sensors, each of which is represented by a cross. The localization technique used depends on the type of information available to the sensor nodes. Time of arrival (TOA), time difference of arrival (TDOA) and angle of arrival (AOA) usually provide the best results [PAK05], however, these quantities are most difficult to obtain, as they require, a good synchronization between timers (for TOA), exchanges between sensors (for TDOA) or multiple antennas (for AOA). Contrary to TOA, TDOA or AOA data, readings of signal strength (RSS) at a given sensor are easily obtained, as they only require low-cost sensors or are already available, as in IEEE 802.11 wireless networks, where these data are provided by the MAC layer [STK05].