A particle filter based framework for indoor wireless localization using custom IEEE 802.15.4 nodes

Locating people close to real-time and with acceptable precision has always been an important part of any organization or industry, especially in law enforcement, manufacturing, healthcare, and logistics. Technologies that have the ability to locate objects or people are called Real Time Location Systems (RTLS). They typically use small low-power transmitters called tags attached to assets (or worn by people) as well as sets of readers that map the location of these tags. Systems that map the longitude and attitude of an object are geo-location systems and generally use the Global Positioning System (GPS) for location mapping. GPS could be used as the location determination portion of an RTLS system (relaying that information would have to rely on another system); unfortunately, GPS signals do not penetrate buildings well and thus GPS will in general not work inside buildings and in dense urban areas. Thus, there is a need for RTLS systems that work in GPS-denied environments.;Several technologies have been proposed to create Real Time Location Systems. Some use dedicated tags and readers while others use existing WLAN networks and add RTLS ability to those networks. We propose a probabilistic approach to localization, based upon Received Signal Strength (RSSI) and inertial information coming from tags (e.g., accelerometer and rotational sensor readings). Global localization is a flavor of localization in which the device is unaware of its initial position and has to determine the same from scratch.;The first step to localize tags in this work involves building a wireless measurement model of the tag with respect to some anchor nodes (access points). The model is built by measuring the RSSI readings of the mobile node relative to the access point at various distances and orientations (rotation away from the access point). These readings form a sample set for sequential Monte-Carlo sampling. Next, a posterior probability distribution for the location of the wireless device is computed over the entire area using Monte-Carlo sampling based Bayesian filtering, also known as particle filters. Location estimates may then be determined from this distribution using the maximum density point or other parameters depending on the estimate needed.;We discuss theory and research leading to the proposed method and describe the experimental hardware/firmware/software system built for the purposes of this work and provide results of real-life experiments.