Is Cooperative Localization in Wireless Body Area Networks Accurate Enough for Motion Capture Applications? Jihad Hamie Institut Mines-Telecom Telecom ParisTech CNRS LTCI UMR 5141 46 rue Barrault 75634 Paris Cedex 13, France Email: jihad.hamie@telecom-paristech.fr Anis Ouni Institut Mines-Telecom Telecom ParisTech CNRS LTCI UMR 5141 46 rue Barrault 75634 Paris Cedex 13, France Email: anis.ouni@telecom-paristech.fr Claude Chaudet Institut Mines-Telecom Telecom ParisTech CNRS LTCI UMR 5141 46 rue Barrault 75634 Paris Cedex 13, France Email: claude.chaudet@telecom-paristech.fr Abstract— The applications of Wireless Body Area Networks (WBANs) encompass several uses such as home activity monitor- ing, postural rehabilitation, or sportive gesture recording, which require motion capture. Such motion capture can be realized by acquiring the positions of on-body nodes, which can be deduced from device-to-device distance measurements. In distributed wireless networks, this ranging is generally realized by measuring characteristics of the radio channel between couples of nodes, such as the signal attenuation on data packets, or their time of flight. These methods have a limited accuracy, though, because of the variability of the radio channel, and produce positions estimations that are usually too coarse for motion capture. In this paper, we present a cooperative solution that improves the localization of such wireless on-body nodes in a Local Coordinate System (LCS). We adapt a cooperative Extended Kalman Filter (EKF) algorithm fed by inter-node range measure- ments through impulse Radio-Ultra Wideband (IR-UWB) Time Of Arrival (TOA) estimation. The cooperative approach relies on the multiplicity of links that results from the presence of multiple nodes on the body. Such collaboration provides spatial diversity and measurements redundancy that have the potential to improve the positioning accuracy. To evaluate this approach, we developed a complete protocol stack from the physical up to the localization application layer, in a discret-event simulator (WSNet). A Time Division Multiple Access (TDMA) Medium Access Control (MAC) is developed for maintaining the peer-to- peer handshake transactions between devices. Simulation results for various connectivity levels, show interesting gains on the average location precision of each node. I. I NTRODUCTION Wireless Body Area Networks (WBANs) are formed by small wireless and embedded devices that are located on or close to a person’s body. Numerous applications are envisioned in various domains such as healthcare, wellness, security, sports or gaming. For various considerations, including the proximity with the human body, wireless communications generally happen at low power. Many applications in these domains could benefit from an accurate positioning of the nodes that follow the movements of the body. An accurate enough devices positioning system could even represent a credible alternative for motion capture applications with re- spect to video acquisition systems that are costly and require a physical installation, or to the use of inertial and magnetic sensors. The upcoming standards relevant for WBAN, such as IEEE 802.15.6 [1] or to some extent IEEE 802.15.4a, all specify an Impulse Radio-Ultra Wideband (IR-UWB) physical layer. Such technology is particularly appealing for radio-based ranging, as its high temporal and fine multi-path resolution capabilities permits the measurement of the Round Trip - Time Of Flight (RT-TOF) of transmitted signals, which yields to a far better accuracy than the signal attenuation-based approaches [2], [3]. The question of localization of the on-body nodes using radio signals however remains difficult considering that, for these power levels, the body itself constitutes a noticeable obstacle. The movements of the body will change the char- acteristics of each wireless link, passing from line of sight to non-line of sight situations and vice-versa. In the traditional lo- calization process, each device positions itself by measuring its distance only with respect to a set of anchor nodes, i.e. nodes whose positions are supposed to be known and fixed. Given the variability of the links, the number of anchors required to reach a good precision is much higher that the theoretical numbers in a perfect propagation environment. Collaborative approaches that rely on the collection, processing and use of all possible device-to-device range measurements appear as viable alternatives in this context, as they increase spatial diversity and measurements redundancy, which could be utilized to alleviate the effects of propagation conditions changes. But applying cooperative localization to the WBAN context is not straightforward. The large dataset still needs to be processed and potentially filtered to remove abnormal mea- surements. Besides, most of the evaluations today have been realized with computer simulations that rely on simplified physical abstractions, as few experimental platforms include IEEE 802.15.6-compliant hardware or use IR-UWB physical