1536-1233 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TMC.2018.2868930, IEEE Transactions on Mobile Computing JOURNAL OF IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. XX, NO. XX, XX 2018 1 Algorithms and Position Optimization for a Decentralized Localization Platform Based on Resource-Constrained Devices Zakaria Kasmi 1 , Naouar Guerchali 2 , Abdelmoumen Norrdine 3 and Jochen Schiller 1 Abstract—As a step towards ubiquitous and mobile computing, a decentralized localization platform allows positioning for objects and persons. The decentralized computation of the position enables to shift the application-level knowledge into a Mobile Station (MS) and avoids the communication with a remote device such as a server. In addition, computing a position on resource-constrained devices is challenging due to the restricted storage, computing capacity and power supply. Therefore, we propose suitable algorithms to compute unoptimized as well as optimized positions on resource-limited MSs. Algorithms for unoptimized positions will be analyzed with respect to the stability, complexity, and memory requirements. The calculated positions are optimized by using the Gauss–Newton (GNM) or Levenberg–Marquardt methods (LVMs). We analyze and compare the GNM with two variants of the LVM algorithm. Furthermore, we develop an adaptive algorithm for the position optimization, which is based on the Singular Value Decomposition (SVD), LVM algorithm, and the Dilution of Precision. This method allows an adaptive selection mechanism for the LVM algorithm. The influence and choice of the right parameter combination of the LVM algorithm will be analyzed and discussed. Finally, we design and evaluate a method to reduce multipath errors on the MS. Index Terms—DecaWave, DWM 1000, Embedded Systems, Internet of Things, IoT, Positioning, Raspberry Pi, RIOT-OS, STM32F4, UWB. ✦ 1 I NTRODUCTION L OCALIZATION enables applications to take advantage of context-aware mobile computing related to the lo- cation of users or objects. The localization can be a value- added service for areas such as the automation of indus- trial systems and equipment, medicine, and networked mo- bility. Furthermore, it can provide contextual information to common devices such as tablets, smart phones, and other wireless devices, in order to support surveillance systems, e-commerce, health monitoring, and intelligent transportation, among others. Localization applications can be classified in the following areas: Emergency, safety and security, tracking and tracing, traffic telematics, personal navigation, management and logistics, billing (e.g., road tolling), commerce, leisure and entertainment, and enquiry and information [1]. GPS (Global Positioning System) and similar systems play an important role in applications such as car navigation or vehicle tracking and monitoring. Its accuracy is restricted in urban areas due to multipath propagation errors as well as limited visibility of the satellites especially with narrow streets and tall buildings [2]. Furthermore, although the use of pseudo-lites, which generate and transmit GPS- like signals, might enable GPS to work indoors; pseudo- lites based solutions must cope with several issues such as • 1 The authors are with the Department of Mathematics and Computer Science, Freie Universit¨ at Berlin, Takustraße 9, 14195 Berlin, Germany. E-mail: {zakaria.kasmi, jochen.schiller}@fu-berlin.de • 2 E-mail: Naouar.Guerchali@rwth-aachen.de • 3 The author is with the Institut f ¨ ur Baubetrieb, Technische Universit¨ at Darmstadt, El-Lissitzky-Straße 1, 64287 Darmstadt, Germany. E-mail: a.norrdine@baubetrieb.tu-darmstadt.de Manuscript received XX xx, 2018; revised XX xx, 2018. multipath effects, precise synchronization, and government restrictions [3], [4]. Since GPS does not work properly in indoor environ- ments, alternative localization techniques have been devel- oped for indoor positioning. Deak et al. present a survey about indoor localization systems, which use various mea- surement methods such as Time of Arrival (TOA) or Re- ceived Signal Strength Indicator (RSSI) [5]. The majority of indoor localization systems necessitate that tags/electronic devices should be mounted on objects or carried by the person being tracked in order to estimate their location [5]. The common features of Indoor Localization Systems (ILSs) are the deployment environment, system range, geometric location model, and the localization techniques [6]. Numerous technologies have been developed for ILSs, which are based on several physical principles and have different performance characteristics. Common technologies used in indoor localization are Wireless LAN (WLAN) or Ultra Wideband (UWB) that are based on electromagnetic waves. Other localization systems use Radio Frequency Identification (RFID), infrared, ultrasound, or computer vi- sion techniques. The main drawbacks of these methods are shadowing, signal propagation errors due to attenuation, multipath, signal delay or bad lighting conditions. Although certain technologies such as UWB are more robust against the mentioned effects, it is impossible to suppress signal propagation errors entirely. Since data processing plays a key role in a localization system, we briefly compare different architectures and pro- cessing techniques. The architecture of localization systems can be classified in three categories: central, decentralized and distributed. In a centralized architecture, the Mobile