Nonlinearities and timescales in neural models of temporal interference stimulation.

In temporal interference (TI) stimulation, neuronal cells react to two interfering sinusoidal electric fields with a slightly different frequency ( f 1 ${f}_{1}$ , f 2 ${f}_{2}$ in the range of about 1-4 kHz, ∣ f 1 - f 2 ∣ $| {f}_{1}-{f}_{2}| $ in the range of about 1-100 Hz). It has been previously observed that for the same input intensity, the neurons do not react to a purely sinusoidal signal at f 1 ${f}_{1}$ or f 2 ${f}_{2}$ . This study seeks a better understanding of the largely unknown mechanisms underlying TI neuromodulation. To this end, single-compartment models are used to simulate computationally the response of neurons to the sinusoidal and TI waveform. This study compares five different neuron models: Hodgkin-Huxley (HH), Frankenhaeuser-Huxley (FH), along with leaky, exponential, and adaptive-exponential integrate-and-fire (IF). It was found that IF models do not entirely reflect the experimental behavior while the HH and FH model did qualitatively replicate the observed neural responses. Changing the time constants and steady state values of the ion gates in the FH model alters the response to both the sinusoidal and TI signal, possibly reducing the firing threshold of the sinusoidal input below that of the TI input. The results show that in the modified (simplified) model, TI stimulation is not qualitatively impacted by nonlinearities in the current-voltage relation. In contrast, ion channels have a significant impact on the neuronal response. This paper offers insights into neuronal biophysics and computational models of TI stimulation.

In Situ Assessment of Uplink Duty Cycles for 4G and 5G Wireless Communications.

In this presented study, we measured in situ the uplink duty cycles of a smartphone for 5G NR and 4G LTE for a total of six use cases covering voice, video, and data applications. The duty cycles were assessed at ten positions near a 4G and 5G base-station site in Belgium. For Twitch, VoLTE, and WhatsApp, the duty cycles ranged between 4% and 22% in time, both for 4G and 5G. For 5G NR, these duty cycles resulted in a higher UL-allotted time due to time division duplexing at the 3.7 GHz frequency band. Ping showed median duty cycles of 2% for 5G NR and 50% for 4G LTE. FTP upload and iPerf resulted in duty cycles close to 100%.

Location-Aware Range-Error Correction for Improved UWB Localization.

In this paper, we present a novel localization scheme, location-aware ranging correction (LARC), to correct ranging estimates from ultra wideband (UWB) signals. Existing solutions to calculate ranging corrections rely solely on channel information features (e.g., signal energy, maximum amplitude, estimated range). We propose to incorporate a preliminary location estimate into a localization chain, such that location-based features can be calculated as inputs to a range-error prediction model. This way, we can add information to range-only measurements without relying on additional hardware such as an inertial measurement unit (IMU). This improves performance and reduces overfitting behavior. We demonstrate our LARC method using an open-access measurement dataset with distances up to 20 m, using a simple regression model that can run purely on the CPU in real-time. The inclusion of the proposed features for range-error mitigation decreases the ranging error 90th percentile (P90) by 58% to 15 cm (compared to the uncorrected range error), for an unseen trajectory. The 2D localization P90 error is improved by 21% to 18 cm. We show the robustness of our approach by comparing results to a changed environment, where metallic objects have been moved around the room. In this modified environment, we obtain a 56% better P90 ranging performance of 16 cm. The 2D localization P90 error improves as much as for the unchanged environment, by 17% to 18 cm, showing the robustness of our method. This method evolved from the first-ranking solution of the 2021 and 2022 International Conference on Indoor Position and Indoor Navigation (IPIN) Competition.

50 Hz Temporal Magnetic Field Monitoring from High-Voltage Power Lines: Sensor Design and Experimental Validation.

A low-cost, tri-axial 50 Hz magnetic field monitoring sensor was designed, calibrated and verified. The sensor was designed using off-the-shelf components and commercially available coils. It can measure 50 Hz magnetic fields originating from high-voltage power lines from 0.08 µT to 364 µT, divided into two measurement ranges. The sensor was calibrated both on-board and in-lab. The on-board calibration takes the circuit attenuation, noise and parasitic components into account. In the in-lab calibration, the output of the developed sensor is compared to the benchmark, a narrowband EHP-50. The sensor was then verified in situ under high-voltage power lines at two independent measurement locations. The measured field values during this validation were between 0.10 µT and 13.43 µT, which is in agreement with other reported measurement values under high-voltage power lines in literature. The results were compared to the benchmark, for which average deviations of 6.2% and 1.4% were found, at the two independent measurement locations. Furthermore, fields up to 113.3 µT were measured in a power distribution sub-station to ensure that both measurement ranges were verified. Our network, four active sensors in the field, had high uptimes of 96%, 82%, 81% and, 95% during a minimum 3-month interval. In total, over 6 million samples were gathered with field values that ranged from 0.08 µT to 45.48 µT. This suggests that the proposed solution can be used for this monitoring, although more extensive long-term testing with more sensors is required to confirm the uptime under multiple circumstances.

Robust IMU-Based Mitigation of Human Body Shadowing in UWB Indoor Positioning.

Ultra-wideband (UWB) indoor positioning systems have the potential to achieve sub-decimeter-level accuracy. However, the ranging performance degrades significantly under non-line-of-sight (NLoS) conditions. The detection and mitigation of NLoS conditions is a complex problem and has been the subject of many works over the past decades. When localizing pedestrians, human body shadowing (HBS) is a particular and specific cause of NLoS. In this paper, we present an HBS mitigation strategy based on the orientation of the body and tag relative to the UWB anchors. Our HBS mitigation strategy involves a robust range error model that interacts with a tracking algorithm. The model consists of a bank of Gaussian Mixture Models (GMMs), from which an appropriate GMM is selected based on the relative body-tag-anchor orientation. The relative orientation is estimated by means of an inertial measurement unit (IMU) attached to the tag and a candidate position provided by the tracking algorithm. The selected GMM is used as a likelihood function for the tracking algorithm to improve localization accuracy. Our proposed approach was realized for two tracking algorithms. We validated the implemented algorithms on dynamic UWB ranging measurements, which were performed in an industrial lab environment. The proposed algorithms outperform other state-of-the-art algorithms, achieving a 37% reduction of the p75 error.

Evaluating 60 GHz FWA Deployments for Urban and Rural Environments in Belgium.

Fixed wireless access (FWA) provides a solution to compete with fiber deployment while offering reduced costs by using the mmWave bands, including the unlicensed 60 GHz one. This paper evaluates the deployment of FWA networks in the 60 GHz band in realistic urban and rural environment in Belgium. We developed a network planning tool that includes novel backhaul based on the IEEE 802.11ay standard with multi-objective capabilities to maximise the user coverage, providing at least 1 Gbps of bit rate while minimising the required network infrastructure. We evaluate diverse serving node locations, called edge nodes (EN), and the impact of environmental factors such as rain and vegetation on the network design. Extensive simulation results show that defining a proper EN's location is essential to achieve viable user coverage higher than 95%, particularly in urban scenarios where street canyons affect propagation. Rural scenarios require nearly 75 ENs per km2 while urban scenarios require four times (300 ENs per km2) this infrastructure. Finally, vegetation can reduce the coverage by 3% or increment infrastructure up to 7%, while heavy rain can reduce coverage by 5% or increment infrastructure by 15%, depending on the node deployment strategy implemented.

MmWave Physical Layer Network Modeling and Planning for Fixed Wireless Access Applications.

The large bandwidths that are available at millimeter-wave frequencies enable fixed wireless access (FWA) applications, in which fixed point-to-point wireless links are used to provide internet connectivity. In FWA networks, a wireless mesh is created and data are routed from the customer premises equipment (CPE) towards the point of presence (POP), which is the interface with the wired internet infrastructure. The performance of the wireless links depends on the radio propagation characteristics, as well as the wireless technology that is used. The radio propagation characteristics depend on the environment and on the considered frequency. In this work, we analyzed the network characteristics of FWA networks using radio propagation models for different wireless technologies using millimeter-wave (mmWave) frequencies of 28 GHz, 60 GHz, and 140 GHz. Different scenarios and environments were considered, and the influence of rain, vegetation, and the number of subscribers was investigated. A network planning algorithm is presented that defines a route for each CPE towards the POP based on a predefined location of customer devices and considering the available capacity of the wireless links. Rain does not have a considerable effect on the system capacity. Even though the higher frequencies exhibit a larger path loss, resulting in a lower power of the received signal, the larger bandwidths enable a higher channel capacity.

Optimal Frequency and Wireless Power Budget for Miniature Receivers in Obese People.

This study investigates wireless power transfer for deep in-body receivers, determining the optimal frequency, power budget, and design for the transmitter and receiver. In particular, the focus is on small, in-body receivers at large depths up to 20 cm for obese patients. This enables long-term monitoring of the gastrointestinal tract for all body types. Numerical simulations are used to investigate power transfer and losses as a function of frequency and to find the optimal design at the selected frequency for an obese body model. From all ISM-frequencies in the investigated range (1 kHz-10 GHz), the value of 13.56 MHz yields the best performance. This optimum corresponds to the transition from dominant copper losses in conductors to dominant losses in conductive tissue. At this frequency, a transmitting and receiving coil are designed consisting of 12 and 23 windings, respectively. With a power transfer efficiency of 2.70×10-5, 18 µW can be received for an input power of 0.68 W while still satisfying exposure guidelines. The power transfer is validated by measurements. For the first time, efficiency values and the power budget are reported for WPT through 20 cm of tissue to mm sized receivers. Compared to WPT at higher frequencies, as commonly used for small receivers, the proposed system is more suitable for WPT to large depths in-body and comes with the advantage that no focusing is required, which can accommodate multiple receivers and uncertainty about receiver location more easily. The received power allows long-term sensing in the gastrointestinal tract by, e.g., temperature, pressure, and pH sensors, motility sensing, or even gastric stimulation.

RF-EMF Exposure near 5G NR Small Cells.

Of particular interest within fifth generation (5G) cellular networks are the typical levels of radiofrequency (RF) electromagnetic fields (EMFs) emitted by 'small cells', low-power base stations, which are installed such that both workers and members of the general public can come in close proximity with them. In this study, RF-EMF measurements were performed near two 5G New Radio (NR) base stations, one with an Advanced Antenna System (AAS) capable of beamforming and the other a traditional microcell. At various positions near the base stations, with distances ranging between 0.5 m and 100 m, both the worst-case and time-averaged field levels under maximized downlink traffic load were assessed. Moreover, from these measurements, estimates were made of the typical exposures for various cases involving users and non-users. Comparison to the maximum permissible exposure limits issued by the International Commission on Non-Ionizing Radiation Protection (ICNIRP) resulted in maximum exposure ratios of 0.15 (occupational, at 0.5 m) and 0.68 (general public, at 1.3 m). The exposure of non-users was potentially much lower, depending on the activity of other users serviced by the base station and its beamforming capabilities: 5 to 30 times lower in the case of an AAS base station compared to barely lower to 30 times lower for a traditional antenna.

Comparison of Low-Cost 5G Electromagnetic Field Sensors.

This paper compares different low-cost sensors that can measure (5G) RF-EMF exposure. The sensors are either commercially available (off-the-shelf Software Defined Radio (SDR) Adalm Pluto) or constructed by a research institution (i.e., imec-WAVES, Ghent University and Smart Sensor Systems research group (S³R), The Hague University of Applied Sciences). Both in-lab (GTEM cell) and in-situ measurements have been performed for this comparison. The in-lab measurements tested the linearity and sensitivity, which can then be used to calibrate the sensors. The in-situ testing confirmed that the low-cost hardware sensors and SDR can be used to assess the RF-EMF radiation. The variability between the sensors was 1.78 dB on average, with a maximum deviation of 5.26 dB. Values between 0.09 V/m and 2.44 V/m were obtained at a distance of about 50 m from the base station. These devices can be used to provide the general public and governments with temporal and spatial 5G electromagnetic field values.

Radio-frequency exposure of the yellow fever mosquito (A. aegypti) from 2 to 240 GHz.

Fifth generation networks (5G) will be associated with a partial shift to higher carrier frequencies, including wavelengths comparable in size to insects. This may lead to higher absorption of radio frequency (RF) electromagnetic fields (EMF) by insects and could cause dielectric heating. The yellow fever mosquito (Aedes aegypti), a vector for diseases such as yellow and dengue fever, favors warm climates. Being exposed to higher frequency RF EMFs causing possible dielectric heating, could have an influence on behavior, physiology and morphology, and could be a possible factor for introduction of the species in regions where the yellow fever mosquito normally does not appear. In this study, the influence of far field RF exposure on A. aegypti was examined between 2 and 240 GHz. Using Finite Difference Time Domain (FDTD) simulations, the distribution of the electric field in and around the insect and the absorbed RF power were found for six different mosquito models (three male, three female). The 3D models were created from micro-CT scans of real mosquitoes. The dielectric properties used in the simulation were measured from a mixture of homogenized A. aegypti. For a given incident RF power, the absorption increases with increasing frequency between 2 and 90 GHz with a maximum between 90 and 240 GHz. The absorption was maximal in the region where the wavelength matches the size of the mosquito. For a same incident field strength, the power absorption by the mosquito is 16 times higher at 60 GHz than at 6 GHz. The higher absorption of RF power by future technologies can result in dielectric heating and potentially influence the biology of this mosquito.

Estimated all-day and evening whole-brain radiofrequency electromagnetic fields doses, and sleep in preadolescents.

OBJECTIVE: To investigate the association of estimated all-day and evening whole-brain radiofrequency electromagnetic field (RF-EMF) doses with sleep disturbances and objective sleep measures in preadolescents. METHODS: We included preadolescents aged 9-12 years from two population-based birth cohorts, the Dutch Generation R Study (n = 974) and the Spanish INfancia y Medio Ambiente Project (n = 868). All-day and evening overall whole-brain RF-EMF doses (mJ/kg/day) were estimated for several RF-EMF sources including mobile and Digital Enhanced Cordless Telecommunications (DECT) phone calls (named phone calls), other mobile phone uses, tablet use, laptop use (named screen activities), and far-field sources. We also estimated all-day and evening whole-brain RF-EMF doses in these three groups separately (i.e. phone calls, screen activities, and far-field). The Sleep Disturbance Scale for Children was completed by mothers to assess sleep disturbances. Wrist accelerometers together with sleep diaries were used to measure sleep characteristics objectively for 7 consecutive days. RESULTS: All-day whole-brain RF-EMF doses were not associated with self-reported sleep disturbances and objective sleep measures. Regarding evening doses, preadolescents with high evening whole-brain RF-EMF dose from phone calls had a shorter total sleep time compared to preadolescents with zero evening whole-brain RF-EMF dose from phone calls [-11.9 min (95%CI -21.2; -2.5)]. CONCLUSIONS: Our findings suggest the evening as a potentially relevant window of RF-EMF exposure for sleep. However, we cannot exclude that observed associations are due to the activities or reasons motivating the phone calls rather than the RF-EMF exposure itself or due to chance finding.

Geodesic Path Model for Indoor Propagation Loss Prediction of Narrowband Channels.

Indoor path loss models characterize the attenuation of signals between a transmitting and receiving antenna for a certain frequency and type of environment. Their use ranges from network coverage planning to joint communication and sensing applications such as localization and crowd counting. The need for this proposed geodesic path model comes forth from attempts at path loss-based localization on ships, for which the traditional models do not yield satisfactory path loss predictions. In this work, we present a novel pathfinding-based path loss model, requiring only a simple binary floor map and transmitter locations as input. The approximated propagation path is determined using geodesics, which are constrained shortest distances within path-connected spaces. However, finding geodesic paths from one distinct path-connected space to another is done through a systematic process of choosing space connector points and concatenating parts of the geodesic path. We developed an accompanying tool and present its algorithm which automatically extracts model parameters such as the number of wall crossings on the direct path as well as on the geodesic path, path distance, and direction changes on the corners along the propagation path. Moreover, we validate our model against path loss measurements conducted in two distinct indoor environments using DASH-7 sensor networks operating at 868 MHz. The results are then compared to traditional floor-map-based models. Mean absolute errors as low as 4.79 dB and a standard deviation of the model error of 3.63 dB is achieved in a ship environment, almost half the values of the next best traditional model. Improvements in an office environment are more modest with a mean absolute error of 6.16 dB and a standard deviation of 4.55 dB.

Combined Antenna-Channel Modeling for the Harsh Horse Hoof Environment.

This paper describes the exploration of the combined antenna-channel model for a horse hoof. An antenna of 25 mm × 40 mm is designed in the ISM 868 MHz band. During the characterization and design of the antenna, the dynamic and harsh environment of the horse hoof is taken into account throughout every step of the procedure because it is impossible to de-embed the antenna from its environment. The antenna and channel model are verified extensively by measurements in phantom and ex vivo. The antenna is verified to be robust against changes in the morphology of the horse's hoof up to 50%. The dynamic environment was captured by considering different soil types and air, and the design was verified to be resilient against changes herein. The antenna performs well within the targeted band, with a fractional bandwidth of 8% and a gain of -2 dBi. Furthermore, a path loss model was constructed for a typical barn environment, and the antenna reaches a range of 250 m in the studied environment based on the LoRa technology. This research is important for monitoring horse health.

Vehicle Localization Using Doppler Shift and Time of Arrival Measurements in a Tunnel Environment.

Most applications and services of Cooperative Intelligent Transport Systems (C-ITS) rely on accurate and continuous vehicle location information. The traditional localization method based on the Global Navigation Satellite System (GNSS) is the most commonly used. However, it does not provide reliable, continuous, and accurate positioning in all scenarios, such as tunnels. Therefore, in this work, we present an algorithm that exploits the existing Vehicle-to-Infrastructure (V2I) communication channel that operates within the LTE-V frequency band to acquire in-tunnel vehicle location information. We propose a novel solution for vehicle localization based on Doppler shift and Time of Arrival measurements. Measurements performed in the Beveren tunnel in Antwerp, Belgium, are used to obtain results. A comparison between estimated positions using Extended Kalman Filter (EKF) on Doppler shift measurements and individual Kalman Filter (KF) on Doppler shift and Time of Arrival measurements is carried out to analyze the filtering methods performance. Findings show that the EKF performs better than KF, reducing the average estimation error by 10 m, while the algorithm accuracy depends on the relevant RF channel propagation conditions and other in-tunnel-related environment knowledge included in the estimation. The proposed solution can be used for monitoring the position and speed of vehicles driving in tunnel environments.

Lessons Learned from a Distributed RF-EMF Sensor Network.

In an increasingly wireless world, spatiotemporal monitoring of the exposure to environmental radiofrequency (RF) electromagnetic fields (EMF) is crucial to appease public uncertainty and anxiety about RF-EMF. However, although the advent of smart city infrastructures allows for dense networks of distributed sensors, the costs of accurate RF sensors remain high, and dedicated RF monitoring networks remain rare. This paper describes a comprehensive study comprising the design of a low-cost RF-EMF sensor node capable of monitoring four frequency bands used by wireless telecommunications with an unparalleled temporal resolution, its application in a small-scale distributed sensor network consisting of both fixed (on building façades) and mobile sensor nodes (on postal vans), and the subsequent analysis of over a year of data between January 2019 and May 2020, during which slightly less than 10 million samples were collected. From the fixed nodes' results, the potential errors were determined that are induced when sampling at lower speeds (e.g., one sample per 15 min) and measuring for shorter periods of time (e.g., a few weeks), as well as an adequate resolution (30 min) for diurnal and weekly temporal profiles which sufficiently preserves short-term variations. Furthermore, based on the correlation between the sensors, an adequate density of 100 sensor nodes per km2 was deduced for future networks. Finally, the mobile sensor nodes were used to identify potential RF-EMF exposure hotspots in a previously unattainable area of more than 60 km2. In summary, through the analysis of a small number of RF-EMF sensor nodes (both fixed and mobile) in an urban area, this study offers invaluable insights applicable to future designs and deployments of distributed RF-EMF sensor networks.

Radiofrequency electromagnetic fields from mobile communication: Description of modeled dose in brain regions and the body in European children and adolescents.

BACKGROUND: Little is known about radiofrequency electromagnetic fields (RF) from mobile technology and resulting dose in young people. We describe modeled integrated RF dose in European children and adolescents combining own mobile device use and surrounding sources. METHODS: Using an integrated RF model, we estimated the daily RF dose in the brain (whole-brain, cerebellum, frontal lobe, midbrain, occipital lobe, parietal lobe, temporal lobes) and the whole-body in 8358 children (ages 8-12) and adolescents (ages 14-18) from the Netherlands, Spain, and Switzerland during 2012-2016. The integrated model estimated RF dose from near-field sources (digital enhanced communication technology (DECT) phone, mobile phone, tablet, and laptop) and far-field sources (mobile phone base stations via 3D-radiowave modeling or RF measurements). RESULTS: Adolescents were more frequent mobile phone users and experienced higher modeled RF doses in the whole-brain (median 330.4 mJ/kg/day) compared to children (median 81.8 mJ/kg/day). Children spent more time using tablets or laptops compared to adolescents, resulting in higher RF doses in the whole-body (median whole-body dose of 81.8 mJ/kg/day) compared to adolescents (41.9 mJ/kg/day). Among brain regions, temporal lobes received the highest RF dose (medians of 274.9 and 1786.5 mJ/kg/day in children and adolescents, respectively) followed by the frontal lobe. In most children and adolescents, calling on 2G networks was the main contributor to RF dose in the whole-brain (medians of 31.1 and 273.7 mJ/kg/day, respectively). CONCLUSION: This first large study of RF dose to the brain and body of children and adolescents shows that mobile phone calls on 2G networks are the main determinants of brain dose, especially in temporal and frontal lobes, whereas whole-body doses were mostly determined by tablet and laptop use. The modeling of RF doses provides valuable input to epidemiological research and to potential risk management regarding RF exposure in young people.

Protocol for personal RF-EMF exposure measurement studies in 5th generation telecommunication networks.

BACKGROUND: The general population is exposed to Radio-Frequency Electromagnetic Fields (RF-EMFs) used by telecommunication networks. Previous studies developed methods to assess this exposure. These methods will be inadequate to accurately assess exposure in 5G technologies or other wireless technologies using adaptive antennas. This is due to the fact that 5G NR (new radio) base stations will focus actively on connected users, resulting in a high spatio-temporal variations in the RF-EMFs. This increases the measurement uncertainty in personal measurements of RF-EMF exposure. Furthermore, a user's exposure from base stations will be dependent on the amount of data usage, adding a new component to the auto-induced exposure, which is often omitted in current studies. GOALS: The objective of this paper is to develop a general study protocol for future personal RF-EMF exposure research adapted to 5G technologies. This protocol will include the assessment of auto-induced exposure of both a user's own devices and the networks' base stations. METHOD: This study draws from lessons learned from previous RF-EMF exposure research and current knowledge on 5G technologies, including studies simulating 5G NR base stations and measurements around 5G NR test sites. RESULTS: To account for auto-induced exposure, an activity-based approach is introduced. In survey studies, an RF-EMF sensor is fixed on the participants' mobile device(s). Based on the measured power density, GPS data and movement and proximity sensors, different activities can be clustered and the exposure during each activity is evaluated. In microenvironmental measurements, a trained researcher performs measurements in predefined microenvironments with a mobile device equipped with the RF-EMF sensor. The mobile device is programmed to repeat a sequence of data transmission scenarios (different amounts of uplink and downlink data transmissions). Based on simulations, the amount of exposure induced in the body when the user device is at a certain location relative to the body, can be evaluated. CONCLUSION: Our protocol addresses the main challenges to personal exposure measurement introduced by 5G NR. A systematic method to evaluate a user's auto-induced exposure is introduced.

Use of Machine Learning for the Estimation of Down- and Up-Link Field Exposure in Multi-Source Indoor WiFi Scenarios.

A novel Machine Learning (ML) method based on Neural Networks (NN) is proposed to assess radio-frequency (RF) exposure generated by WiFi sources in indoor scenarios. The aim was to build an NN capable of addressing the complexity and variability of real-life exposure setups, including the effects of not only down-link transmission access points (APs) but also up-link transmission by different sources (e.g. laptop, printers, tablets, and smartphones). The NN was fed with easy to be found data, such as the position and type of WiFi sources (APs, clients, and other users) and the position and material characteristics (e.g. penetration loss) of walls. The NN model was assessed using an additional new layout, distinct from that one used to build and optimize the NN coefficients. The NN model achieved a remarkable field prediction accuracy across exposure conditions in both layouts, with a median prediction error of -0.4 to 0.6 dB and a root mean square error of 2.5-5.1 dB, compared with the target electric field estimated by a deterministic indoor network planner. The proposed approach performs well for the different layouts and is thus generally used to assess RF exposure in indoor scenarios. © 2021 The Authors. Bioelectromagnetics published by Wiley Periodicals LLC on behalf of Bioelectromagnetics Society.

Multi-Floor Indoor Pedestrian Dead Reckoning with a Backtracking Particle Filter and Viterbi-Based Floor Number Detection.

We present a smartphone-based indoor localisation system, able to track pedestrians over multiple floors. The system uses Pedestrian Dead Reckoning (PDR), which exploits data from the smartphone's inertial measurement unit to estimate the trajectory. The PDR output is matched to a scaled floor plan and fused with model-based WiFi received signal strength fingerprinting by a Backtracking Particle Filter (BPF). We proposed a new Viterbi-based floor detection algorithm, which fuses data from the smartphone's accelerometer, barometer and WiFi RSS measurements to detect stairs and elevator usage and to estimate the correct floor number. We also proposed a clustering algorithm on top of the BPF to solve multimodality, a known problem with particle filters. The proposed system relies on only a few pre-existing access points, whereas most systems assume or require the presence of a dedicated localisation infrastructure. In most public buildings and offices, access points are often available at smaller densities than used for localisation. Our system was extensively tested in a real office environment with seven 41 m × 27 m floors, each of which had two WiFi access points. Our system was evaluated in real-time and batch mode, since the system was able to correct past states. The clustering algorithm reduced the median position error by 17% in real-time and 13% in batch mode, while the floor detection algorithm achieved a 99.1% and 99.7% floor number accuracy in real-time and batch mode, respectively.