Mohammed Khider

Unscented Kalman filter and Magnetic Angular Rate Update (MARU) for an improved Pedestrian Dead-Reckoning

The Extended Kalman Filter (EKF) has been the state of the art in Pedestrian Dead-Reckoning for foot-mounted Inertial Measurements Units. However due to the non-linearity in the propagation of the orientation the EKF is not the optimal Bayesian filter. We propose the usage of the Unscented Kalman Filter (UKF) as the integration algorithm for the inertial measurements. The UKF improves the mean and covariance propagation needed for the Kalman filter. Although the UKF provides a better estimate of the orientation, with Zero velocity UPdaTes (ZUPT) measurements, the yaw and the bias in the gyroscope associated with it becomes unobserved and might generate errors in the positioning. We studied the changes in the magnetic field during the stance phase and their relationship with the turn rates to propose three measurements using the magnetometer signal that will be called Magnetic Angular Rate Updates (MARUs). The first measurement uses the change in the angle of the magnetic field in the horizontal plane to measure the change in the yaw and provides a simple measurement for the UKF implementation. The second measurement relates the change in the magnetic field vector to the turn rate and provides information on the bias of the gyroscope for an UKF. The last measurement uses a first order approximation to generate a linear relationship with the gyroscope bias and therefore it can be used in an EKF. Finally we proposed a metric for the reliability of the stance as a way to use the pre and post stance information but adjusting the covariance of the measurements gradually from swing to stance. These methods were tested on real and simulated signals and they have shown improvements over the original PDR algorithms.

Continuous location and direction estimation with multiple sensors using particle filtering

In this paper we discuss the use of particle filtering to estimate the values of several state variables describing a users context. Since particle filtering algorithms are computationally efficient realizations of Bayesian filters they perform exceptionally well to optimally combine the a priori knowledge stemming from behavioral models, such as movement models, and the noisy measurements from sensors. The estimate at each time step is obtained in the form of a probability density function that represents the entire information and quantifies the inherent uncertainty about the context. The concept has been realized in simulations and experiments. In this paper, the applied movement model is presented with simulated measurements from GPS and compass sensors to illustrate the concept

Simultaneous Localization and Mapping for pedestrians using distortions of the local magnetic field intensity in large indoor environments

We present a Simultaneous Localization and Mapping (SLAM) algorithm based on measurements of the ambient magnetic field strength (MagSLAM) that allows quasi-real-time mapping and localization in buildings, where pedestrians with foot-mounted sensors are the subjects to be localized. We assume two components to be present: firstly a source of odometry (human step measurements), and secondly a sensor of the local magnetic field intensity. Our implementation follows the FastSLAM factorization using a particle filter. We augment the hexagonal transition map used in the pre-existing FootSLAM algorithm with local maps of the magnetic field strength, binned in a hierarchical hexagonal structure. We performed extensive experiments in a number of different buildings and present the results for five data sets for which we have ground truth location information. We consider the results obtained using MagSLAM to be strong evidence that scalable and accurate localization is possible without an a priori map.

A high precision reference data set for pedestrian navigation using foot-mounted inertial sensors

In this paper we present a reference data set that we are making publicly available to the indoor navigation community [8]. This reference data is intended for the analysis and verification of algorithms based on foot mounted inertial sensors. Furthermore, we describe our data collection methodology that is applicable to the analysis of a broad range of indoor navigation approaches. We employ a high precision optical reference system that is traditionally being used in the film industry for human motion capturing and in applications such as analysis of human motion in sports and medical rehabilitation. The data set provides measurements from a six degrees of freedom foot mounted inertial MEMS sensor array, as well as synchronous high resolution data from the optical tracking system providing ground truth for location and orientation. We show the use of this reference data set by comparing the performance of algorithms for an essential part of pedestrian dead reckoning systems for positioning, namely identification of the rest phase during the human gait cycle.

Improving Simultaneous Localization and Mapping for pedestrian navigation and automatic mapping of buildings by using online human-based feature labeling

In this paper we present an extension to odometry based SLAM for pedestrians that incorporates human-reported measurements of recognizable features, or “places” in an environment. The method which we have called “PlaceSLAM” builds on the Simultaneous Localization and Mapping (SLAM) principle in that a spatial representation of such places can be built up during the localization process. We see an important application to be in mapping of new areas by volunteering pedestrians themselves, in particular to improve the accuracy of “FootSLAM” which is based on human step estimation (odometry). We present a description of various flavors of PlaceSLAM and derive a Bayesian formulation and particle filtering implementation for the most general variant. In particular we distinguish between two important cases which depend on whether the pedestrian is required to report a places identifier or not. Our results based on experimental data show that our approach can significantly improve the accuracy and stability of FootSLAM and this with very little additional complexity. After mapping has been performed, users of such improved FootSLAM maps need not report places themselves.

Particle filter based positioning with 3GPP-LTE in indoor environments

Global navigation satellite systems (GNSSs) can deliver very good position estimates under optimum conditions. However, especially in urban and indoor scenarios with severe multipath propagation and blocking of satellites by buildings the accuracy loss can be very large. Often, a position with GNSS is impossible in these scenarios. On the other hand, cellular wireless communication systems such as the third generation partnership project (3GPP) long-term evolution (LTE) provide excellent coverage in urban and most indoor environments. Thus, this paper researches timing based positioning algorithms, in this case time difference of arrival (TDoA), using 3GPP-LTE measurements. Several approaches and algorithms exist to solve the navigation equation for cellular systems, for instance Bayes filtering methods such as Kalman or particle filter. This paper specifically considers and develops a particle filter for 3GPP-LTE TDoA positioning. To obtain better positioning results, a 3GPP-LTE TDoA error model is derived. This error model is afterwards included in the likelihood function of the particle filter. The last part of this paper, evaluates the positioning performances of the developed particle filter in an indoor scenario. These evaluations show clearly the possibility of using 3GPP-LTE measurements for indoor positioning.

Towards operational systems for continuous navigation of rescue teams

Rescue workers, such as fire fighters often face conditions that impose a combination of threats on them: severe physical and mental stress, limited or no visibility, unknown layout of buildings may induce disorientation. Systems capable of continuously locating all members of a rescue team could help to maintain orientation, and in cases of accident, help other team members or team leaders to localize colleagues in danger. The paper points out the discrepancy between state of the art in research projects, primarily in the fields of wearable computing and indoor navigation, and the lack of operational systems that have been fielded or have undergone sustained usage. The paper shows, that a successful operational system for continuous navigation of rescue teams requires a combination of several sensors, motion models, computing, communication means, and human machine interfaces. Based on an analysis of rescue operations and interviews with rescue workers, relevant issues and requirements are derived. A reference architecture for an operational system for continuous navigation of rescue teams is outlined. Its implementation and feasibility aspects are discussed. We discuss in how far the required technological components are available and affordable and where shortcomings have to be resolved by future research and development.

Optimal sampling frequency and bias error modeling for foot-mounted IMUs

The use of foot-mounted inertial measurement units (IMUs) has shown promising results in providing accurate human odometry as a component of accurate indoor pedestrian navigation. The specifications of these sensors, such as the sampling frequency have to meet requirements related to human motion. We investigate the lowest usable sampling frequency: To do so, we evaluate the frequency distribution of different human motion like crawling, jumping or walking in different scenarios such as escalators, lifts, on carpet or grass, and with different footwear. These measurements indicate that certain movement patterns, as for instance going downstairs, upstairs, running or jumping contain more high frequency components. When using only a low sampling rate this high frequency information is lost. Hence, it is important to identify the lowest usable sampling frequency and sample with it if possible. We have made a set of walks to illustrate the resulting odometries at different frequencies, after applying an Unscented Kalman Filter (UKF) using Zero Velocity Updates. The odometry error is highly dependent on the drift of the individual accelerometers and gyroscopes. In order to obtain better odometry it is necessary to perform a detailed analysis of the sensor noise processes. We resorted to computing the Allan variance for three different IMU chipsets of various quality specification. From this we have derived a bias model for the UKF and evaluated the benefit of applying this model to a set of real data from walk.

A pedestrian navigation system using a map-based angular motion model for indoor and outdoor environments

By incorporating known floor-plans in sequential Bayesian positioning estimators, such as particle filters, long-term positioning accuracy can be achieved as long as the map is sufficiently accurate and the environment sufficiently constrains pedestrians’ motion. Instead of using binary decisions to eliminate particles when crossing a wall, map-based angular probability density functions (PDFs) are used in this article that are capable of weighting the possible headings of the pedestrian according to local infrastructure. In addition, we will include outdoor maps by processing satellite images of the region. We will show that the angular PDFs will help to obtain better performance in critical multi-modal navigation scenarios and in the outdoor area when including maps.

On-line training of the path-loss model in Bayesian WLAN indoor positioning

Received signal strengths have been widely exploited in indoor positioning due to the massive presence of wireless local networks in buildings. Theoretical propagation models such as the path-loss model can be used in order to avoid long training phases as in fingerprinting approaches. The main issue concerning the employiment of the path-loss model is that the values of some parameters, i.e., the transmit power and the decay exponent, depend on many factors, such as the device, building structure and other environmental features. In this paper, we propose a Bayesian positioning algorithm based on the Rao-Blackwellized particle filter, where the parameters of the path-loss model are estimated independently for each AP in addition to localizing the user. Both parameters are described by discrete random variables with uniform priors. We validate ou proposal by means of simulations and two different experiments; finally, some remarks on complexity are also given.