Maria Garcia Puyol

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.

Pedestrian Simultaneous Localization and Mapping in Multistory Buildings Using Inertial Sensors

Pedestrian navigation is an important ingredient for efficient multimodal transportation, such as guidance within large transportation infrastructures. A requirement is accurate positioning of people in indoor multistory environments. To achieve this, maps of the environment play a very important role. FootSLAM is an algorithm based on the simultaneous localization and mapping (SLAM) principle that relies on human odometry, i.e., measurements of a pedestrians steps, to build probabilistic maps of human motion for such environments and can be applied using crowdsourcing. In this paper, we extend FootSLAM to multistory buildings following a Bayesian derivation. Our approach employs a particle filter and partitions the map space into a grid of adjacent hexagonal prisms with eight faces. We model the vertical component of the odometry errors using an autoregressive integrated moving average (ARIMA) model and extend the geographic tree-based data structure that efficiently stores the probabilistic map, allowing real-time processing. We present the multistory FootSLAM maps that were created from three data sets collected in different buildings (one large office building and two university buildings). Hereby, the user was only carrying a single foot-mounted inertial measurement unit (IMU). We believe the resulting maps to be strong evidence of the robustness of FootSLAM. This paper raises the future possibility of crowdsourced indoor mapping and accurate navigation using other forms of human odometry, e.g., obtained with the low-cost and nonintrusive sensors of a handheld smartphone.

Collaborative Mapping for Pedestrian Navigation in Security Applications

In rescue missions or law enforcement applications, accurate determination of every team member’s position and providing this information on a map may significantly improve mutual situation awareness and potentially reduce the risk of accidentally harming a team member. Furthermore, it could help keep track of the areas that have been already visited, helping the coordination of the mission at hand.

Managing large-scale mapping and localization for pedestrians using inertial sensors

Pedestrian navigation in indoor environments without a pre-installed infrastructure still presents many challenges. There are different approaches that address the problem using prior knowledge about the environment when the building plans or similar are available. Since this is not always the case, a family of technologies based on the principle of Simultaneous Localization and Mapping (SLAM) has been proposed. In this paper we will present some estimates on how a mapping process based on FootSLAM - a form of SLAM for pedestrians - might scale for a large-scale collaborative effort eventually encompassing most of our public indoor space, where the mapping entities are humans. Our assumptions on pedestrian motion and area visiting rate together with calculations based on the computational requirements of pedestrian SLAM algorithms allow us to make estimates with regard to the feasibility, scalability and computational cost of wide-scale mapping of indoor areas by pedestrians.

Complexity-reduced FootSLAM for indoor pedestrian navigation using a geographic tree-based data structure

FootSLAM or simultaneous localisation and mapping (SLAM) for pedestrians is a technique that addresses the indoor positioning and mapping problem based on human odometry (aka pedestrian dead reckoning), for example with a foot-mounted inertial sensor. FootSLAM follows the FastSLAM factorisation, using a Rao-Blackwellised particle filter to simultaneously estimate the building layout and the pedestrians pose – his position and orientation. To that end, FootSLAM divides the 2D space into a grid of uniform and adjacent hexagons and counts the number of times that each particle crosses the edges of the hexagons it visits. As we shall show, the complexity of FootSLAM grows quadratically with time, preventing the mapping of large areas. In this paper, we present a new geographic tree-based data structure, called H-tree, to reduce the quadratic-in-time computational growth rate of naïve FootSLAM to t times log t. In addition, we introduce a compact representation (alphabet) for the set of six counters that are used to map the transitions of the particles across the edges of each hexagon. This alphabet is particularly effective during the exploration phases of FootSLAM that requires much particle diversity. In this contribution, the computational savings of the H-tree are presented both theoretically and with real-world data. In practice, we believe that FootSLAM can be applied in quasi real-time applications that require rapid mapping of unknown areas. Additionally, the mass market offline mapping process can be undertaken much more efficiently.

Complexity-reduced FootSLAM for indoor pedestrian navigation

FootSLAM or simultaneous localization and mapping (SLAM) for pedestrians is a technique that addresses the indoor positioning and mapping problem based on human odometry (aka pedestrian dead reckoning), e.g. with a foot-mounted inertial sensor. FootSLAM follows the FastSLAM factorization, using a Rao-Blackwellized particle filter to simultaneously estimate the building layout and the pedestrians pose - his position and orientation. To that end, FootSLAM divides the 2D space into a grid of uniform and adjacent hexagons and counts the number of times each particle crosses the edges of the hexagons it visits. As we shall show, the complexity of FootSLAM grows quadratically with time, preventing the mapping of large areas.

Maps-based angular PDFs used as prior maps for FootSLAM

FootSLAM (Simultaneous Localization and Mapping) is a new technology that addresses the indoor mapping and positioning challenge that relies only on sensors that the person carries. In this paper, we propose to use maps-based angular Probability Density Functions (PDFs), using prior knowledge of the building layout as prior maps for FootSLAM and show how they can be integrated into the FootSLAM weight update. The angular PDFs - that represent PDFs for a human step direction at a given location - are derived from a diffusion algorithm based on maps. The advantage of using prior maps is that the FootSLAM algorithm reaches convergence faster and more accurately.

Characterization of planar-intensity based heading likelihood functions in magnetically disturbed indoor environments

Heading information is a critical input to pedestrian dead reckoning. Unlike in most outdoor environments, the magnetic field inside of buildings is often strongly perturbed and inhomogeneous. Hence, straightforward approaches to use measurements of two-axis and three-axis magnetometers perform poorly. In recent measurements we have observed statistical properties of the magnetic field indicating that knowledge of the measured horizontal magnetic intensity is informative about the expected deviation of the measured magnetic heading. This statistical dependence has been quantified based on indoor measurements that have been collected in several offices and corridors in 3 buildings having different building orientations. A decrease in the spread of the horizonal angle is exhibited for larger horizontal intensities suggesting that measurements with large horizontal intensities are more reliable. We provide an approach to determine a likelihood function for the measured magnetic heading as a function of the local magnetic intensity in indoor environments. We show how the Expectation Maximization (EM) algorithm is used to construct a parametric two dimensional distribution of heading and planar-intensity, which can serve as a heading likelihood function in Bayesian positioning estimators, based upon which, greater weight is given to the less disturbed (strong-intensity) heading measurements and lower weight to the more erroneous ones (low-intensity). Drawing on our empirical data we show the performance improvement achieved with this new likelihood function.

Measuring the Uncertainty of Probabilistic Maps Representing Human Motion for Indoor Navigation

Indoor navigation and mapping have recently become an important field of interest for researchers because global navigation satellite systems (GNSS) are very often unavailable inside buildings. FootSLAM, a SLAM (Simultaneous Localization and Mapping) algorithm for pedestrians based on step measurements, addresses the indoor mapping and positioning problem and can provide accurate positioning in many structured indoor environments. In this paper, we investigate how to compare FootSLAM maps via two entropy metrics. Since collaborative FootSLAM requires the alignment and combination of several individual FootSLAM maps, we also investigate measures that help to align maps that partially overlap. We distinguish between the map entropy conditioned on the sequence of pedestrian’s poses, which is a measure of the uncertainty of the estimated map, and the entropy rate of the pedestrian’s steps conditioned on the history of poses and conditioned on the estimated map. Because FootSLAM maps are built on a hexagon grid, the entropy and relative entropy metrics are derived for the special case of hexagonal transition maps. The entropy gives us a new insight on the performance of FootSLAM’s map estimation process.