Marcel Kyas

A reference system for indoor localization testbeds

We present a low-cost robot system capable of performing robust indoor localization while carrying components of another system which shall be evaluated. Using off-the-shelf components, the ground truth positioning data provided by the robot can be used to evaluate a variety of localization systems and algorithms. Not needing any pre-installed components in its environment, it is very easy to setup. The robot system relies on wheel-odometry data of a Roomba robot, and visual distance measurements of two Kinects. The Robot Operating System (ROS) is used for the localization process according to a precise pre-drawn floor plan that may be enhanced with Simultaneous Localization and Mapping (SLAM). The system is able to estimate its position with an average error of 6.7 cm. It records its own positioning data as well as the data from the system under evaluation and provides simple means for analysis. It is also able to re-drive a previous test run if reproducable conditions are needed.

Dynamic Classes: Modular Asynchronous Evolution of Distributed Concurrent Objects

Many long-lived and distributed systems must remain available yet evolve over time, due to, e.g., bugfixes, feature extensions, or changing user requirements. To facilitate such changes, formal methods can help in modeling and analyzing runtime software evolution. This paper presents an executable object-oriented modeling language which supports runtime software evolution. The language, based on Creol, targets distributed systems by active objects, asynchronous method calls, and futures. A dynamic class construct is proposed in this setting, providing an asynchronous and modular upgrade mechanism. At runtime, class redefinitions gradually upgrade existing instances of a class and of its subclasses. An upgrade may depend on previous upgrades of other classes. For asynchronous runtime upgrades, the static picture may differ from the actual runtime system. An operational semantics and a type and effect system are given for the language. The type analysis of an upgrade infers and collects dependencies on previous upgrades. These dependencies are exploited as runtime constraints to ensure type safety.

Comparing centralized Kalman filter schemes for indoor positioning in wireless sensor network

Sensor devices suffer severe interference due to multi path effect, non line of sight (NLOS) and variations of the wireless propagation environment in indoor positioning systems. These interferences lead to measurement error during sensor measurement. Kalman filter (KF) and extended Kalman filter (EKF) have been widely used in tracking systems to reduce measurement noise. However, KF and EKF assume the measurement noise follows normal distribution, and the real noise distribution should be based on experimental statistical results. Besides fluctuating wireless condition makes the system unstable in indoor environments. We analyze the time of flight (TOF) measurement statistic model in experiments and design KF and EKF models for indoor positioning system according to the statistic model. We introduce our system architecture for wireless sensor networks (WSN) to overcome KFs drawbacks, which divides the positioning system into three components: measurement, pre-processing and data-processing. Measurement component measures the range based on TOF method. We developed a voting filter (VF) and an averaging filter (AF) in preprocessing to reduce measurement noise for later processing. During data-processing, Kalman filter and extended Kalman filter are used to track the positions. We also implement another scheme, low pass filter (LPF) with KF or EKF, to estimate the positions with the knowledge of geographic information. Three realistic experiments are set up using the sensor equipment nanoPAN 5375 to evaluate these methods. Comparing the experimental results, low pass filter with EKF is most suitable for indoor positioning.

Quantitative and Spatial Evaluation of Distance-Based Localization Algorithms

Indoor localization, especially in wireless networks (WN) has become an important research focus in computer science during the past ten years. Several approaches exist to estimate a node’s position relative to other devices. Most approaches are based on distance measurements and localization algorithms. In this chapter we provide an overview of common and new localization algorithms. A detailed investigation on the error distribution and the real world behaviour of these algorithms is presented. We also provide a discussion of the evaluation results that leads to open questions and future research approaches.

Executable interface specifications for testing asynchronous creol components

We propose and explore a formal approach for black-box testing asynchronously communicating components in open environments. Asynchronicity poses a challenge for validating and testing components. We use Creol, a high-level, object-oriented language for distributed systems and present an interface specification language to specify components in terms of traces of observable behavior. n nThe language enables a concise description of a component’s behavior, it is executable in rewriting logic and we use it to give test specifications for Creol components. In a specification, a clean separation between interaction under control of the component or coming from the environment is central, which leads to an assumption-commitment style description of a component’s behavior. The assumptions schedule the inputs, whereas the outputs as commitments are tested for conformance with the specification. The asynchronous nature of communication in Creol is respected by testing only up-to a notion of observability. The existing Creol interpreter is combined with our implementation of the specification language to obtain a specification-driven interpreter for testing.

The Membership Degree Min-Max localization algorithm

We introduce the Membership Degree Min-Max (MD-Min-Max) localization algorithm as a precise and simple lateration algorithm for indoor localization. MD-Min-Max is based on the well known Min-Max algorithm that uses a bounding box to compute the position. We present an analysis of the Min-Max algorithm and show strengths and weaknesses in the spatial distribution of the position error. MD-Min-Max uses a Membership Function (MF) based on an estimated error distribution of the distance measurements to gain a higher precision than Min-Max. The algorithm has the same complexity as Min-Max and can be used for indoor localization even on small devices, e.g. in Wireless Sensor Networks (WSNs). To evaluate the performance of the algorithm we compare it with other Min-Max algorithms in simulations and in a large real world deployment of a WSN.

A non-parametric modeling of Time-of-flight ranging error for indoor network localization

For indoor network positioning, the ranging error modeling plays an important role in positioning algorithm optimization, simulation setup, system parameters calibration, performance evaluation and test-bed configuration, etc. However, the error model is commonly assumed as a Normal distribution or other distributions which cannot capture the negative value, the positive bias and the right-side tail phenomenon of the indoor ranging error. We use a non-parametric modeling for the ranging error, as the indoor wireless propagation is non-analytical. Seven error models are configured by distribution fitting with measurements from both stationary and mobile Time-of-flight (TOA) experiments implemented in typical indoor scenarios. Then the configured models are evaluated by Kolmogorov-Smirnov (KS) goodness-of-fit hypothesis test, indicating that a biased statistical model is sufficient to characterize indoor ranging and has a good fitting to real-world measurements. Further, the KS test results are verified by comparing the simulated and experimental positioning results. Our modeling method works for most indoor scenarios, and the modeling results can significantly improve the effective of simulations to reality.

A statistics-based least squares (SLS) method for non-line-of-sight error of indoor localization

The main challenge of indoor wireless positioning is the large positive bias of non-line-of-sight (NLOS) ranging, which directly degrades the localization accuracy. It is necessary for a practical positioning algorithm to learn the specific of indoor ranging from statistics. Thus, we analyze the ranging characteristics based on real-world indoor experiments with both static and mobile cases, which finds that bounding-box algorithm is robust to the NLOS bias. Then, we propose a twostep statistics-based least squares (SLS) method consisting of a NLOS bias elimination and a linear least squares (LLS) process. SLS first removes the NLOS bias by an intermediate estimation obtained from bounding-box algorithm, then uses a weighted LLS estimator to handle the remained ranging error. The difference between SLS and other NLOS mitigation approaches is that SLS aims to remove the bias away from the NLOS range while the others try to less emphasize or discard the NLOS range. SLS is compared with three NLOS mitigation algorithms on a sensor test-bed in a typical hallway. Results demonstrate that an effective NLOS mitigation of SLS.

The effects of human body shadowing in RF-based indoor localization

In radio frequency based indoor human localization systems with body mounted sensors, the human body can cause non-line-of-sight (NLOS) effects which might result in severe range estimation and localization errors. However, previous studies on the impact of the human body only conducted static experiments in controlled environments. We confirm known effects and conduct real-world experiments in a typical indoor human tracking scenario using 2.4 GHz time of flight (TOF) range measurements. We analyze the effect on the raw measurements and on the localization results using the localization algorithms Centroid, NLLS, MD-Min-Max, and Geo-n. The experiment design is focused on incident management, where an infrastructure might only be installed in front of the building. We show that these effects have considerable impact on the localization accuracy of the person.

Weighted Least-Squares by Bounding-Box (B-WLS) for NLOS Mitigation of Indoor Localization

The major problem of indoor localization is the imprecise ranging, which directly degrades the localization accuracy. The ordinary least-squares (OLS) estimator is able to handle unbiased and homoscedastic ranging errors, but incapable for the bias and heteroscedasticity as characterized by real-world ranging of indoor scenarios, especially the non-line-of-sight (NLOS) error. A potential improvement of LS is to weight each element related to the corresponding ranging error, known as the weighted LS (WLS). However, current weighting metrics are either impractical to get or still involve the NLOS error. We propose to weight each element in a linear LS (LLS) estimator by the difference between the measured ranges and Bounding-box results, named B-WLS. Compared with five LS type algorithms in simulations and a mobile target experiments, results demonstrate that B-WLS efficiently enables the LLS estimation to suppress to the NLOS error.