Michailas Romanovas

A modular and mobile system for indoor localization

The work presents a system for sensor data and complementary information fusion for localization in indoor environments. The system is based on modular sensor units, which can be attached to a person and contains various sensors, such as range sensors, inertial and magnetic sensors, a GPS receiver and a barometer. The measurements are processed using Bayesian Recursive Estimation algorithms and combined with available a priori knowledge such as map information or human motion models and constraints. The processing can be done locally, since all necessary data are available on the mobile unit. This system provides a platform for implementation, combination and evaluation of various localization principles and can be used for a variety of applications, such as indoor and outdoor pedestrian navigation, localization of other objects such as vehicles as well as robotics applications.

Multi-modal sensor data and information fusion for localization in indoor environments

The work presents the development of a framework for sensor data and complementary information fusion for localization in indoor environments. The framework is based on a modular and flexible sensor unit, which can be attached to a person and which contains various sensor types, such as range sensors, inertial and magnetic sensors or barometers. All measurements are processed within Bayesian Recursive Estimation algorithms and combined with available a priori knowledge such as map information or human motion models.

A study on indoor pedestrian localization algorithms with foot-mounted sensors

The work presents a foot-mounted sensor system for a combined indoor/outdoor pedestrian localization. The approach is based on a zero-velocity update scheme formulated as an Extended or Unscented Kalman filter with quaternion orientation representation and employs a custom low-cost sensor unit. Both filters are compared in terms of speed and accuracy on a representative trajectory. A detailed discussion is provided with respect to different filter state formulations, stance still detection mechanisms and associated filter parameters. The presented pure inertial system is augmented with magnetic field measurements for heading correction. The challenging localization scenario with an elevator is addressed by augmenting the system with a barometric pressure sensor for height error correction. The work also demonstrates how the basic algorithm version can be extended with reference systems such as GPS and passive RFID tags on the floor for absolute position drift correction.

Using multi-dimensional dynamic time warping for TUG test instrumentation with inertial sensors

The Timed Up and Go (TUG) is a clinical test used widely to measure balance and mobility, e.g. in Parkinsons disease (PD). The test includes a sequence of functional activities, namely: sit-to-stand, 3-meters walk, 180° turn, walk back, another turn and sit on the chair. Meanwhile the stopwatch is used to score the test by measuring the time which the patients with PD need to perform the test. Here, the work presents an instrumented TUG using a wearable inertial sensor unit attached on the lower back of the person. The approach is used to automate the process of assessment compared with the manual evaluation by using visual observation and a stopwatch. The developed algorithm is based on the Dynamic Time Warping (DTW) for multi-dimensional time series and has been applied with the augmented feature for detection and duration assessment of turn state transitions, while a 1-dimensional DTW is used to detect the sit-to-stand and stand-to-sit phases. The feature set is a 3-dimensional vector which consists of the angular velocity, derived angle and features from Linear Discriminant Analysis (LDA). The algorithm was tested on 10 healthy individuals and 20 patients with PD (10 patients with early and late disease phases respectively). The test demonstrates that the developed technique can successfully extract the time information of the sit-to-stand, both turns and stand-to-sit transitions in the TUG test.

TUG Test Instrumentation for Parkinson's disease patients using Inertial Sensors and Dynamic Time Warping.

The Timed Up and Go (TUG) test is a clinical tool widely used to evaluate balance and mobility, e.g. in Parkinson’s disease (PD). This test includes a sequence of functional activities, namely: sit-to-stand, 3-meters walk, 180° turning, walk back, and turn-to-sit. The work introduces a new method to instrument the TUG test using a wearable inertial sen-sor unit (DynaPort Hybrid, McRoberts B.V., NL) attached on the lower back of the person. It builds on Dynamic Time Warping (DTW) for detection and duration assessment of associated state transitions. An automatic assessment to sub-stitute a manual evaluation with visual observation and a stopwatch is aimed at to gain objective information about the patients. The algorithm was tested on data of 10 healthy individuals and 20 patients with Parkinsons disease (10 pa-tients for early and late disease phases respectively). The algorithm successfully extracted the time information of the sit-to-stand, turn and turn-to-sit transitions.

Efficient orientation estimation algorithm for low cost inertial and magnetic sensor systems

The presented work develops a high-dynamic quaternion attitude estimation Unscented Kalman Filter suitable for the implementation in low cost sensor systems, comprising accelerometers, gyroscopes and magnetic field sensors. The adoption of a spherical σ-point selection strategy and the implementation of the square-root version reduces the requirements for computational resources. A special handling of angular rate, gyroscope bias, and translational accelerations within the process model is implemented and evaluated. Issues regarding the quaternion mean calculations, noise representation as well as control noise scaling are discussed. The performance of the designed filter is assessed using real hand motion reference data and correspondingly generated noisy sensor measurements.

Observing Relative Motion With Three Accelerometer Triads

A gyroscope-free inertial measurement unit (GF-IMU) detects the relative motion of a body based on acceleration measurements only. It consists of multiple transducers attached at distinct locations within the body that together form an accelerometer array. In this paper, we employ only three accelerometer triads in order to completely capture the transversal and angular acceleration as well as the angular velocity. By modeling the GF-IMU as a nonlinear control system, we are able to conduct an observability analysis, which shows that this approach is capable of capturing an arbitrary spatial motion. We also show that additional triads only provide redundant information. Based on the control system formulation, we derive the models required to employ a nonlinear Kalman filter as a state observer. As the system description is of a general form they are suitable for any accelerometer array regardless of the number and placement of the transducers. Hence, the presented Kalman filter approach is applicable to all observable GF-IMU configurations. The measurements taken with a prototype on a 3-D rotation table confirm the observability analysis. The evaluations also show that the approach using three accelerometer triads achieves an estimation accuracy comparable with that of arrays employing a higher number of triads.

Application of fractional sensor fusion algorithms for inertial mems sensing

Abstract The work presents an extension of the conventional Kalman filtering concept for systems of fractional order (FOS). Modifications are introduced using the Grunwald‐Letnikov (GL) definition of the fractional derivative (FD) and corresponding truncation of the history length. Two versions of the fractional Kalman filter (FKF) are shown, where the FD is calculated directly or by augmenting the state vector with the estimate of the FD. The filters are compared to conventional integer order (IO) Position (P‐KF) and Position‐Velocity (PV‐KF) Kalman filters as well as to an adaptive Interacting Multiple‐Model Kalman Filter (IMM‐KF). The performance of the filters is assessed based on a hand and a head motion data set. The feasibility of the given approach is shown.

Pedestrian Indoor Localization Using Foot Mounted Inertial Sensors in Combination with a Magnetometer, a Barometer and RFID

A system for pedestrian indoor localization is presented, which uses the data of an inertial sensor unit mounted on the foot of a person walking through an indoor or outdoor environment. The inertial sensor data are integrated to a position/orientation information using a classical strapdown navigation approach, while several additional sensor data and constraints, such as Zero Velocity Updates, magnetometer and barometer readings and the detection of spatially distributed RFID tags, are incorporated to the solution using an Unscented Kalman Filter. The work presents a custom sensor system development, describes the developed algorithms and evaluates several methods to reduce the drift, which usually comes with the integration of low cost inertial sensors.

A low-cost shoe-mounted Inertial Navigation System with magnetic disturbance compensation

The work evaluates the performance of a custom-built shoe-mounted pedestrian indoor localization system based on commercially available low-cost inertial and magnetic sensors. This self-contained approach employs quaternions for attitude representation, whereas the estimation problem is formulated as a Kalman filter (KF) for INS strapdown mechanization equations. The paper compares three nonlinear KF techniques based on Unscented KF, Extended KF and Error-state Extended KF. Additionally, for poorly observable INS estimates, some magnetic field disturbance compensation mechanisms are implemented and tested for longer walking scenarios. The performance of the developed techniques was tested on a set of both indoor and outdoor paths.