Tuomo Kröger

Accelerometer assisted robust wireless signal positioning based on a hidden Markov model

Reliable and accurate indoor positioning remains nowadays as one of the greatest challenges in the area of personal navigation and location based services (LBS). This manuscript proposes methods to improve the accuracy and robustness of indoor positioning using signal strength measurements of Wireless Local Area Networks (WLAN), and presents three aspects of contributions. First, the Weibull function is employed to represent the distribution of the signal strength over time. Thus, the impact of the signal strength variation on the fingerprinting database is mitigated, and fewer samples are required for training the database. Second, the accelerometer sensor is utilized to provide the pedestrian dynamics information, which is used to improve the positioning accuracy and reliability. Lastly, hidden Markov model (HMM) based particle filters are performed to compute the positioning solution through combining the signal strength measurements with the pedestrian dynamics information. Through the experimental evaluation of three scenarios, the proposed methods were found to improve significantly the accuracy and robustness of WLAN positioning. Due to their affordable computational load, the positioning methods proposed can be implemented for indoor navigation on mass-market mobile devices without any extra cost requirements.

Information filter with speed detection for indoor Bluetooth positioning

This paper studies indoor Bluetooth network. wireless positioning using a network of Bluetooth signals. Fingerprints of received signal strength indicators are used for localization. Due to the relatively long interval between the available consecutive Bluetooth signal strength measurements, we propose a method of information filtering with speed detection, which combines the estimation information from the received signal strength (RSS) measurements with the prior information from the motion model. Speed detection is further assisted to correct the outliers of position estimation. The field tests show that the new algorithm proposed applying information filter with speed detection improves the horizontal positioning accuracy of indoor navigation with about 17% compared to the static fingerprinting positioning method, achieving a 4.2 m positioning accuracy on the average, and about 16% improvement compared to the point Kalman filter.

Knowledge-based error detection and correction method of a Multi-sensor Multi-network positioning platform for pedestrian indoor navigation

For pedestrian indoor navigation, an accurate 2D/3D position is a premise for any further processing. Currently, indoor navigation is a challenging task for standalone GNSS technology. FGI has integrated self-contained sensors with wireless locating technology to investigate a hybrid indoor positioning solution. However, the infrastructure indoors inflicts multiple disturbances to the positioning sensors. A knowledge-based error detection and correction method is applied to detect and eliminate the occurring gross errors. Six modes of user dynamics are extracted from measurements of a barometer and an accelerometer, and such contexts can improve the positioning accuracy and enhance the user experience of the final navigation application.

The evaluation of WiFi positioning in a Bluetooth and WiFi coexistence environment

An increasing number of WiFi and Bluetooth terminals, tags, and other mobile devices drives a growing demand for integration and coexistence of these two technologies. This paper gives the preliminary results of WiFi positioning in a WiFi and Bluetooth coexistence environment. The received signal strength indication is introduced as an observation applied to our WiFi positioning. Then, we present the basis of a fingerprinting approach to WiFi positioning and analyze the characters and protocols of WiFi and Bluetooth networks. Furthermore, we present the interference in the WiFi and Bluetooth coexistence environments.

Motion Restricted Information Filter for Indoor Bluetooth Positioning

This paper studies wireless positioning using a network of Bluetooth signals. Fingerprints of received signal strength indicators RSSI are used for localization. Due to the relatively long interval between the available consecutive Bluetooth signal strength measurements, the authors propose a method of information filtering with speed detection, which combines the estimation information from the RSSI measurements with the prior information from the motion model. Speed detection is further assisted to correct the outliers of position estimation. The field tests show that the new algorithm proposed applying information filter with speed detection improves the horizontal positioning accuracy of indoor navigation with about 17% compared to the static fingerprinting positioning method, achieving a 4.2 m positioning accuracy on the average, and about 16% improvement compared to the point Kalman filter.

Method of pedestrian dead reckoning using speed recognition

Accurate and robust ubiquitous localization is one of the most demanding challenges that the navigation research community faces currently. The GNSS offers a perfect solution for open sky environment, however, in GPS-denied or unfavourable environments, for example urban canyons or indoors, the traditional GNSS standalone solution cannot provide the users position at a reasonable accuracy. Therefore, dead reckoning algorithms are one of the research fields which have attracted most research attention in the last few years. A multi-sensor based PDR (pedestrian dead reckoning) algorithm is presented in this paper, which makes full use of the measurements from a barometer and a 3-axis accelerometer. Users dynamics and motion modes are recognized with different speeds. The walking distance is then calculated based on this information and the human physiological characteristics. Hence, the PDR solution can be propagated and a continuous position solution of the user is available. Seven user dynamics with 3 different speeds are tested with the system. The algorithm recognizes the speed of the pedestrian both in horizontal and vertical direction, which makes the algorithm suitable to be applied in a multi-floor building which is a more complex task than navigation in a one floor construction.

Lane detection based on a visual-aided multiple sensors platform

Currently, GPS is the dominating technology for commercial car navigation applications, which offers acceptable navigation service for door to door navigation. However, within GPS degraded or denied areas, GPS standalone may not support sufficient accuracy and availability for a satisfactory user experience. Lane detection is an important and fundamental functionality to complete various future intelligent transportation system applications. In this paper, multiple candidate solutions for lane detection are investigated including single frequency precise point positioning (PPP), low-cost gyroscope, and visual-aided methods. These methods are compared against a traditional standalone GPS solution. The results of driving tests show that single frequency PPP can improve position accuracy by efficiently mitigating the ionospheric error but still cannot fulfill the lane detection functionality while the gyroscope and visual-aided methods show promising results.

Pulsewidth coding approach for multi-sensor synchronization of urban mobile mapping system

Mobile mapping system has become an irresistible trend used for 3D urban environment modeling, transportation engineering, road survey and other applications during last few years. A typical car-borne laser scanner based mobile mapping system today can be considered as a multi-sensor system that integrates various navigation devices and data acquisition sensors on a rigid, moving platform like a van or any other vehicle for determining the positions of the surrounding objects along the driving trajectory. All these sensors in the mobile mapping system are synchronized to the GPS time via geo-referencing system. With traditional synchronizing method, the hardware restriction of the geo-referencing system limits the number (kind) of sensors to be synchronized. The paper presents a synchronization approach which applies pulsewidth coding into synchronizing pulse generation. The approach enables geo-referencing system to be synchronized with more sensors and breaks the limitation of hardware. The accuracy of the approach is analyzed after a test bench is presented. The advantages and disadvantages of the approach are discussed correspondingly.