Barometer-Driven Height Estimation for Indoor Positioning Including Elevator and Escalator Detection Mechanisms

Kurzfassung

Indoor Positioning Systems have emerged as important technologies for locating and tracking objects or individuals within enclosed areas where global navigation satellite systems (GNSSs) such as global positioning system (GPS) are not available. These systems address a wide range of applications across various sectors, such as retail, healthcare, manufacturing, public safety, smart buildings, and other branches. For instance, location awareness might provide navigation services in airports, train stations, and large parking facilities or assist first responders in locating individuals during emergencies in complex structures. Besides horizontal location determination, accurate height estimation is crucial for any three-dimensional (3D) indoor navigation solution and is not yet fully solved. Traditional positioning systems, such as GNSSs, often fail to provide reliable and accurate positioning information indoors due to signal blockage, signal attenuation and multipath effects. This limitation has driven the need for alternative indoor positioning solutions. Over the last years, micro-electro-mechanical system (MEMS) sensors such as inertial measurement units (IMUs) have become more and more popular and gained significant importance for indoor positioning purposes. Compared to other techniques, IMU-based systems are relatively low-cost, offer infrastructure-free operation, and can be easily integrated into wearable devices, thanks to their compact centimeter- to millimeter-sized form factor. IMUs generally consist of accelerometers and gyroscopes, supplying 3D data on linear acceleration and rotational movement, respectively. In certain systems, a magnetometer is also included to provide heading information relative to the Earth's magnetic field, complementing the orientation capabilities of the IMU. The measurements of these sensors are affected by noise and inherent biases. Hence, when the position is calculated through double integration of the acceleration data, these inaccuracies can lead to significant exponential drift over time, compromising the reliability of the positional estimates. Mounting IMUs on the foot is advantageous for indoor positioning for several compelling reasons. Foot-mounted IMUs excel at accurately detecting the stance and swing phases of the foot while walking. During the stance phase, when the foot is in contact with the ground, the sensor registers a period of zero velocity. This characteristic enables the implementation of Zero-Velocity Updates (ZUPT), which are crucial for correcting accumulated errors and addressing the issue of drift. The process of estimating a person's position and orientation using body-mounted sensors is commonly referred to as pedestrian dead reckoning (PDR). A notable system in the realm of indoor positioning is the so called "NavShoe", developed by the German Aerospace Center (DLR). This technology utilizes a foot-mounted IMU to estimate the user's position and orientation within indoor environments. The NavShoe employs an Unscented Kalman Filter (UKF) to fuse inertial measurements and perform PDR. This advanced filtering technique propagates the mean and covariance of the estimates by utilizing multiple sigma points surrounding the estimated mean, subsequently measuring the mean and covariance of these propagated points. The UKF operates with a state vector comprising 15 states, which includes Euler angles, gyroscope bias, position, velocity, and accelerometer bias, each represented in three dimensions. During ZUPT phases, the UKF executes a virtual measurement update by assuming zero velocity, ensuring step-wise recalibration and enhancing overall positioning accuracy. Besides determining the position on a 2D map, height estimation is a crucial component of indoor positioning systems, particularly in multi-story buildings where vertical accuracy can significantly impact navigation and emergency response. Unlike horizontal positioning, which primarily involves tracking movement across a flat plane, vertical positioning requires precise determination of a user's elevation or floor level within a building. This information is vital for applications ranging from navigation assistance in complex structures to enabling first responders to quickly locate individuals during emergencies. However, estimating height displacement using PDR presents unique challenges that make it more difficult than horizontal displacement estimation. One of the primary difficulties stem from the nature of accelerometer data and the integration process required to derive position from acceleration measurements. Accelerometers measure linear acceleration, which must be integrated twice to obtain position. Any small error or bias in the acceleration data can lead to significant drift over time, especially in the vertical axis. This drift is intensified by the need to separate gravitational acceleration from actual movement-related accelerations accurately. Even minor inaccuracies in this separation can result in large positional errors after integration. To address the issue of height drift, we propose a height update algorithm that modifies the user's height estimate only upon detecting stair climbs by analyzing the user's step height through calculating the height difference between two consecutive steps. If the system does not identify stairs, the UKF will retain the previous height estimate . Given that the step height analysis depends on the PDR-based height, which is frequently updated by this algorithm, it is essential to keep both the stair detection algorithm and height update algorithm distinct from one another. This approach is anticipated to enhance both stair detection accuracy and PDR-based height estimation simultaneously. To the best of the author's knowledge, this represents the first investigation into such a method. Considering the availability of an integrated barometer within the IMU, atmospheric pressure data can provide valuable information for height estimation. Barometric height measurement is a technique used to determine the altitude or elevation of a location based on the observed atmospheric pressure, typically measured using capacitive- or piezo-resistive-based MEMS barometers. This method relies on the relationship between atmospheric pressure and altitude, as described by the hydrostatic equation and the ideal gas law. However, it is important to note that barometric pressure can be influenced by various factors, such as sensor bias, measurement noise, weather conditions, temperature changes, and air circulation within the building. To mitigate these effects, advanced filtering and calibration techniques are often employed, such as using reference barometers or differential pressure updates. By compensating for an initial elevation reference, this technique enables the measurement of relative and differential height changes, which can serve as a supplementary source of height information for the UKF measurement vector. Given that pedestrians frequently utilize elevators and escalators in indoor settings, the ability to accurately identify these modes of vertical transportation is crucial. The integration of barometric pressure sensors with the existing IMU technology in systems like the NavShoe can significantly enhance height estimation and facilitate the detection of elevators and escalators. By analyzing vertical speed and acceleration derived from pressure changes in conjunction with inertial data, we aim to develop a robust algorithm capable of distinguishing between these two forms of vertical movement based on their unique characteristics. This dual-sensor approach not only improves navigation accuracy but also enhances safety in emergency situations by providing precise location estimates within complex indoor environments. To evaluate the proposed algorithms, we conduct indoor measurements involving participants who walk up and down stairs, and utilize elevators and escalators. The results validate the effectiveness of the barometer as a reliable source for height estimation and as a method for detecting both elevators and escalators. The barometric readings consistently indicate a stable height, even in the presence of fluctuations in acceleration-based height estimates that stemmed from incorrectly detected stairs. In summary, this paper aims to improve height estimation for inertial indoor positioning across various floor changing scenarios . Unlike current state-of-the-art algorithms, which often face challenges with accuracy, we explore a more robust height update algorithm that integrates barometric measurements for both height estimation and the detection of elevators and escalators. Preliminary results from multiple indoor measurements show promising outcomes, demonstrating significant enhancements in the accuracy of height estimation.