Kalman filter is a commonly used method in the Global Navigation Satellite System (GNSS)/Inertial Navigation System (INS) integrated navigation system, in which the process noise covariance matrix has a significant influence on the positioning accuracy and sometimes even causes the filter to diverge when using the process noise covariance matrix with large errors. Though many studies have been done on process noise covariance estimation, the ability of the existing methods to adapt to dynamic and complex environments is still weak. To obtain accurate and robust localization results under various complex and dynamic environments, we propose an adaptive Kalman filter navigation algorithm (which is simply called RL-AKF), which can adaptively estimate the process noise covariance matrix using a reinforcement learning approach. By taking the integrated navigation system as the environment, and the opposite of the current positioning error as the reward, the adaptive Kalman filter navigation algorithm uses the deep deterministic policy gradient to obtain the most optimal process noise covariance matrix estimation from the continuous action space. Extensive experimental results show that our proposed algorithm can accurately estimate the process noise covariance matrix, which is robust under different data collection times, different GNSS outage time periods, and using different integration navigation fusion schemes. The RL-AKF achieves an average positioning error of 0.6517 m within 10 s GNSS outage for GNSS/INS integrated navigation system and 14.9426 m and 15.3380 m within 300 s GNSS outage for the GNSS/INS/Odometer (ODO) and the GNSS/INS/Non-Holonomic Constraint (NHC) integrated navigation systems, respectively.

Recently, the demand for human activity recognition has become more and more urgent. It is widely used in indoor positioning, medical monitoring, safe driving, etc. Existing activity recognition approaches require either the location information of the sensors or the specific domain knowledge, which are expensive, intrusive, and inconvenient for pervasive implementation. In this paper, a human activity recognition algorithm based on SDAE (Stacking Denoising Autoencoder) and LightGBM (LGB) is proposed. The SDAE is adopted to sanitize the noise in raw sensor data and extract the most effective characteristic expression with unsupervised learning. The LGB reveals the inherent feature dependencies among categories for accurate human activity recognition. Extensive experiments are conducted on four datasets of distinct sensor combinations collected by different devices in three typical application scenarios, which are human moving modes, current static, and dynamic behaviors of users. The experimental results demonstrate that our proposed algorithm achieves an average accuracy of 95.99%, outperforming other comparative algorithms using XGBoost, CNN (Convolutional Neural Network), CNN + Statistical features, or single SDAE.

Accurate stride-length estimation is a fundamental component in numerous applications, such as pedestrian dead reckoning, gait analysis, and human activity recognition. The existing stride-length estimation algorithms work relatively well in cases of walking a straight line at normal speed, but their error overgrows in complex scenes. Inaccurate walking-distance estimation leads to huge accumulative positioning errors of pedestrian dead reckoning. This paper proposes TapeLine, an adaptive stride-length estimation algorithm that automatically estimates a pedestrian’s stride-length and walking-distance using the low-cost inertial-sensor embedded in a smartphone. TapeLine consists of a Long Short-Term Memory module and Denoising Autoencoders that aim to sanitize the noise in raw inertial-sensor data. In addition to accelerometer and gyroscope readings during stride interval, extracted higher-level features based on excellent early studies were also fed to proposed network model for stride-length estimation. To train the model and evaluate its performance, we designed a platform to collect inertial-sensor measurements from a smartphone as training data, pedestrian step events, actual stride-length, and cumulative walking-distance from a foot-mounted inertial navigation system module as training labels at the same time. We conducted elaborate experiments to verify the performance of the proposed algorithm and compared it with the state-of-the-art SLE algorithms. The experimental results demonstrated that the proposed algorithm outperformed the existing methods and achieves good estimation accuracy, with a stride-length error rate of 4.63% and a walking-distance error rate of 1.43% using inertial-sensor embedded in smartphone without depending on any additional infrastructure or pre-collected database when a pedestrian is walking in both indoor and outdoor complex environments (stairs, spiral stairs, escalators and elevators) with natural motion patterns (fast walking, normal walking, slow walking, running, jumping).

With the explosive development of pervasive computing and the Internet of Things (IoT), indoor positioning and navigation have attracted immense attention over recent years. Pedestrian dead reckoning (PDR) is a potential autonomous localization technology that obtains position estimation employing built-in sensors. However, most existing PDR methods assume that the smartphone is held horizontally and points to the walking direction. To solve reckoning errors caused by inconsistency of headings between walking heading and pointing of smartphone, we design an accurate and robust PDR method based on walking patterns, which is identified by multihead convolutional neural networks. In addition to adaptively adjust the threshold of step detection and select the most suitable step length model according to the results of walking pattern recognition, a novel heading estimation approach independent of device orientation is proposed. To mitigate accumulative errors, we proposed an online trajectory calibration method based on forward and backward magnetic fingerprint trajectory matching. We conduct extensive and well-designed experiments in typical scenarios, and the experimental results indicate that the 75th percentile localization accuracy of the three scenarios is 1.06, 1.08, and 1.22 m, respectively, using the commercial smartphone embedded sensor without any dedicated infrastructures or training data. Despite the intricate pedestrian locomotion, the proposed PDR method has great potential in pedestrian positioning.

In the recent years, the availability of accurate vehicle position becomes more urgent. The global navigation satellite systems/inertial navigation system (GNSS/INS) is the most used integrated navigation scheme for land vehicles, which utilizes the Kalman filter (KF) to optimally fuse GNSS measurement and INS prediction for accurate and robust localization. However, the uncertainty of the process noise covariance and the measurement noise covariance has a significant impact on Kalman filtering performance. Traditional KF-based integrated navigation methods configure the process noise covariance and measurement noise covariance with predefined constants, which cannot adaptively characterize the various and dynamic environments, and obtain accurate and continuous positioning results under complex environments. To obtain accurate and robust localization results under various complex and dynamic environments, in this article, we propose a novel noise covariance estimation algorithm for the GNSS/INS-integrated navigation using multitask learning model, which can simultaneously estimate the process noise covariance and measurement noise covariance for the KF. The predicted multiplication factors are used to dynamically scale process noise covariance matrix and measurement noise covariance matrix respectively according to the inputs of raw inertial measurement. Extensive experiments are conducted on our collected practical road data set under three typical complex urban scenarios, such as, avenues, viaducts, and tunnels. Experimental results demonstrate that compared with the traditional KF-based integrated navigation algorithm with predefined fixed settings, our proposed method reduces 77.13% positioning error.

Smartphone based indoor positioning has greatly helped people in finding their positions in complex and unfamiliar buildings. One popular positioning method is by utilizing indoor magnetic field, because this feature is stable and infrastructure-free. In this method, the magnetometer embedded on the smartphone measures indoor magnetic field and queries its position. However, the environments of the magnetometer are rather harsh. This harshness mainly consists of coarse-grained hard/soft-iron calibrations and sensor electronic noise. The two kinds of interferences decrease the position distinguishability of the magnetic field. Therefore, it is important to extract location features from magnetic fields to reduce these interferences. This paper analyzes the main interference sources of the magnetometer embedded on the smartphone. In addition, we present a feature distinguishability measurement technique to evaluate the performance of different feature extraction methods. Experiments revealed that selected fingerprints will improve position distinguishability.