Inertial sensors are widely utilized in smartphones, drones, vehicles, and wearable devices, playing a crucial role in enabling ubiquitous and reliable localization. Inertial sensor-based positioning is essential in various applications, including personal navigation, location-based security, and human-device interaction. However, low-cost MEMS inertial sensors’ measurements are inevitably corrupted by various error sources, leading to unbounded drifts when integrated doubly in traditional inertial navigation algorithms, subjecting inertial positioning to the problem of error drifts. In recent years, with the rapid increase in sensor data and computational power, deep learning techniques have been developed, sparking significant research into addressing the problem of inertial positioning. Relevant literature in this field spans across mobile computing, robotics, and machine learning. In this article, we provide a comprehensive review of deep learning-based inertial positioning and its applications in tracking pedestrians, drones, vehicles, and robots. We connect efforts from different fields and discuss how deep learning can be applied to address issues such as sensor calibration, positioning error drift reduction, and multi-sensor fusion. This article aims to attract readers from various backgrounds, including researchers and practitioners interested in the potential of deep learning-based techniques to solve inertial positioning problems. Our review demonstrates the exciting possibilities that deep learning brings to the table and provides a roadmap for future research in this field.

Accurate and robust localization is a fundamental need for mobile agents. Visual-inertial odometry (VIO) algorithms exploit the information from camera and inertial sensors to estimate position and translation. Recent deep learning based VIO models attract attentions as they provide pose information in a data-driven way, without the need of designing hand-crafted algorithms. Existing learning based VIO models rely on recurrent models to fuse multimodal data and process sensor signal, which are hard to train and not efficient enough. We propose a novel learning based VIO framework with external memory attention that effectively and efficiently combines visual and inertial features for states estimation. Our proposed model is able to estimate pose accurately and robustly, even in challenging scenarios, e.g., on overcast days and water-filled ground, which are difficult for traditional VIO algorithms to extract visual features. Experiments validate that it outperforms both traditional and learning based VIO baselines in different scenes.

Robust and accurate zero-velocity detection can improve the performance of zero-velocity-aided foot-mounted inertial navigation system. To ensure the accuracy of zero-velocity detection, we propose a novel detector based on contrastive learning. This detector roughly eliminates the inertial data that must not be the zero-velocity event in advance, to reduce the computation cost. Then the detector uses the remaining inertial data to detect the zero-velocity event via a trained contrastive neural network. The contrastive neural network uses the triplet network and is trained by comparing with the anchor data which consists of the known static inertial data from the period of initial alignment. The classifier will finally determine whether the output of the triplet network is the zero-velocity event. Two experiments were conducted to evaluate this novel detector, showing that it can adaptively and accurately detect the zero-velocity event. The horizontal position errors of the two experiments are respectively 1.33m over a 953m outdoor path with walking and 3.74m over a 1143m indoor/outdoor path with combined motion of low dynamic and high dynamic.

Aiming at the problems that the DVL calibration process in SINS/DVL integrated navigation is tedious and the measurement information is contaminated by outliers during the alignment of the moving base, we studied IABC fast calibration algorithm and MSTEKF alignment algorithm respectively. Firstly, we optimize the initial population using chaos strategy to improve the calibration accuracy of IABC and avoid the algorithm falling into local optimum. After the DVL calibration, we combined the M-estimation and STEKF to obtain MSTEKF, which was used to suppress the outlier interference in the alignment of the moving base. The experimental results show that IABC can effectively improve the DVL calibration accuracy, and the running speed of the algorithm is increased by about 41.47% compared with PSO. Compared with KF and SHAKF, MSTEKF can effectively suppress outlier interference, and the position alignment error is reduced by 90.00% and 61.31%, respectively.

In this paper, deep learning is introduced into the zero-speed detection of individual navigation. Through the combination of one-dimensional CNN and LSTM, the accuracy of zero-speed detection is further improved, and other cumbersome adjustment procedures based on thresholds and other methods are avoided. At the same time, a method of establishing a zero-speed label using ultrasonic ranging method is also demonstrated. Since the error of running and other sports mainly comes from the device error and not the algorithm itself, in this paper, the 6-axis IMU data is first imported into the one-dimensional CNN and is divided into a part that can perform zero-speed detection and a part that is not suitable for zero-speed detection. It can be used for zero speed detection and reuse LSTM based zero speed detection. Compared with other algorithms, this method uses the deep learning method to train the appropriate model in advance and can be used without adjusting the reference. It is especially suitable for the urgency of the soldier’s mission. The experimentally proven SHOE algorithm has a navigation accuracy of less than 3.5m/186.06m without adjustment, and the accuracy of the proposed method is stable above 1.0m/186.06m, which improves the overall navigation accuracy.

Accurately and reliably positioning pedestrians in satellite-denied conditions remains a significant challenge. Pedestrian dead reckoning (PDR) is commonly employed to estimate pedestrian location using low-cost inertial sensor. However, PDR is susceptible to drift due to sensor noise, incorrect step detection, and inaccurate stride length estimation. This work proposes ReLoc-PDR, a fusion framework combining PDR and visual relocalization using graph optimization. ReLoc-PDR leverages time-correlated visual observations and learned descriptors to achieve robust positioning in visually-degraded environments. A graph optimization-based fusion mechanism with the Tukey kernel effectively corrects cumulative errors and mitigates the impact of abnormal visual observations. Real-world experiments demonstrate that our ReLoc-PDR surpasses representative methods in accuracy and robustness, achieving accurte and robust pedestrian positioning results using only a smartphone in challenging environments such as less-textured corridors and dark nighttime scenarios.