Inertial Measurement Unit Data Research Articles

ASL-SLAM: A LiDAR SLAM With Activity Semantics-Based Loop Closure

A practical back-end module with loop closure detection is very useful and important for a LiDAR simultaneous localization and mapping (SLAM) system to perform high-precision positioning and mapping tasks. However, most existing loop closure detection methods are based on images or point clouds, and these methods may produce errors when the structure or texture is similar. To overcome this problem, we propose a complete LiDAR SLAM system, including a front-end odometry module based on normal distribution transform (NDT)-LOAM and a back-end optimization module with loop closure based on activity semantics. Through the analysis and calculation of inertial measurement unit (IMU) data from SLAM platforms such as unmanned ground vehicles (UGVs), the activity semantics of turning and passing over a speed bump are detected based on the peak <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${z}$ </tex-math></inline-formula> -axis angular velocity and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${z}$ </tex-math></inline-formula> -axis acceleration, respectively. Then, according to this activity semantics information and its unique and definite attributes, we establish correct loop closure detection using rough geometric detection, activity semantics matching, and point cloud rematching for validation. Finally, graph optimization theory is utilized to reduce the global cumulative error, improve the global trajectory accuracy and map consistency, and obtain the final global motion trajectory and point cloud map. We collected a dataset for evaluation, which contains indoor data, outdoor data, and indoor–outdoor integration data, and we also evaluated our method on the KITTI dataset. The experimental results for different scenes show that the addition of activity semantics can effectively help loop closure detection and improve LiDAR SLAM system performance.

Research on positioning and mapping algorithm of sliding window optimization for substation monitoring robot

This paper proposed a positioning and mapping algorithm of sliding window optimization for substation monitoring robots to solve the problems of low precision location and poor robustness of the existing laser odometer in power inspection outdoor scene mapping. A tightly coupled simultaneous localization and mapping (SLAM) algorithm was proposed based on 16-wire LiDAR and inertial measurement unit (IMU). Firstly, the paper estimated the IMU and corrected the motion distortion of the laser point cloud by linear interpolation. Secondly, scene features were extracted by curvature and classified according to different feature properties. The local map was constructed in the sliding window using the inter-frame matching module. Finally, the joint optimization function was built using the distance and IMU data obtained by matching the frame with the local map. The paper used the KITTI and self-recorded datasets to conduct the experiments. The results show that the improved method’s accuracy outperforms the lightweight and ground-optimized LiDAR odometry and Mapping (Lego-LOAM) and LiDAR inertial odometry and mapping (LIO-Mapping) and draws a broad application prospect in power inspection field.

Real-time machine learning-based recognition of human thermal comfort-related activities using inertial measurement unit data

The real-time detection of indoor thermal comfort can bring significant benefits for energy-efficient controls over the heating, ventilation and air conditioning (HVAC) systems. At present, few studies have been conducted to propose non-intrusive and cost-effective solutions for the real-time sensing of individual thermal comfort. Leveraging advances in machine learning, this study proposes a data-driven method for real-time recognition of thermal comfort-related activities based on human inertial measurement unit (IMU) data. Using wearable devices at human wrists, experiments have been designed to collect prototype IMU data on 30 thermal comfort-related activities from building occupants. An end-to-end data analytic framework, which consists of offline training and online detection modules, has been developed for practical applications. Various feature engineering and state-of-the-art machine learning techniques have been implemented and tested to derive optimal data-driven solutions, leading to an activity recognition accuracy of 86.2%. The methods proposed provide an automated and feasible approach for thermal comfort-related activity recognition, which can be integrated with HVAC systems for smart and customized controls, such as personalized ventilation and air-conditioning.

Lower Extremity Inverse Kinematics Results Differ Between Inertial Measurement Unit- and Marker-Derived Gait Data.

In-lab, marker-based gait analyses may not represent real-world gait. Real-world gait analyses may be feasible using inertial measurement units (IMUs) in combination with open-source data processing pipelines (OpenSense). Before using OpenSense to study real-world gait, we must determine whether these methods estimate joint kinematics similarly to traditional marker-based motion capture (MoCap) and differentiate groups with clinically different gait mechanics. Healthy young and older adults and older adults with knee osteoarthritis completed this study. We captured MoCap and IMU data during overground walking at 2 speeds. MoCap and IMU kinematics were computed with OpenSim workflows. We tested whether sagittal kinematics differed between MoCap and IMU, whether tools detected between-group differences similarly, and whether kinematics differed between tools by speed. MoCap showed more anterior pelvic tilt (0%-100% stride) and joint flexion than IMU (hip: 0%-38% and 61%-100% stride; knee: 0%-38%, 58%-89%, and 95%-99% stride; and ankle: 6%-99% stride). There were no significant tool-by-group interactions. We found significant tool-by-speed interactions for all angles. While MoCap- and IMU-derived kinematics differed, the lack of tool-by-group interactions suggests consistent tracking across clinical cohorts. Results of the current study suggest that IMU-derived kinematics with OpenSense may enable reliable evaluation of gait in real-world settings.

Smartphone Inertial Measurement Unit Data Features for Analyzing Driver Driving Behavior

Driving behavior is an important aspect of maintaining and sustaining safe transport on the roads. It also directly affects fuel consumption, traffic flow, public health, and air pollution along with psychology and personal mental health. For advanced driving assistance systems (ADAS) and autonomous vehicles, predicting driver behavior helps to facilitate interaction between ADAS and the human driver. Consequently, driver behavior prediction has emerged as an important research topic and has been investigated largely during the past few years. Often, the investigations are based on simulators and controlled environments. Driving behavior can be inferred using control actions, visual monitoring, and inertial measurement unit (IMU) data. This study leverages the IMU data recorded using a smartphone placed inside the vehicle. The dataset contains the accelerometer and gyroscope data recorded from the real traffic environment. Extensive experiments are performed regarding the use of a different set of features, the combination of original and derived features, and binary vs multi-class classification problems; a total of six scenarios are considered. Results reveal that ’timestamp’ is the most important feature and using it with accelerometer and gyroscope features can lead to a 100% accuracy for driver behavior prediction. Without using the ’timestamp’ feature, the number of wrong predictions for ’slow’ and ’normal’ classes is high due to the feature space overlap. Although derived features can help elevate the performance of the models, the models show inferior performance to that of using the ’timestamp’ feature. Deep learning models tend to show poor performance than machine learning models where random forest and extreme gradient boosting machines show a 100% accuracy for multi-class classification.

Spatial-Temporal Analysis of Gait in Amyotrophic Lateral Sclerosis Using Foot-Worn Inertial Sensors: An Observational Study

Introduction: Amyotrophic lateral sclerosis (ALS) is a progressive neurodegenerative disease that alters gait and increases the risk of falls. The current model of care involves in-person multidisciplinary clinic visits to, in part, assess alterations in gait, evaluate safety, and make recommendations for management. Clinic visits, however, are relatively infrequent, and multidisciplinary evaluations can be physically demanding for patients. To better understand how gait changes over time in those with ALS and enable healthcare providers to properly respond to these changes, remote monitoring of functional mobility would be advantageous. Methods: The objective of this study was to remotely track long-term changes in walking speed using wearable inertial measurement units (IMUs). Nine ALS patients and 6 healthy controls submitted twice-weekly home walking recordings for 24 and 4 weeks, respectively. An IMU data processing method was developed and validated against laboratory-measured walking speed. Results: For both ALS patients and healthy controls, home walking speed was less than clinic walking speed by an average of 0.19 m/s (p = 0.0024). Over 24 weeks, home walking speed significantly decreased for 5 of 9 ALS patients at an average of −0.021 m/s/months (p = 0.005). Those who eventually transitioned to using assistive device (AD) while on the study demonstrated a greater decrease in walking speed than those who did not. Conclusions: Remote longitudinal gait monitoring of ALS patients is feasible with the use of an IMU. Decreases in walking speed were detected in the majority of patients, most strongly in those who eventually transitioned to an AD. Home walking speed may more accurately represent the walking abilities of ALS patients in their real-life environments, a finding which further supports the case for remote monitoring in ALS.

MOBILE LASER SCANNING WITH LOW-COST NAVIGATION SENSORS: COMPENSATING FOR LOW-GRADE IMU WITH DUAL-GNSS AND TIGHTLY-COUPLED LIDAR

Abstract. Kinematic laser scanning enables efficient and accurate acquisition of 3D data, specifically point clouds. Transforming the laser ranging measurements from the scanner coordinate system to a georeferenced coordinate system requires integration with auxiliary navigation systems, in this case using GNSS (global navigation satellite system) and an IMU (inertial measurement unit). GNSS and IMU data are commonly fused via Kalman filter, and the resulting trajectory is used for georeferencing. Errors in the trajectory propagate through the georeferencing and cause discrepancies in the point cloud. To mitigate this, we present a holistic integration method incorporating GNSS, IMU and LiDAR measurements in a single adjustment, which can be seamlessly adapted to a setup with two GNSS receivers and antennas. This tight coupling of LiDAR and IMU together with the dual-GNSS setup allows the use of an ultra low-cost IMU while still achieving high-quality point clouds. We demonstrate this methodology on two datsets, where we discuss in particular the boresight calibration of such a system, and the impact of the LiDAR measurements and the dual-GNSS set-up on the trajectory and the point cloud.

Reconstructing 3D human pose and shape from a single image and sparse IMUs.

This article presents an approach to improve 3D human pose estimation by fusing a single image with sparse inertial measurement units (IMUs). Based on a dual-stream feature extract network, we design a model-attention network with a residual module to closely couple the dual-modal feature from a static image and sparse inertial measurement units. The final 3D pose and shape parameters are directly obtained by a regression strategy. Extensive experiments are conducted on two benchmark datasets for 3D human pose estimation. Compared to state-of-the-art methods, the per vertex error (PVE) of human mesh reduces by 9.4 mm on Total Capture dataset and the mean per joint position error (MPJPE) reduces by 7.8 mm on the Human3.6M dataset. The quantitative comparison demonstrates that the proposed method could effectively fuse sparse IMU data and images and improve pose accuracy.

Physical match demands of four LIQUI-MOLY Handball-Bundesliga teams from 2019-2022: effects of season, team, match outcome, playing position, and halftime.

Due to the development in team handball, there is a need to optimize the physical capacities of team handball players for which knowledge of the physical match demands is essential. The aim of this study was to investigate the physical match demands of four LIQUI-MOLY Handball-Bundesliga (HBL) teams across three seasons with respect to the effects of season, team, match outcome, playing position, and halftime. A fixed installed local positioning system (Kinexon) was used, collecting 2D positional and 3D inertial measurement unit data at 20 and 100 Hz, respectively. The physical match demands were operationalized by basic (e.g., distance, speed, and acceleration) and more advanced variables (e.g., jumps, throws, impacts, acceleration load, and metabolic power). A total of 347 matches (213 with an additional ball tracking) were analyzed from four teams (one top, two middle, and one lower ranked) during three consecutive seasons (2019-2022). One-way ANOVAs were calculated to estimate differences between more than two groups (e.g., season, team, match outcome, playing position). Mean differences between halftimes were estimated using Yuen's test for paired samples. Large effects were detected for the season (), team (), and playing position (). Medium effects were found for match outcome () and halftime (). For the first time, we provide a comprehensive analysis of physical match demands in handball players competing in the LIQUI-MOLY Handball-Bundesliga. We found that physical match demands differ on that top-level with up to large effect sizes concerning the season, team, match outcome, playing position, and halftime. Our outcomes can help practitioners and researchers to develop team and player profiles as well as to optimize talent identification, training, regeneration, prevention, and rehabilitation procedures.

Fault-Tolerant Control of a Variable-Pitch Quadrotor under Actuator Loss of Effectiveness and Wind Perturbations

The actuator fault-tolerant control problem for a variable-pitch quadrotor is addressed under uncertain conditions. Following a model-based approach, the plant nonlinear dynamics are faced with a disturbance observer-based control and a sequential quadratic programming control allocation, where only kinematic data of the onboard inertial measurement unit are required for the fault-tolerant control, i.e., it does not require the measurement of the motor speed nor the current drawn by the actuators. In the case of almost horizontal wind, a single observer handles both faults and the external disturbance. The estimation of the wind is fed forward by the controller, while the actuator fault estimation is exploited in the control allocation layer, which copes with the variable-pitch nonlinear dynamics, thrust saturation, and rate limits. Numerical simulations in the presence of measurement noise show the capability of the scheme to handle multiple actuator faults in a windy environment.

Pedestrian Localization with Stride-Wise Error Estimation and Compensation by Fusion of UWB and IMU Data.

Indoor positioning enables mobile machines to perform tasks (semi-)automatically, such as following an operator. However, the usefulness and safety of these applications depends on the reliability of the estimated operator localization. Thus, quantifying the accuracy of positioning at runtime is critical for the application in real-world industrial contexts. In this paper, we present a method that produces an estimate of the current positioning error for each user stride. To accomplish this, we construct a virtual stride vector from Ultra-Wideband (UWB) position measurements. The virtual vectors are then compared to stride vectors from a foot-mounted Inertial Measurement Unit (IMU). Using these independent measurements, we estimate the current reliability of the UWB measurements. Positioning errors are mitigated through loosely coupled filtering of both vector types. We evaluate our method in three environments, showing that it improves positioning accuracy, especially in challenging conditions with obstructed line of sight and sparse UWB infrastructure. Additionally, we demonstrate the mitigation of simulated spoofing attacks on UWB positioning. Our findings indicate that positioning quality can be judged at runtime by comparing user strides reconstructed from UWB and IMU measurements. Our method is independent of situation- or environment-specific parameter tuning, and as such represents a promising approach for detecting both known and unknown positioning error states.

Fusion of video and inertial sensing data via dynamic optimization of a biomechanical model

Inertial sensing and computer vision are promising alternatives to traditional optical motion tracking, but until now these data sources have been explored either in isolation or fused via unconstrained optimization, which may not take full advantage of their complementary strengths. By adding physiological plausibility and dynamical robustness to a proposed solution, biomechanical modeling may enable better fusion than unconstrained optimization. To test this hypothesis, we fused video and inertial sensing data via dynamic optimization with a nine degree-of-freedom model and investigated when this approach outperforms video-only, inertial-sensing-only, and unconstrained-fusion methods. We used both experimental and synthetic data that mimicked different ranges of video and inertial measurement unit (IMU) data noise. Fusion with a dynamically constrained model significantly improved estimation of lower-extremity kinematics over the video-only approach and estimation of joint centers over the IMU-only approach. It consistently outperformed single-modality approaches across different noise profiles. When the quality of video data was high and that of inertial data was low, dynamically constrained fusion improved estimation of joint kinematics and joint centers over unconstrained fusion, while unconstrained fusion was advantageous in the opposite scenario. These findings indicate that complementary modalities and techniques can improve motion tracking by clinically meaningful margins and that data quality and computational complexity must be considered when selecting the most appropriate method for a particular application.

Real-time propeller fault detection for multirotor drones based on vibration data analysis

This article presents a Fault Detection (FD) method to deal with propeller faults on multirotor drones in real-time. Several solutions have been proposed in the literature, however, they depend on additional sensors and/or dedicated hardware to deal with heavy computational complexity. So, they cannot be implemented in off-the-shelf commercial devices, i.e., without the aid of additional on-board sensors and/or extra computational power. The proposed method, instead, requires the on-board Inertial Measurement Unit (IMU) data only: by combining Finite Impulse Response (FIR), together with sparse classifiers, only a subset of the features is actually needed online and the FD is thus feasible in real-time.Design and tests are based on real flight data from a hexarotor, equipped with a conventional ArduPilot-based controller. The classification accuracy in testing is up to 93.37% (98.21%) with a binary tree (Linear Support Vector Machine (LSVM)). Moreover, the space and time complexity of the proposed method is low: on a PixHawk Cube flight controller, it requires less than 2% of the cycle time, and can then run in real-time. Finally, the proposed fault detection solution is model-free and it can be easily generalized to other multirotor vehicles.

Simplified Markerless Stride Detection Pipeline (sMaSDP) for Surface EMG Segmentation.

To diagnose mobility impairments and select appropriate physiotherapy, gait assessment studies are often recommended. These studies are usually conducted in confined clinical settings, which may feel foreign to a subject and affect their motivation, coordination, and overall mobility. Conducting gait studies in unconstrained natural settings instead, such as the subject's Activities of Daily Life (ADL), could provide a more accurate assessment. To appropriately diagnose gait deficiencies, muscle activity should be recorded in parallel with typical kinematic studies. To achieve this, Electromyography (EMG) and kinematic are collected synchronously. Our protocol sMaSDP introduces a simplified markerless gait event detection pipeline for the segmentation of EMG signals via Inertial Measurement Unit (IMU) data, based on a publicly available dataset. This methodology intends to provide a simple, detailed sequence of processing steps for gait event detection via IMU and EMG, and serves as tutorial for beginners in unconstrained gait assessment studies. In an unconstrained gait experiment, 10 healthy subjects walk through a course designed to mimic everyday walking, with their kinematic and EMG data recorded, for a total of 20 trials. Five different walking modalities, such as level walking, ramp up/down, and staircase up/down are included. By segmenting and filtering the data, we generate an algorithm that detects heel-strike events, using a single IMU, and isolates EMG activity of gait cycles. Applicable to different datasets, sMaSDP was tested in healthy gait and gait data of Parkinson's Disease (PD) patients. Using sMaSDP, we extracted muscle activity in healthy walking and identified heel-strike events in PD patient data. The algorithm parameters, such as expected velocity and cadence, are adjustable and can further improve the detection accuracy, and our emphasis on the wearable technologies makes this solution ideal for ADL gait studies.

PDGes: An Interpretable Detection Model for Parkinson’s Disease Using Smartphones

Parkinson’s disease (PD) is a neurodegenerative disorder that severely affects the motor system of patients. Early PD detection will greatly improve the quality of lives. However, existing automatic PD detection systems either rely on customized sensors or require users to perform special activities, using machine learning models whose prediction process is not understandable by medical professionals. In this article, we develop a non-disruptive PD detection system on smartphones based on interpretable prediction models. We design an application named PDGes to passively collect touchscreen and Inertial Measurement Unit data of users’ tapping and swiping actions on smartphones. Meaningful features that reflect finger dexterity , tremor , stiffness , and hand movement are extracted to build the prediction model. To better comprehend the decisions made by the model, we conduct a systematic analysis of feature importance to help validate the conformity of the model with clinical PD diagnosis. We collected data from 108 volunteers to evaluate the performance of PDGes . The experiment results show that PDGes achieves a detection accuracy of more than 94.5% on different smartphones.

Semisupervised Learning for Noise Suppression Using Deep Reinforcement Learning of Contrastive Features

In this work, we present DeDRLSSL a generic semi-supervised noise suppression framework. The proposed model is based on a reinforcement learning (RL) system for learning contrastive features to refine the features utilized in consistency matching for semi-supervised learning (SSL). The proposed method outperforms the state-of-the-art supervised models in terms of error compensation for Inertial Measurement Unit (IMU) data from various evaluation metrics and improves the baselines for yaw estimation on average by 38% and 28% across the benchmarks for 30% and 50% of labeled data, respectively. Our approach can be adapted to any SSL approach to compensate for the problem of label scarcity.

Deep Learning for Laying Hen Activity Recognition Using Wearable Sensors

Laying hen activities in modern intensive housing systems can dramatically influence the policies needed for the optimal management of such systems. Intermittent monitoring of different behaviors during daytime cannot provide a good overview, since daily behaviors are not equally distributed over the day. This paper investigates the application of deep learning technology in the automatic recognition of laying hen behaviors equipped with body-worn inertial measurement unit (IMU) modules in poultry systems. Motivated by the human activity recognition literature, a sophisticated preprocessing method is tailored on the time-series data of IMU, transforming it into the form of so-called activity images to be recognized by the deep learning models. The diverse range of behaviors a laying hen can exhibit are categorized into three classes: low-, medium-, and high-intensity activities, and various recognition models are trained to recognize these behaviors in real-time. Several ablation studies are conducted to assess the efficacy and robustness of the developed models against variations and limitations common for an in situ practical implementation. Overall, the best trained model on the full-feature acquired data achieves a mean accuracy of almost 100%, where the whole process of inference by the model takes less than 30 milliseconds. The results suggest that the application of deep learning technology for activity recognition of individual hens has the potential to accurately aid successful management of modern poultry systems.

SensePred: Guiding Video Prediction by Wearable Sensors

Video prediction has been studied in several earlier works. However, there are two inherent limitations in existing solutions, one being the ambiguity concern and the other being the degradation problem in lengthy predictions. To overcome these limitations, this work studies using wearable inertial sensors to guide video prediction. We create a data set called Pedestrians with IMU (Ped-IMU) that records people walking around with the wearable devices to collect the relation between inertial measurement unit (IMU) and video data and propose a SensePred model to conquer these limitations. The model takes full wearable sensor information and partial video information as inputs and predicts the missing video information. Simulation results validate the effectiveness of our model.

LSTM-Based Projectile Trajectory Estimation in a GNSS-Denied Environment.

This paper presents a deep learning approach to estimate a projectile trajectory in a GNSS-denied environment. For this purpose, Long-Short-Term-Memories (LSTMs) are trained on projectile fire simulations. The network inputs are the embedded Inertial Measurement Unit (IMU) data, the magnetic field reference, flight parameters specific to the projectile and a time vector. This paper focuses on the influence of LSTM input data pre-processing, i.e., normalization and navigation frame rotation, leading to rescale 3D projectile data over similar variation ranges. In addition, the effect of the sensor error model on the estimation accuracy is analyzed. LSTM estimates are compared to a classical Dead-Reckoning algorithm, and the estimation accuracy is evaluated via multiple error criteria and the position errors at the impact point. Results, presented for a finned projectile, clearly show the Artificial Intelligence (AI) contribution, especially for the projectile position and velocity estimations. Indeed, the LSTM estimation errors are reduced compared to a classical navigation algorithm as well as to GNSS-guided finned projectiles.

Gyro-Net: IMU Gyroscopes Random Errors Compensation Method Based on Deep Learning

To solve the problem of inaccurate orientation estimation after long-term operations of the Inertial Measurement Unit (IMU), we present a learning-based method (called Gyro-Net) to estimate and compensate for IMU gyroscope random errors. We firstly introduce a semi-dense network structure, which extracts different scale features by IFES Block (IMU Feature Extraction & Selection Block), and adopts skip-connections and transition layers to adjust the feature pipeline. In this way, we can reuse features between different blocks before and after feature extraction, selection and compression. Driven by a proposed absolute and relative loss, the network can be trained and achieve the reduction of cumulative estimated orientation errors. The experimental results in public datasets show that our method can effectively and accurately estimate the orientation from raw IMU data. Moreover, we apply the network output directly to Open-VINS, and the results show that Gyro-Net can improve the accuracy of pose estimation for Open-VINS, especially in scenarios where camera-based estimation often struggles (e.g., fast motion, drastic lighting, viewpoint changes and motion blur).