Autonomous vehicles need a real-time traffic tracking system in order to interact with multiple moving vehicles in urban situations. Most existing works rely on camera sensors to interpret the surrounding environment. However, camera sensors degrade in conditions of lighting, shadows, and extreme weathers such that they can hardly detect objects. The light detection and ranging (LiDAR) sensors promise to be a good substitute, as they enable highly precise and robust localization across a wide range of conditions. This paper presents a new LiDAR-based online multiple maneuvering vehicle tracking problem and proposes a novel online multi-model smooth variable structure filter to address the problem. The real-time experiments show that our method is able to deliver superior performance compared to other conventional methods.

Bearing fault diagnosis is a well-developed field and an active area of research in which the combination of model-free machine learning techniques with vibration data has become a popular approach. However, vibration data from rotating machines has the potential to contain domain shifts beyond the accepted causes in this research area (different part models, operating conditions and sensor locations) which can enable data leakage between training and test datasets. To demonstrate the impact of data leakage, six common bearing diagnosis methods are applied to two datasets using three data splitting methods to compare classification performance. Diagnosis is preformed using Principal Component Analysis (PCA), Supervised Principal Component Analysis (SPCA) and Linear Discriminant Analysis (LDA) in combination with frequency analysis and envelope analysis feature extraction methods. Datasets from McMaster University and Paderborn University are used as experimental data sources, and produce vastly differing results (over a 40% drop in accuracy) depending on the selected dataset splitting method, revealing a previously unknown domain shift. Despite great results for diagnosis methods using frequency response analysis on the data from McMaster, these results are not expected to generalize due to possible data leakage. Out of fifty-five previous works using the Paderborn dataset, ten are identified as likely to be affected and only six properly address the problem. Recommendations are given for future experiment design, model creation and model evaluation.

Human self-motion perception largely relies on the integration of the visual, vestibular and proprioceptive systems. Much behavioral research has been conducted in order to understand this integration process; however, little is known about the online processes in humans during self-motion perception. Of the few studies to physically move human participants with full-body motion while recording the brain, most have used EEG due to its relative mobility. Past research provides evidence that multisensory self-motion perception elicits theta, alpha, and beta oscillations. It is important, however, to understand the individual contribution of each sense to fully understand how these oscillatory frequencies contribute to self-motion perception. To our knowledge, there has yet to be a study that directly compares the EEG correlates of visual self-motion with a no-motion physical input, versus physical-self motion with a no-motion visual input. We recorded event-related spectral power within a motion simulator controlled by a MOOG Stewart platform. Participants were given a visual or physical stimulus and made heading direction judgments. Compared to physical-only trials, visual-only trials produced earlier theta ERS and alpha ERD early in the trial, and more robust beta ERS late in the trial. We suggest beta-band power is likely associated with the process of visual-vestibular weighting. Moreover, within the right motor area, we found differences in theta power associated with left versus right headings. Theta ERS in the right motor area appears to be associated with heading processing for both the visual and vestibular systems but is minimally affected by multisensory weighting.

Machines designed to perform the same tasks using different technologies can be organized into families based on their similarities or differences. We are interested in identifying common properties and differences of such machines from raw sensor data for analysis and fault diagnostics. The usual first step is a feature extraction process that requires an understanding of the machine’s harmonics, bearing frequencies, etc. In this paper, we present a model-free path from the raw sensor data to statistically meaningful feature vectors. This is accomplished by defining a transform independent of the operating frequency and performing statistical reductions to identify the components with the largest variances, resulting in a low dimensional statistically meaningful feature space. To obtain an insight into the family relationships we perform a clustering. As the data set has some labeled characteristics we define an entropy-based measure to evaluate a clustering using the a priori-known labels, resulting in a symmetric measurement uniquely defining the clustering goal. Applying this hierarchically we obtain the family tree. The methods are presented can be applied in general situations. As a case study we apply them to a real data set of vibrating screens.

Many autonomous vehicles and advanced driver-assistance systems are equipped with front-facing cameras that detect and track objects using deep-learning-based algorithms. However, the localization capability of monocular cameras is often overlooked. In this paper, a novel method for estimating the pixel-wise error in a detected objects location versus its ground truth is proposed. As the object moves away from the camera, the pixel errors are shown to be normally distributed with unique spreads along the images vertical axis (y-pixel). The pixel error appears to be smaller as objects get farther away, while at the same distance range, objects have similar error distribution across the cameras horizontal view. The horizontal axis (x-pixel) error appears to be smaller while the distance moves further away. However, the x-pixel location along a constant y-pixel row has no impact on the error distribution. The estimated x and y-pixel error distributions can in turn be used to form a spatial error distribution for finding the location of a detected object within a certain confidence interval. The spatial errors are then projected onto the world coordinate system using a camera transformation matrix to give a more realistic sense of what this error means. The results show that location estimation using monocular cameras generates an elliptical error distribution around the object with a larger error in the y-pixel direction compared to the x-pixel direction. This error distribution can be important to fuse information from multiple range-detecting sensors as well as multi-vehicle and multi-object tracking. The uncertainty characterization for position measurement, as demonstrated in this paper is an essential element of tracking and, is sensor and algorithm dependent.