Search

A highly accurate reference vehicle state is a requisite for the evaluation and validation of Autonomous Driving (AD) and Advanced Driver Assistance Systems (ADASs). This highly accurate vehicle state is usually obtained by means of Inertial Navigation Systems (INSs) that obtain position, velocity, and Course Over Ground (COG) correction data from Satellite Navigation (SatNav). However, SatNav is not always available, as is the case of roofed places, such as parking structures, tunnels, or urban canyons. This leads to a degradation over time of the estimated vehicle state. In the present paper, a methodology is proposed that consists on the use of a Machine Learning (ML)-method (Transformer Neural Network—TNN) with the objective of generating highly accurate velocity correction data from On-Board Diagnostics (OBD) data. The TNN obtains OBD data as input and measurements from state-of-the-art reference sensors as a learning target. The results show that the TNN is able to infer the velocity over ground with a Mean Absolute Error (MAE) of 0.167 kmh (0.046 ms) when a database of 3,428,099 OBD measurements is considered. The accuracy decreases to 0.863 kmh (0.24 ms) when only 5000 OBD measurements are used. Given that the obtained accuracy closely resembles that of state-of-the-art reference sensors, it allows INSs to be provided with accurate velocity correction data. An inference time of less than 40 ms for the generation of new correction data is achieved, which suggests the possibility of online implementation. This supports a highly accurate estimation of the vehicle state for the evaluation and validation of AD and ADAS, even in SatNav-deprived environments.

The prediction of the motion of traffic participants is a crucial aspect for the research and development of Automated Driving Systems (ADSs). Recent approaches are based on multi-modal motion prediction, which requires the assignment of a probability score to each of the multiple predicted motion hypotheses. However, there is a lack of ground truth for this probability score in the existing datasets. This implies that current Machine Learning (ML) models evaluate the multiple predictions by comparing them with the single real trajectory labeled in the dataset. In this work, a novel data-based method named Probabilistic Traffic Motion Labeling (PROMOTING) is introduced in order to (a) generate probable future routes and (b) estimate their probabilities. PROMOTING is presented with the focus on urban intersections. The generation of probable future routes is (a) based on a real traffic dataset and consists of two steps: first, a clustering of intersections with similar road topology, and second, a clustering of similar routes that are driven in each cluster from the first step. The estimation of the route probabilities is (b) based on a frequentist approach that considers how traffic participants will move in the future given their motion history. PROMOTING is evaluated with the publicly available Lyft database. The results show that PROMOTING is an appropriate approach to estimate the probabilities of the future motion of traffic participants in urban intersections. In this regard, PROMOTING can be used as a labeling approach for the generation of a labeled dataset that provides a probability score for probable future routes. Such a labeled dataset currently does not exist and would be highly valuable for ML approaches with the task of multi-modal motion prediction. The code is made open source.

Motion capture of unrestrained moving animals is a major analytic tool in neuroethology and behavioral physiology. At present, several motion capture methodologies have been developed, all of which have particular limitations regarding experimental application. Whereas marker-based motion capture systems are very robust and easily adjusted to suit different setups, tracked species, or body parts, they cannot be applied in experimental situations where markers obstruct the natural behavior (e.g., when tracking delicate, elastic, and/or sensitive body structures). On the other hand, marker-less motion capture systems typically require setup- and animal-specific adjustments, for example by means of tailored image processing, decision heuristics, and/or machine learning of specific sample data. Among the latter, deep-learning approaches have become very popular because of their applicability to virtually any sample of video data. Nevertheless, concise evaluation of their training requirements has rarely been done, particularly with regard to the transfer of trained networks from one application to another. To address this issue, the present study uses insect locomotion as a showcase example for systematic evaluation of variation and augmentation of the training data. For that, we use artificially generated video sequences with known combinations of observed, real animal postures and randomized body position, orientation, and size. Moreover, we evaluate the generalization ability of networks that have been pre-trained on synthetic videos to video recordings of real walking insects, and estimate the benefit in terms of reduced requirement for manual annotation. We show that tracking performance is affected only little by scaling factors ranging from 0.5 to 1.5. As expected from convolutional networks, the translation of the animal has no effect. On the other hand, we show that sufficient variation of rotation in the training data is essential for performance, and make concise suggestions about how much variation is required. Our results on transfer from synthetic to real videos show that pre-training reduces the amount of necessary manual annotation by about 50%., Competing Interests: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest., (Copyright © 2021 Arent, Schmidt, Botsch and Dürr.)

A current trend in automotive research is autonomous driving. For the proper testing and validation of automated driving functions a reference vehicle state is required. Global Navigation Satellite Systems (GNSS) are useful in the automation of the vehicles because of their practicality and accuracy. However, there are situations where the satellite signal is absent or unusable. This research work presents a methodology that addresses those situations, thus largely reducing the dependency of Inertial Navigation Systems (INSs) on the SatNav. The proposed methodology includes (1) a standstill recognition based on machine learning, (2) a detailed mathematical description of the horizontation of inertial measurements, (3) sensor fusion by means of statistical filtering, (4) an outlier detection for correction data, (5) a drift detector, and (6) a novel LiDAR-based Positioning Method (LbPM) for indoor navigation. The robustness and accuracy of the methodology are validated with a state-of-the-art INS with Real-Time Kinematic (RTK) correction data. The results obtained show a great improvement in the accuracy of vehicle state estimation under adverse driving conditions, such as when the correction data is corrupted, when there are extended periods with no correction data and in the case of drifting. The proposed LbPM method achieves an accuracy closely resembling that of a system with RTK.

Feedback is essential for skill acquisition as it helps identifying and correcting performance errors. Nowadays, Virtual Reality can be used as a tool to guide motor learning, and to provide innovative types of augmented feedback that exceed real world opportunities. Concurrent feedback has shown to be especially beneficial for novices. Moreover, watching skilled performances helps novices to acquire a motor skill, and this effect depends on the perspective taken by the observer. To date, however, the impact of watching one's own performance together with full body superimposition of a skilled performance, either from the front or from the side, remains to be explored. Here we used an immersive, state-of-the-art, low-latency cave automatic virtual environment (CAVE), and we asked novices to perform squat movements in front of a virtual mirror. Participants were assigned to one of three concurrent visual feedback groups: participants either watched their own avatar performing full body movements or were presented with the movement of a skilled individual superimposed on their own performance during movement execution, either from a frontal or from a side view. Motor performance and cognitive representation were measured in order to track changes in movement quality as well as motor memory across time. Consistent with our hypotheses, results showed an advantage of the groups that observed their own avatar performing the squat together with the superimposed skilled performance for some of the investigated parameters, depending on perspective. Specifically, for the deepest point of the squat, participants watching the squat from the front adapted their height, while those watching from the side adapted their backward movement. In a control experiment, we ruled out the possibility that the observed improvements were due to the mere fact of performing the squat movements-irrespective of the type of visual feedback. The present findings indicate that it can be beneficial for novices to watch themselves together with a skilled performance during execution, and that improvement depends on the perspective chosen., (Copyright © 2019 Hülsmann, Frank, Senna, Ernst, Schack and Botsch.)

In this paper, we present a method for automated estimation of a human face given a skull remain. Our proposed method is based on three statistical models. A volumetric (tetrahedral) skull model encoding the variations of different skulls, a surface head model encoding the head variations, and a dense statistic of facial soft tissue thickness (FSTT). All data are automatically derived from computed tomography (CT) head scans and optical face scans. In order to obtain a proper dense FSTT statistic, we register a skull model to each skull extracted from a CT scan and determine the FSTT value for each vertex of the skull model towards the associated extracted skin surface. The FSTT values at predefined landmarks from our statistic are well in agreement with data from the literature. To recover a face from a skull remain, we first fit our skull model to the given skull. Next, we generate spheres with radius of the respective FSTT value obtained from our statistic at each vertex of the registered skull. Finally, we fit a head model to the union of all spheres. The proposed automated method enables a probabilistic face-estimation that facilitates forensic recovery even from incomplete skull remains. The FSTT statistic allows the generation of plausible head variants, which can be adjusted intuitively using principal component analysis. We validate our face recovery process using an anonymized head CT scan. The estimation generated from the given skull visually compares well with the skin surface extracted from the CT scan itself., Competing Interests: The authors have declared that no competing interests exist.

Cartoon characters are omnipresent in popular media. While few studies have scientifically investigated their processing, in computer graphics, efforts are made to increase realism. Yet, close approximations of reality have been suggested to evoke sometimes a feeling of eeriness, the "uncanny valley" effect. Here, we used high-density electroencephalography to investigate brain responses to professionally stylized happy, angry, and neutral character faces. We employed six face-stylization levels varying from abstract to realistic and investigated the N170, early posterior negativity (EPN), and late positive potential (LPP) event-related components. The face-specific N170 showed a u-shaped modulation, with stronger reactions towards both most abstract and most realistic compared to medium-stylized faces. For abstract faces, N170 was generated more occipitally than for real faces, implying stronger reliance on structural processing. Although emotional faces elicited highest amplitudes on both N170 and EPN, on the N170 realism and expression interacted. Finally, LPP increased linearly with face realism, reflecting activity increase in visual and parietal cortex for more realistic faces. Results reveal differential effects of face stylization on distinct face processing stages and suggest a perceptual basis to the uncanny valley hypothesis. They are discussed in relation to face perception, media design, and computer graphics.

Neuropsychological assessment of human visual processing capabilities strongly depends on visual testing conditions including room lighting, stimuli, and viewing-distance. This limits standardization, threatens reliability, and prevents the assessment of core visual functions such as visual processing speed. Increasingly available virtual reality devices allow to address these problems. One such device is the portable, light-weight, and easy-to-use Oculus Rift. It is head-mounted and covers the entire visual field, thereby shielding and standardizing the visual stimulation. A fundamental prerequisite to use Oculus Rift for neuropsychological assessment is sufficient test-retest reliability. Here, we compare the test-retest reliabilities of Bundesen's visual processing components (visual processing speed, threshold of conscious perception, capacity of visual working memory) as measured with Oculus Rift and a standard CRT computer screen. Our results show that Oculus Rift allows to measure the processing components as reliably as the standard CRT. This means that Oculus Rift is applicable for standardized and reliable assessment and diagnosis of elementary cognitive functions in laboratory and clinical settings. Oculus Rift thus provides the opportunity to compare visual processing components between individuals and institutions and to establish statistical norm distributions.

Background: To increase the ecological validity of neuropsychological instruments the use of virtual reality (VR) applications can be considered as an effective tool in the field of cognitive neurorehabilitation. Despite the growing use of VR programs, only few studies have considered the application of everyday activities like shopping or travelling in VR training devices., Methods: We developed a novel 360°-VR supermarket, which is displayed on a circular arrangement of 8 touch-screens--the "OctaVis". In this setting, healthy human adults had to memorize an auditorily presented shopping list (list A) and subsequently buy all remembered products of this list in the VR supermarket. This procedure was accomplished on three consecutive days. On day four, a new shopping list (list B) was introduced and participants had to memorize and buy only products of this list. On day five, participants had to buy all remembered items of list A again, but without new presentation of list A. Additionally, we obtained measures of participants' presence, immersion and figural-spatial memory abilities. We also tested a sample of patients with focal epilepsy with an extended version of our shopping task, which consisted of eight days of training., Results: We observed a comprehensive and stable effect of learning for the number of correct products, the required time for shopping, and the length of movement trajectories in the VR supermarket in the course of the training program. Task performance was significantly correlated with participants' figural-spatial memory abilities and subjective level of immersion into the VR., Conclusions: Learning effects in our paradigm extend beyond mere verbal learning of the shopping list as the data show evidence for multi-layered learning (at least visual-spatial, strategic, and verbal) on concordant measures. Importantly, learning also correlated with measures of figural-spatial memory and the degree of immersion into the VR. We propose that cognitive training with the VR supermarket program in the OctaVis will be efficient for the assessment and training of real-life cognitive abilities in healthy subjects and patients with epilepsy. It is most likely that our findings will also apply for patients with cognitive disabilities resulting from other neurological and psychiatric syndromes.