No description available.

Autonomous robots are set to become ubiquitous in the future, with applications ranging from autonomous cars to assistive household robots. These systems must operate in close proximity of dynamic and static objects, including humans and other non-autonomous systems, adding complexity to their decision-making processes. The behaviour of such objects is often stochastic and hard to predict. Making robust decisions under such uncertain scenarios can be challenging for these autonomous robots. In the past, researchers have used deterministic approach to predict the motion of surrounding objects. However, these approaches can be over-confident and do not capture the stochastic behaviour of surrounding objects necessary for safe decision-making. In this dissertation, we show the importance of probabilistic prediction of surrounding dynamic objects and their incorporation into planning for safety-critical decision making. We utilise Bayesian inference models such as Monte Carlo dropout and deep ensemble to probabilistically predict the motion of surrounding objects. Our probabilistic trajectory forecasting model showed improvement over standard deterministic approaches and could handle adverse scenarios such as sensor noise and occlusion during prediction. The uncertainty-inclusive prediction of surrounding objects has been incorporated into planning. The inclusion of predicted states of surrounding objects with associated uncertainty enables the robot make proactive decisions while avoiding collisions. / Doctor of Philosophy / In future, humans will greatly rely on the assistance of autonomous robots in helping them with everyday tasks. Drones to deliver packages, cars for driving to places autonomously and household robots helping with day-to-day activities. In all such scenarios, the robot might have to interact with their surrounding, in particular humans. Robots working in close proximity to humans must be intelligent enough to make safe decisions not affecting or intruding the human. Humans, in particular make abrupt decisions and their motion can be unpredictable. It is necessary for the robot to understand the intention of human for navigating safely without affecting the human. Therefore, the robot must capture the uncertain human behaviour and predict its future motion so that it can make proactive decisions. We propose to capture the stochastic behaviour of humans using deep learning based prediction models by learning motion patterns from real human trajectories. Our method not only predicts future trajectory of humans but also captures the associated uncertainty during prediction. In this thesis, we also propose how to predict human motion under adverse scenarios like bad weather leading to noisy sensing as well as under occlusion. Further, we integrate the predicted stochastic behaviour of surrounding humans into the planning of the robot for safe navigation among humans.

Driving Behavior Analysis (DBA) plays a pivotal role in designing intelligent transportation systems, enhancing road safety, and advancing Autonomous Vehicles (AVs). Driver identification, as a key aspect of DBA, has the potential to provide unprecedented opportunities for enhanced security and driver profiling. However, the current solutions for driver identification suffer from demanding extensive data collection, limited scalability, and inadequate generalization. Furthermore, DBA is also essential for training AVs, addressing the main challenges they face: accurately perceiving their surroundings to make informed decisions and to navigate safely, and effectively handling unforeseen scenarios. In the first part of this thesis, we concentrate on behavior analysis for driver identification and verification and design two novel schemes aiming to reduce data dependency and enhance the generalization ability of existing approaches. First, we propose a novel driver identification model, called DriverRep, which reduces data dependency by presenting a fully unsupervised triplet loss training. DriverRep is the first model that extracts the latent representations associated with each driver, called driver embeddings, in an unsupervised manner. In addition, we develop a novel model to tackle driver verification and impostor detection tasks based on DBA and extracted driver embeddings. In the second part, we focus on behavior prediction for AVs and their surrounding agents. First, we tackle behavior prediction in dynamic and complex scenarios by introducing three novel prediction models for forecasting drivers intentions and behaviors at unsignalized intersections. We then address social reasoning by proposing a novel prediction model that analyzes agent interactions using graph neural networks, making the scene understanding process more informative for AVs. Our proposed prediction model, called STAG, explicitly activates social modeling with a directed graph representation while considering spatial and temporal inter-agent correlations. We further design a novel prediction system, namely CAPHA, which conditions the future behavior of agents on grid-based plans modeled as a Markov decision process and solves the prediction task via inverse reinforcement learning to produce scene compliant behaviors. Moreover, we introduce a novel goal-based prediction model, called GMP, which encodes interactions between agents and dynamic and static context information to estimate the distribution of target goals, efficiently considering the inherent uncertainty in agents behavior. Extensive quantitative and qualitative comparisons have been conducted between the developed solutions and related benchmark schemes using various scenarios and environments. The obtained results demonstrate the potential of these solutions for the understudy tasks of DBA and real-world applications.

Emerging connected and autonomous vehicle technologies (CAV) provide an opportunity to improve highway capacity and reduce adverse impacts of stop-and-go traffic. To realize the potential benefits of CAV technologies, this study provides insightful methodological and managerial tools in microscopic and macroscopic traffic scales. In the macroscopic scale, this dissertation proposes an analytical method to formulate highway capacity for a mixed traffic environment where a portion of vehicles are CAVs and the remaining are human-driven vehicles (HVs). The proposed analytical mixed traffic highway capacity model is based on a Markov chain representation of spatial distribution of heterogeneous and stochastic headways. This model captures not only the full spectrum of CAV market penetration rates but also all possible values of CAV platooning intensities that largely affect the spatial distribution of different headway types. Numerical experiments verify that this analytical model accurately quantifies the corresponding mixed traffic capacity at various settings. This analytical model allows for examination of the impact of different CAV technology scenarios on mixed traffic capacity. We identify sufficient and necessary conditions for the mixed traffic capacity to increase (or decrease) with CAV market penetration rate and platooning intensity. These theoretical results caution scholars not to take CAVs as a sure means of increasing highway capacity for granted but rather to quantitatively analyze the actual headway settings before drawing any qualitative conclusion. In the microscopic scale, this study develops innovative control strategies to smooth highway traffic using CAV technologies. First, it formulates a simplified traffic smoothing model for guiding movements of CAVs on a general one-lane highway segment. The proposed simplified model is able to control the overall smoothness of a platoon of CAVs and approximately optimize traffic performance in terms of fuel efficiency and driving comfort. The elegant theoretical properties for the general objective function and the associated constraints provides an efficient analytical algorithm for solving this problem to the exact optimum. Numerical examples reveal that this exact algorithm has an efficient computational performance and a satisfactory solution quality. This trajectory-based traffic smoothing concept is then extended to develop a joint trajectory and signal optimization problem. This problem simultaneously solves the optimal CAV trajectory function shape and the signal timing plan to minimize travel time delay and fuel consumption. The proposed algorithm simplifies the vehicle trajectory and fuel consumption functions that leads to an efficient optimization model that provides exact solutions. Numerical experiments reveal that this algorithm is applicable to any signalized crossing points including intersections and work-zones. Further, the model is tested with various traffic conditions and roadway geometries. These control approaches are then extended to a mixed traffic environment with HVs, connected vehicles (CVs), and CAVs by proposing a CAV-based speed harmonization algorithm. This algorithm develops an innovative traffic prediction model to estimate the real-time status of downstream traffic using traffic sensor data and information provided by CVs and CAVs. With this prediction, the algorithm controls the upstream CAVs so that they smoothly hedge against the backward deceleration waves and gradually merge into the downstream traffic with a reasonable speed. This model addresses the full spectrum of CV and CAV market penetration rates and various traffic conditions. Numerical experiments are performed to assess the algorithm performance with different traffic conditions and CV and CAV market penetration rates. The results show significant improvements in damping traffic oscillations and reducing fuel consumption.

La prédiction de trajectoires d'avions à partir des données disponibles au sol est un problème critique pour le contrôle aérien. Une prédiction fiable et efficace est un prérequis pour l'implémentation d'outils automatiques pour la détection et la résolution de conflits entre les trajectoires. Dans ce contexte, nous proposons de nouvelles méthodes non paramétriques pour la prédiction d'intervalle contenant une proportion attendue des données avec un haut niveau de confiance. Dans un premier temps, nous traitons le problème de l'estimation d'une distribution de probabilité à partir d'un petit échantillon. En considérant l'interprétation des distributions de possibilité comme une famille de distributions de probabilité, nous décrivons un ensemble de distributions de possibilité qui résument différents types d'intervalles statistiques. Ensuite, nous proposons un cadre de travail pour vérifier si un modèle, construit à partir de données, respecte les propriétés de recouvrement requises par les intervalles de prédiction. Nous introduisons aussi deux mesures pour comparer des modèles de prédiction d'intervalle qui ont des tailles moyennes et des taux de recouvrement différents. A partir de nos travaux sur les intervalles statistiques (et leurs distributions de possibilité associés), nous présentons une nouvelle méthode pour induire des intervalles de prédictions bornés pour des méthodes de régression des moindres carrés non paramétriques sans assumer que la prédiction est non biaisée et que les erreurs sont homoscédastiques. Nos intervalles de prédiction sont construits en utilisant des intervalles de tolérances sur les erreurs dans le voisinage du point à prédire. Pour cela, nous décrivons une méthode de sélection de voisinage à taille fixe ou de voisinage à taille variable dépendant de la quantité d'informations autour du point. Nous obtenons un algorithme qui induit, dans la majorité des cas, les intervalles de prédiction fiables les plus petits possibles. Les méthodes que nous proposons sont comparées avec les méthodes les plus connues au niveau théorique et au niveau pratique. Une évaluation est effectuée sur neuf bases de données. La taille, l'efficacité, la fiabilité et la précision des intervalles prédits sont comparés. Ces expérimentations montrent que nos approches sont significativement plus précises et fiables que les autres. Enfin nous appliquons nos méthodes au problème de la prédiction de trajectoires d'avions et nous comparons les résultats avec ceux des méthodes classiques et des modèles physiques.

L'organisme Eurocontrol prévoit une forte hausse du trafic aérien européen d'ici l'année 2035. Cette hausse de trafic justifie le développement de nouveaux concepts et outils pour pouvoir assurer les services dus aux usagers de l'espace aérien. La prévision de trajectoires d'avion est au coeur de ces évolutions. Parmi ces outils, les outils de détection et résolution de conflits utilisent les trajectoires prédites pour anticiper les pertes de séparation entre avions et proposer des solutions aux contrôleurs aériens. L'horizon de prédiction utilisé pour cette application est de l'ordre de dix à vingt minutes. Parmi les algorithmes réalisant une détection et résolution de conflits, certains sont mis en œuvre au sol, obligeant ainsi les prédictions à être calculées en n'utilisant que les informations disponibles dans les systèmes sols. Dans ces systèmes, la masse des avions ainsi que les profils de vitesse ou de poussée des moteurs ne sont pas connus. Ainsi, le calcul d'une trajectoire prédite avec un modèle physique se fait en utilisant des valeurs de référence pour les paramètres inconnus. Dans ce cadre, nous nous intéressons à la phase de montée pour laquelle ces paramètres influent grandement sur la trajectoire de l'avion. Ce travail s'appuie sur le modèle physique BADA développé et maintenu par Eurocontrol. Ce modèle physique modélise, entre autres, les performances des avions. Il fournit également des valeurs de référence pour les paramètres inconnus comme la masse de l'avion, son profil de vitesse en montée, ou la commande de poussée des moteurs. Ce modèle, largement utilisé dans le monde entier, est particulièrement imprécis pour la phase de montée, car les valeurs réelles de ces paramètres sont parfois très éloignées des valeurs de référence. Dans cette thèse, nous proposons soit d'estimer directement certains paramètres, comme la masse, à partir des points passés de la trajectoire, soit d'utiliser des méthodes d'apprentissage supervisé afin d'apprendre, à partir d'exemples, des modèles prédisant les valeurs des paramètres manquants (masse, loi de poussée, vitesses cibles). Ces différentes méthodes sont testées sur des données radar Mode-C et Mode-S sur plusieurs types d'avions. Les prédictions obtenues avec ces méthodes sont comparées à celles obtenues avec les paramètres de référence. Elles sont également comparées avec les prédictions obtenues par des méthodes de régression prédisant directement l'altitude de l'avion plutôt que les paramètres du modèle physique. Nos méthodes permettent de réduire, suivant le type de l'avion, de 50 % à 85 % par rapport à la méthode BADA de référence, la racine de l'erreur quadratique moyenne sur l'altitude prédite à un horizon de dix minutes. / The Eurocontrol organization forecasts a strong increase of the European air traffic till the year 2035. This growth justifies the development of new concepts and tools in order to ensure services to airspace users. Trajectory prediction is at the core of these developments. Among these tools, conflict detection and resolution tools use trajectory predictions to anticipate losses of separation between aircraft and propose solutions to air traffic controllers. For such applications, the time horizon of the prediction is about ten to twenty minutes. Among conflict detection and resolution algorithms, some are operated in ground-based systems. The trajectory predictions must then be computed using only the information that is available to ground systems. In these systems, the mass, the speed profile and the thrust setting are unknown. Thus, using a physical model, the trajectory predictions are computed using reference values for unknown parameters. In this context, we are focusing on the climb phase. In this phase these unknown parameters have a great influence on the aircraft trajectory. This work relies on a physical model of the aircraft performances : BADA, developed and maintained by Eurocontrol. It also provides reference values for unknown parameters such as the mass, the speed profile and the thrust setting. This widely used model is particularly inaccurate for the climb phase as the actual values for the unknown parameters might be very different from the reference values. In this thesis, we propose to estimate directly the mass, an unknown parameter, using a physical model and past points of the trajectory. We also use supervised learning methods in order to learn, from examples, some models predicting the unknown parameters (mass, speed profile and thrust setting). These different approaches are tested using Mode-C Radar data and Mode-S Radar data with different aircraft types. The obtained predictions are compared with the ones obtained with the BADA reference values. These predictions are also compared with predictions obtained by directly predicting the future altitude instead of the unknown parameters of the physical model. These methods, depending on the aircraft type, reduces the root mean square error on the predicted altitude at a 10 min horizon by 50 % to 85 % when compared to the root mean square error obtained using BADA with the reference values.

<p>The motivation for this research stems from the promise of coupling multi-sensory systems and advanced data analytics to enhance holistic situational awareness and thus prevent fatal accidents in the construction industry. The construction industry is one of the most dangerous industries in the U.S. and worldwide. Occupational Safety and Health Administration (OSHA) reports that the construction sector employs only 5% of the U.S. workforce, but accounts for 21.1% (1,008 deaths) of the total worker fatalities in 2018. The struck-by accident is one of the leading causes and it alone led to 804 fatalities between 2011 and 2015. A critical contributing factor to struck-by accidents is the lack of holistic situational awareness, attributed to the complex and dynamic nature of the construction environment. In the context of construction site safety, situational awareness consists of three progressive levels: perception – to perceive the status of construction entities on the jobsites, comprehension – to understand the ongoing construction activities and interactions among entities, and projection – to predict the future status of entities on the dynamic jobsites. In this dissertation, holistic situational awareness refers to the achievement at all three levels. It is critical because with the absence of holistic situational awareness, construction workers may not be able to correctly recognize the potential hazards and predict the severe consequences, either of which will pose workers in great danger and may result in construction accidents. While existing studies have been successful, at least partially, in improving the perception of real-time states on construction sites such as locations and movements of jobsite entities, they overlook the capability of understanding the jobsite context and predicting entity behavior (i.e., movement) to develop the holistic situational awareness. This presents a missed opportunity to eliminate construction accidents and save hundreds of lives every year. Therefore, there is a critical need for developing holistic situational awareness of the complex and dynamic construction sites by accurately perceiving states of individual entities, understanding the jobsite contexts, and predicting entity movements.<br></p><p>The overarching goal of this research is to minimize the risk of struck-by accidents on construction jobsite by enhancing the holistic situational awareness of the unstructured and dynamic construction environment through a novel data-driven approach. Towards that end, three fundamental knowledge gaps/challenges have been identified and each of them is addressed in a specific objective in this research.<br></p> <p>The first knowledge gap is the lack of methods in fusing heterogeneous data from multimodal sensors to accurately perceive the dynamic states of construction entities. The congested and dynamic nature of construction sites has posed great challenges such as signal interference and line of sight occlusion to a single mode of sensor that is bounded by its own limitation in perceiving the site dynamics. The research hypothesis is that combining data of multimodal sensors that serve as mutual complementation achieves improved accuracy in perceiving dynamic states of construction entities. This research proposes a hybrid framework that leverages vision-based localization and radio-based identification for robust 3D tracking of multiple construction workers. It treats vision-based tracking as the main source to obtain object trajectory and radio-based tracking as a supplementary source for reliable identity information. It was found that fusing visual and radio data increases the overall accuracy from 88% and 87% to 95% and 90% in two experiments respectively for 3D tracking of multiple construction workers, and is more robust with the capability to recover the same entity ID after fragmentation compared to using vision-based approach alone.<br></p> <p>The second knowledge gap is the missing link between entity interaction patterns and diverse activities on the jobsite. With multiple construction workers and equipment co-exist and interact on the jobsite to conduct various activities, it is extremely difficult to automatically recognize ongoing activities only considering the spatial relationship between entities using pre-defined rules, as what has been done in most existing studies. The research hypothesis is that incorporating additional features such as attentional cues better represents entity interactions and advanced deep learning techniques automates the learning of the complex interaction patterns underlying diverse activities. This research proposes a two-step long short-term memory (LSTM) approach to integrate the positional and attentional cues to identify working groups and recognize corresponding group activities. A series of positional and attentional cues are modeled to represent the interactions among entities, and the LSTM network is designed to (1) classify whether two entities belong to the same group, and (2) recognize the activities they are involved in. It was found that by leveraging both positional and attentional cues, the accuracy increases from 85% to 95% compared with cases using positional cues alone. Moreover, dividing the group activity recognition task into a two-step cascading process improves the precision and recall rates of specific activities by about 3%-12% compared to simply conducting a one-step activity recognition.<br></p> <p>The third knowledge gap is the non-determining role of jobsite context on entity movements. Worker behavior on a construction site is goal-based and purposeful, motivated and influenced by the jobsite context including their involved activities and the status of other entities. Construction workers constantly adjust their movements in the unstructured and dynamic workspace, making it challenging to reliably predict worker trajectory only considering their previous movement patterns. The research hypothesis is that combining the movement patterns of the target entity with the jobsite context more accurately predicts the trajectory of the entity. This research proposes a context-augmented LSTM method, which incorporates both individual movement and workplace contextual information, for better trajectory prediction. Contextual information regarding movements of neighboring entities, working group information, and potential destination information is concatenated with movements of the target entity and fed into an LSTM network with an encoder-decoder architecture to predict trajectory over multiple time steps. It was found that integrating contextual information with target movement information can result in a smaller final displacement error compared to that obtained only considering the previous movement, especially when the length of prediction is longer than the length of observation. Insights are also provided on the selection of appropriate methods.<br></p><p>The results and findings of this dissertation will augment the holistic situational awareness of site entities in an automatic way and enable them to have a better understanding of the ongoing jobsite context and a more accurate prediction of future states, which in turn allows the proactive detection of any potential collisions.<br></p>

<div>With the growing demand of air traffic, it becomes more important and critical than ever to develop advanced techniques to control and monitor air traffic in terms of safety and efficiency. Especially, trajectory prediction can play a significant role on the improvement of the safety and efficiency because predicted trajectory information is used for air traffic management such as conflict detection and resolution, sequencing and scheduling. </div><div><div>In this work, we propose a new framework by integrating</div><div>the two methods, called hybrid data-driven and physics-based trajectory prediction. The proposed algorithm is applied to real air traffic surveillance data to demonstrate its performance.</div></div>

No description available.

Metamodeling is regarded as a powerful analysis tool to learn the input-output relationship of a system based on a limited amount of data collected when experiments with real systems are costly or impractical. As a popular metamodeling method, Gaussian process regression (GPR), has been successfully applied to analyses of various engineering systems. However, GPR-based metamodeling for time-dependent systems (TDSs) is especially challenging due to three reasons. First, TDSs require an appropriate account for temporal effects, however, standard GPR cannot address temporal effects easily and satisfactorily. Second, TDSs typically require analytics tools with a sufficiently high computational efficiency to support online decision making, but standard GPR may not be adequate for real-time implementation. Lastly, reliable uncertainty quantification is a key to success for operational planning of TDSs in real world, however, research on how to construct adequate error bounds for GPR-based metamodeling is sparse. Inspired by the challenges encountered in GPR-based analyses of two representative stochastic TDSs, i.e., load forecasting in a power system and trajectory prediction for unmanned aerial vehicles (UAVs), this dissertation aims to develop novel modeling, sampling, and statistical analysis techniques for enhancing the computational and statistical efficiencies of GPR-based metamodeling to meet the requirements of practical implementations. Furthermore, an in-depth investigation on building uniform error bounds for stochastic kriging is conducted, which sets up a foundation for developing robust GPR-based metamodeling techniques for analyses of TDSs under the impact of strong heteroscedasticity. / Ph.D. / Metamodeling has been regarded as a powerful analysis tool to learn the input-output relationship of an engineering system with a limited amount of experimental data available. As a popular metamodeling method, Gaussian process regression (GPR) has been widely applied to analyses of various engineering systems whose input-output relationships do not depend on time. However, GPR-based metamodeling for time-dependent systems (TDSs), whose input-output relationships depend on time, is especially challenging due to three reasons. First, standard GPR cannot properly address temporal effects for TDSs. Second, standard GPR is typically not computationally efficient enough for real-time implementations in TDSs. Lastly, research on how to adequately quantify the uncertainty associated with the performance of GPR-based metamodeling is sparse. To fill this knowledge gap, this dissertation aims to develop novel modeling, sampling, and statistical analysis techniques for enhancing standard GPR to meet the requirements of practical implementations for TDSs. Effective solutions are provided to address the challenges encountered in GPR-based analyses of two representative stochastic TDSs, i.e., load forecasting in a power system and trajectory prediction for unmanned aerial vehicles (UAVs). Furthermore, an in-depth investigation on quantifying the uncertainty associated with the performance of stochastic kriging (a variant of standard GPR) is conducted, which sets up a foundation for developing robust GPR-based metamodeling techniques for analyses of more complex TDSs.