Search results
(1 - 2 of 2)
- Title
- DEEP LEARNING IN ENGINEERING MECHANICS: WAVE PROPAGATION AND DYNAMICS IMPLEMENTATIONS
- Creator
- Finol Berrueta, David
- Date
- 2019
- Description
-
With the advent of Artificial Intelligence research in the 1960s, the need for intelligent systems that are able to truly comprehend the...
Show moreWith the advent of Artificial Intelligence research in the 1960s, the need for intelligent systems that are able to truly comprehend the physical world around them became relevant. Significant milestones in the realm of machine learning and, in particular, deep learning during the past decade have led to advanced data-driven models that are able to approximate complex functions from pure observations. When it comes to the application of physics-based scenarios, the vast majority of these models rely on statistical and optimization constructs, leaving minimal room in their development for the physics-driven frameworks that more traditional engineering and science fields have been developing for centuries. On the other hand, the more traditional engineering fields, such as mechanics, have evolved on a different set of modeling tools that are mostly based on physics driven assumptions and equations, typically aided by statistical tools for uncertainty handling. Deep learning models can provide significant implementation advantages in commercial systems over traditional engineering modeling tools in the current economies of scale, but they tend to lack the strong reliability their counterparts naturally allow. The work presented in this thesis is aimed at assessing the potential of deep learning tools, such as Convolutional Neural Networks and Long Short-Term Memory Networks, as data-driven models in engineering mechanics, with a major focus on vibration problems. In particular, two implementation cases are presented: a data driven surrogate model to a Phononic eigenvalue problem, and a physics-learning model in rigid-body dynamics scenario. Through the applications presented, this work that shows select deep learning architectures can appropriately approximate complex functions found in engineering mechanics from a system’s time history or state and generalize to set expectations outside training domains. In spatio-temporal systems, it is also that shown local learning windows along space and time can provide improved model reliability in their approximation and generalization performance
Show less
- Title
- Computationally Efficient Predictive Control Strategies for Autonomous Vehicles
- Creator
- Bhattacharyya, Viranjan
- Date
- 2021
- Description
-
This thesis aims at developing computationally efficient (hence real-time applicable) control strategies for autonomous vehicles in the...
Show moreThis thesis aims at developing computationally efficient (hence real-time applicable) control strategies for autonomous vehicles in the presence of uncertainty, while incorporating high fidelity vehicle dynamics. The motivation for the control strategies is to ensure safety and improve energy efficiency of the vehicles. In this research, an effort has been made to develop control strategies to strike a balance between these competing factors. The specific contributions are: development of a new hierarchical control framework that can guarantee avoidance of red-light idling in the presence of uncertainty in preceding vehicle information/prediction in connected environment (hence improves system mobility); exploitation of a data-driven modeling approach for identifying a linear predictor for the nonlinear vehicle dynamics, which facilitates formulation of a convex equivalent problem of the original non-convex problem (hence facilitates computational tractability); introduction of a novel vehicle dynamics-aware fast game-theoretic planner for behavior and motion planning of vehicles in uncertain and unconnected environments. This thesis explores both the possible directions of future autonomous vehicles: connected and unconnected autonomous vehicles. In particular, the first problem relates to longitudinal fuel efficient driving (eco-driving) in a connected urban environment, where the connected and automated vehicles (CAVs) aim at the improvement of fuel efficiency and reduction of red-light idling (stop and go motion). The CAVs also focus on ensuring collision avoidance with the preceding vehicles despite the prediction uncertainty in future trajectory of preceding vehicles. This problem assumes vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communication, and is a longitudinal control problem. The next problem considers the uncertainty in prediction of future states of neighbouring vehicles in an unconnected environment and involves both lateral and longitudinal control. Following previous research, the interactive nature of driving is modeled using game-theory and a computationally efficient game-theoretic planner is introduced. Simulation results show the efficacy of the proposed methods in terms of computational tractability and fuel-efficiency.
Show less