Theses and Dissertations
Permanent URI for this collection
Browse
Recent Submissions
1 - 5 of 24051
-
ItemDifferentiable and tolerant barrier states for improved exploration of safety-embedded differential dynamic programming with chance constraints(Georgia Institute of Technology, 2023-05-10)A great challenge exists at the intersection of perception and controls – integrating the uncertainty present in perception-based state and obstacle estimation into safe control and trajectory optimization. First, we present the tolerant discrete barrier state (T-DBaS), a novel safety-embedding technique for trajectory optimization with enhanced exploratory capabilities. This approach generalizes the standard discrete barrier state (DBaS) method by accommodating temporary constraint violation during the optimization process while still approximating its safety guarantees. Towards applying T-DBaS to safety-critical au- tonomous robotics, we combine it with Differential Dynamic Programming (DDP), leading to the proposed safe trajectory optimization method T-DBaS-DDP, which inherits the con- vergence and scalability properties of the solver. Despite this, the tolerant barrier function parameters require tuning to reach peak performance for a wide array of constraints. To alleviate this requirement, we tune the T-DBaS parameters with the parameterized trajec- tory optimizer Pontryagin Differentiable Programming (PDP), proposing T-DBaS-PDP, an interpretable and generalizable solver for a variety of optimal control problems. In order to integrate perception uncertainty into safe optimal control, we learn the safety of the sys- tem via gaussian processes to create an interpretable, data-driven, and safety-guaranteeable framework. We implement this framework on differential drive and quadrotor dynamics and show its improvement over hand-tuned T-DBaS-DDP.
-
ItemDigital holography for exploring instabilities and breakup of liquid jets in supersonic crossflows(Georgia Institute of Technology, 2023-05-04)Direct injection studies of liquid jets in supersonic crossflows (JICFs) are critical for understanding combustion in scramjet engines. Exploring these fluid dynamic interactions is not only an important step towards characterizing fundamental liquid breakup properties but also key for improving engine design and increasing efficiency. Current engine designs lack precise injector optimization and, therefore deliver inefficient fuel sprays. To remedy this, previous studies in the literature have examined how supersonic crossflows affect gaseous and liquid jet breakup characteristics using backlit imaging or schlieren techniques. In this work, we aim to study jet instabilities and droplet breakup characteristics in JICFs for the first time using digital in-line holography techniques. Experiments are conducted in a heated Mach 1.71 crossflow with a transitional regime liquid jet (slenderness ratio L/D of 19) with a diameter of 0.5 mm. High-speed and high-resolution digital in-line holography techniques are utilized to spatially resolve the jet breakup characteristics near the injection point. Results show that the front-edge instability wavelength spacing ranges from 68.3 to 104.5 microns, decreasing as the injected liquid pressure increases from 100 to 500 psi. These results show an inverse relationship between these instabilities and the injected pressure. Both windward and leeward droplet velocities and sizes are also measured using digital holography and analyzed to determine trends. Findings show a clear relationship between the liquid jet injection pressure and the velocity profile of the droplets on the windward side of the jet in the streamwise direction. Droplet size distributions showed small droplet diameters ranging from 3.8 to 25 microns. The unique experimental results acquired in this work can be used to understand entrainment effects, improve mathematical multiphase flow breakup models, optimize injector geometry, and refine future scramjet engine designs.
-
ItemNavigating to Objects: Simulation, Data, and Models(Georgia Institute of Technology, 2023-05-03)General-purpose robots that can perform a diverse set of embodied tasks in a diverse set of environments have to be good at visual exploration. Consider the canonical example of asking a household robot, ‘Where are my keys?’. To answer this (assuming the robot does not remember the answer from memory), the robot would have to search the house, often guided by intelligent priors – e.g. peeking into the washroom or kitchen might be sufficient to be reasonably sure the keys are not there, while exhaustively searching the living room might be much more important since keys are more likely to be there. While doing so, the robot has to internally keep track of where all it has been to avoid redundant search, and it might also have to interact with objects, e.g. check drawers and cabinets in the living room (but not those in the washroom or kitchen!). This example illustrates fairly sophisticated exploration, involving a careful interplay of various implicit objectives (semantic priors, exhaustive search, efficient navigation, interaction, etc.) which are hard to learn using Reinforcement Learning (RL). In this thesis, we focus on learning such embodied object-search strategies from human demonstrations which implicitly captures intelligent behavior we wish to impart to our agents. In Part I, we present a large-scale study of imitating human demonstrations on tasks that require a virtual robot to search for objects in new environments – (1) ObjectGoal Navigation (e.g. ‘find & go to a chair’) and (2) PICK&PLACE (e.g. ‘find mug, pick mug, find counter, place mug on counter’). In Part 2, we extend our focus to improving agents trained using human demonstrations in a tractable way. Towards this, we present PIRLNav, a two-stage learning scheme for BC pretraining on human demonstrations followed by RL-finetuning. Finally, using this BC→RL training recipe, we present a rigorous empirical analysis where we investigate whether human demonstrations can be replaced with ‘free’ (automatically generated) sources of demonstrations, e.g. shortest paths (SP) or task-agnostic frontier exploration (FE) trajectories.
-
ItemMicrostructural Prediction of Additively Manufactured Multi-Phase Materials(Georgia Institute of Technology, 2023-05-03)Material modeling is the central theme of this thesis. Experimentation of titanium alloys and composites provided background knowledge of the time and financial costs associated with testing and re-testing properties in a forensic, trial-and-error manner. The model discussed in this thesis merges material properties and process parameters to generate a unique microstructure for the titanium 6Al-4V (Ti-6-4) alloy produced using selective laser melting additive manufacturing SLM-AM. The material texture is generated by first calculating melt pool geometries using Rosenthal Solution equations and Bunge matrix transformations. The result is a single-phase representation of a liquidus, BCC beta titanium phase deposited over a random-orientation substrate. The texture is then transformed into a two-phase alpha (HCP)-beta microstructure through transformation pathways modeled based on mechanisms discovered in other studies. The final texture product can then be input into other models capable of computing mechanical properties based on texture inputs. Though no model can be fully comprehensive in simulating material nature and behavior, the model in this thesis adapts enough experimental data and follows enough phenomenological observations within the field of material science and engineering to produce simulated samples capable of achieving realistic property values. The model is scripted in a manner where it can be adapted for alternative materials by inputting different properties and tailoring the process settings. Multiple benefits arise from being able to model material microstructures without the need to physically test real-world samples. There is a substantial time savings in being able to quickly adjust properties and formulations. Expensive equipment, materials, and labor can all be avoided, and a larger testing matrix can be executed through this approach.
-
ItemEssays on Strategic Use of Intellectual Property Rights(Georgia Institute of Technology, 2023-05-02)This dissertation examines various aspects of firms' strategic responses in the management of their intellectual property rights to external factors, with each chapter focusing on a different aspect of this complex topic. The first chapter of the dissertation investigates whether markets for technology can provide an alternative to in-house innovation as a response to foreign escape competition. The findings indicate that while external technologies are important for innovative firms, low-productivity firms experience a negative impact on the demand for external technologies when exposed to import competition. The second chapter focuses on the tradeoff incumbents face in using patents as barriers to entry or as ex-post responses to competitors' entry moves. Using the U.S. pharmaceutical industry as the main empirical setting, this chapter shows that incumbents intentionally fragment and delay the full disclosure of their intellectual property rights through continuation patents. They disproportionately reveal continuation patents after a competitor entry threat becomes concrete, tailoring their response to the threat they have received and successfully delaying competitor entry through litigation. Finally, the third chapter investigates how firms manage information asymmetry in their patent prosecution source using exposure to patent litigation. This chapter shows that firms exposed to patent litigation are more likely to change the sourcing of patent prosecution legal services relative to unexposed firms working with the same prosecuting law firm.