FAQ
Open Access Blog
Policies
Introduction
Texas ScholarWorks was established to provide open, online access to the products of the University's research and scholarship, to preserve these works for future generations, to promote new models of scholarly communication, and to help deepen community understanding of the value of higher education.
UT Tower and campus image credit: Earl McGehee, CC-BY, https://www.flickr.com/photos/ejmc/7452145850
Communities in TSW
Select a community to browse its collections.
Recent Submissions
"Fractional" Vocational Working and Learning in Project Teams: "Project Assemblage" as a Unit of Analysis?
(Springer Link, 2023-11-03) Spinuzzi, Clay; Guile, David
Situated and Activity theories have exercised a significant influence in the field of vocational learning for some considerable time, both sharing a focus on bounded forms of work and forms of learning that facilitate learning in, or to changes to, bounded forms of work. Yet much learning occurs in unbounded contexts often referred to as projectification, where collaborations occur only for the life of a project thereby creating new contingent contexts for learning . Given the existence of this form of working and learning, what type of unit of analysis (UoA) is required to analyse that vocational working and learning in the context of projectification? To address this question, the paper advances the following inter-theoretical argument. Firstly, it is timely to develop a new unit of analysis (UoA) to capture the fractional (intermittent, discontinuous and concurrent) working and learning dynamics associated with the forms of projectification, where funding has to be procured in order to commence. Secondly, that unit of analysis is constituted by the concept of project assemblage, which is based on ideas from Actor Network Theory, Cultural-historical Activity Theory and Cultural Sociology. Thirdly, this new UoA enables researchers to identify the way in which project teams, where members are coming in-and-out, learn to use their different forms of specialist activity to enact objects, why team members will have different backgrounds and understandings of their work, why objects may not cohere, even though team members may treat them as unified and coherent, and how team members learn to incorporate one another’s insights and suggestions, and establish a finalized object.
Impact of Power Outages Depends on Who Loses It: Equity-Informed Grid Resilience Planning
(2023-10-06) Hasenbein, John J.; Kutanoglu, Erhan; Toplu-Tutay, Gizem
This research presents a novel approach for enhancing power grid resilience with a focus on social
equity in light of increasing natural disasters. We recognize that natural disasters such as hurricanes
and floods can disproportionately affect disadvantaged communities, exacerbating existing disparities.
Our research aims to bridge this gap by integrating tailored equity metrics into resilience
planning.
Our methodology utilizes a two-stage stochastic optimization model for hurricane-induced flood
mitigation, which optimizes substation hardening and power flow decisions. The goal of this model
is to minimize both the expected load shed and the expected well-being loss metrics of socially
vulnerable communities (affected population and duration of loss) in the aftermath of flooding. We
explore the trade-off between these objectives.
What sets our research apart is the integration of realistic flood scenario generation, a large-scale
synthetic power grid of Texas, and multiple methodologies in resource allocation, community impact
modeling, power flow modeling, and equity metric development, as well as comprehensive
computational experiments.
The findings highlight the importance of the composite objective function in altering power flow
decisions to prioritize electricity provision and save communities in disadvantaged areas even without
investing in substation hardening (i.e., just managing load shedding with more attention to such
vulnerabilities).
The results also quantify the equity and load shed benefits of substation hardening as a function of
the investment budget with a parameterized analysis. With an attention to equity, power outages
increase in nonvulnerable communities — a trade-off made to mitigate well-being loss in the most
vulnerable areas. Notably, more attention to equity provides a lower or equal number of people saved
per 1 MW increase in the load shed, underscoring the concept of diminishing returns. Our findings
highlight the importance of strategically allocating a limited budget and consistently prioritizing
the hardening of substations serving more vulnerable populations.
We further explore a justice model inspired by the government’s Justice40 initiative but find it less
effective than our equity-informed models at preventing well-being loss.
Our findings offer valuable insights for policymakers, grid operators, and utilities striving for a more
resilient and equitable power grid. We believe that our research will not only contribute to equitable
power grid resilience but also provide practical solutions to address the pressing challenges posed
by climate change and natural disasters.
Robust machine learning against unseen distributional shifts
(2023-06-20) Wang, Haotao; Wang, Zhangyang; Bovik, Alan C; Kim, Hyeji; Liu, Qiang; Zhou, Jiayu
The performance decay experienced by deep neural networks (DNNs) when confronted with distributional shifts poses a significant challenge for their deployment in real-world scenarios. The limited out-of-distribution (OOD) generalization ability of DNNs acts as a critical bottleneck, especially considering the vast diversity and unpredictable changes inherent in the real-world test data distribution. To tackle this challenge, robust training emerges as a highly effective strategy for enhancing the OOD generalization ability of DNNs, enabling them to excel in complex and unpredictable real-world environments. This dissertation revolves around my extensive research on robust training, encompassing three fundamental aspects: robust data augmentation, robust model architecture design, and robust learning algorithms. Within the realm of robust data augmentation, my research delves beyond traditional augmentation methods. I explore innovative techniques that jointly generates diverse and hard training samples. Such data augmentation strategy enhances the robustness of DNNs, equipping them with the ability to generalize effectively even in the face of unforeseen changes in the data distribution. In terms of robust model architecture design, I investigate approaches that empower DNNs with inherent resilience to distributional shifts. Specifically, I design more robust normalization layers for deep neural networks, and utilize model ensembles to increase the models' capability to encode diverse training samples and generalize to unseen test cases. Moreover, I investigate advanced training algorithms that explicitly emphasize robustness. These algorithms involve leveraging adversarial training paradigms to expose and mitigate vulnerabilities, developing novel loss functions such as contrastive loss that prioritize performance on challenging or out-of-distribution samples, and robust regularization terms to prevent the models from overfitting to non-robust features. Most importantly, I show these robust learning strategies are not mutually independent with each other. Instead, it is crucial to adapt them with each other to achieve the best performance gain.
Climate impacts and food systems planning : Austin Farms’ resilience to heat and drought
(2023-08-15) Patracuolla, Gabrielle Rose; Lieberknecht, Katherine E.; Paterson, Robert G.
The innate human connection to nature is crucial to understanding how people interact with their food systems in the urban context. Increasing heat and drought will continue to affect the production and growth of agriculture in Austin, Texas as well as people’s ability to connect to the food system. Farmers in Austin are worried about increased impacts on production, distribution and income. Future projections show increases in heat and drought causing concern for the future of food production in the region. The City of Austin and other entities can help to create stability in an increasingly unstable climate.
Information-theoretic path planning and navigation
(2023-06-12) Pedram, Ali Reza; Tanaka, Takashi, Ph. D.; Sentis, Luis; Beaman, Joseph; Chen, Dongmei; Alambeigi, Farshid; Bakolas, Efstathios
Thanks to the wide availability of low-cost and high-performance sensing devices, obtaining a large amount of sensor data has become easier for robots in many applications. Nevertheless, operating a sensor at full capacity may not be the best strategy for resource-constrained robots, especially if it drains the robot’s scarce power or computational resources with little benefit. As sensor modalities increase, how to achieve a given task with minimum perceptual resources (e.g., with reduced sensing frequencies or sensor gains) becomes an increasingly relevant question. Specially in navigation tasks, it is crucial for autonomous agents to find the path plans which require low perception to be followed. This dissertation proposes a two-stage paradigm to reduce the perception cost in autonomous navigation. In the first stage, referred to as the offline path planning stage, a reference path plan requiring moderate navigation perception is sought in uncertain configuration space (the product space of mean and covariance). In the second stage, the online path following stage, the sensor modalities and closed-loop controller are deployed to closely follow the reference path. In the first phase, we introduce a novel path length function accounting for both travel and the expected perception costs. The continuity of the path length function with respect to the topology of the total variation metric is shown. We formulate the path planning problem as a shortest path problem with respect to the introduced cost conditioned by a collision avoidance constraint. We present sampling-based planning algorithms to solve the formulated shortest path problem and propose several improvements to increase their computational efficiency. Leveraging the continuity of the path length, we prove the proposed algorithms find the optimal path almost surely (with probability one) in the limit of a large number of samples. The first phase ends with a smoothing algorithm developed to eliminate the sharp turns in the paths sought by the sampling-based algorithms. For path smoothing, we devise novel collision avoidance constraints that guarantee safety in continuous-time (in transition between states). This dissertation formulates the path following stage as a joint control-sensing problem and develops an attention mechanism to follow the reference path with minimum perception cost (required bits of information). Several numerical simulations are provided to showcase the utility of the proposed algorithms and paradigm in reducing the perception costs, like the frequency of sensor measurements or the number of sensors that must be used simultaneously in navigation.