About

Shirley Dyke is the Donald A. and Patricia A. Coates Professor of Innovation in Mechanical Engineering and Professor of Civil and Construction Engineering at Purdue University. Her research focuses on innovation within engineering disciplines, contributing to advancements in mechanical and civil engineering fields. As a distinguished faculty member, she is involved in teaching, research, and leadership activities that promote engineering innovation and excellence.

Research topics

• Computer Science
• Artificial Intelligence
• Engineering
• Composite material
• Physics
• Statistical physics
• Electrical engineering
• Mathematics
• Applied mathematics
• Materials science
• Simulation
• Control engineering

Selected publications

• Reinforcement Learning-Based Framework to Support Multi-Agent Teaming in Space Missions

  2026-01-08

  articleSenior author

  Future space missions will increasingly depend on effective multi-agent teaming, particularly human–robot collaboration, to sustain autonomous operations in deep-space environments. Addressing the urgent need for adaptive and data-driven decision capabilities, this paper presents a reinforcement learning–based framework for supporting multi-agent teaming by generating teaming policies that determine how human and robotic agents coordinate to complete complex mission tasks. The framework models each task as a structure of precedence-constrained subtasks with stochastic execution outcomes and supports both individual and collaborative action modes under uncertainty. To validate its capability, two representative maintenance and repair (M&R) case studies are examined: a linear structural-repair task and a mission-realistic life-support maintenance scenario. Results show that the learned policies are precedence-aware, allocate roles across agents, and efficiently balance collaboration and parallelization to minimize redundant actions and improve task reliability. These outcomes demonstrate that the proposed framework effectively captures the temporal, structural, and probabilistic dynamics of human–robot teaming for future space mission operations.

  PublisherDOI

• Investigating Large Language Model-Based Decision Making for Deep Space Habitat Systems

  2026-03-07

  articleSenior author

  Deep space habitats are complex systems with tightly coupled interdependencies exposed to hazardous environments. Effective decision-making in these systems requires models that can scale with multiple fault scenarios while incorporating system-level expertise. Due to increased communication delays and bandwidth constraints, conventional ground control strategies and expert interventions may become less practical, highlighting the need for autonomous onboard decision-making capabilities. In this work, we investigate the feasibility of using Large Language Models (LLMs) for autonomous decision support within the Human-centered Autonomous Resilient Space Habitat (HARSH), a cyber-physical testbed at the Resilient Extra-Terrestrial Habitats Institute (RETHi). As a pilot study, we replace existing heuristics and priority rules in HARSH's health management system (HMS) with LLM-generated decisions. The LLM determines agent actions based on structured outputs from the diagnostic reasoner, guided by prompt engineering and a predefined action set. Structured outputs and function-calling features of OpenAI's API are employed to obtain and execute LLM responses. To quantify uncertainty in LLM decision making, we perform a statistical study using ten representative failure modes and three model variants: GPT-4.1-nano, GPT-4.1, and o3. Accuracy and response time are evaluated across 1,10, and 100 decision samples. Results show that GPT-4.1 achieves accuracy comparable to o3 with significantly lower response time, while GPT-4.1-nano is faster but less accurate. Synthetic scenario simulations with single- and multi-fault conditions are then conducted under two prompt versions differing in information content. Under the best prompt, LLM decisions match heuristic decisions, while under the worst prompt accuracy degrades, especially when multiple faults occur. The LLM-based decision maker is integrated into the HMS and evaluated in multi-fault experiments with and without emulated communication delays between the ground-based fault detection and diagnostics module and the on-board command and control. Resilience metrics are estimated, compared, and contrasted at both component and system levels. Finally, challenges, lessons learned, and limitations are discussed, with future work focusing on extending the information available to the LLM and testing unanticipated fault detection use cases.

  PublisherDOI

• Leveraging Lunar Testbeds to Enable Resilient and Autonomous Space Habitats

  2026-05-13

  article1st authorCorresponding

  The Resilient ExtraTerrestrial Habitats Institute (RETHi), funded through NASA’s Space Technology Mission Directorate, has been addressing the challenges of space habitation since its founding in 2019. As humanity prepares to explore the Moon, Mars, and even deeper into space, resilience, autonomy, and reusability grow in importance. The harsh and unpredictable environments on the Moon and Mars, with threats like micrometeorite impacts, seismic activity, radiation, reduced gravity, abrasive dust, and extreme thermal conditions, will require that these systems are designed to withstand such extreme conditions and respond when anomalies occur. Furthermore, these systems must be capable of operating independently, with minimal communication with ground control. To meet these challenges, RETHi centered its research around three thrusts—Resilience, Awareness, and Robotics. To assist with the research tasks, rapid prototyping solutions, and validation of new approaches, three complementary testbeds were created such that a suite of experimental and computational capabilities regarding resilient design and autonomous operation is in place. Together, the physics-based habitat simulator (HabSim), the Control-oriented Dynamic Computational Model (CDCM), and the Human-Centered Autonomous Resilient Space Habitat (HARSH) have generated significant insights on the autonomous operation and resilience of a habitat. This paper will highlight several of the main findings, as well as ideas for future utilization by academia, government, and private sectors.

  PublisherDOI

• A Cyber-Physical Testing Framework for Evaluating Space Habitat Performance

  2026-03-07

  articleSenior author
  PublisherDOI

• Reactive, Resource-Aware Scheduler for Crew Ingress in Deep Space Habitats

  2026-05-13

  article

  The recent NASA Ignition activities have reinvigorated the nation's enthusiasm for deep space exploration. However, the extended duration and traveling distances of these missions require autonomous operation and resilience to ensure crew safety during critical habitat transitions predicted in current mission scenarios. Among the most demanding of these transitions is crew ingress—the process of activating a dormant habitat for crewed operations—which requires sequencing multiple subsystems under constrained energy availability and the potential for disruptions such as micrometeorite impacts or equipment failure. In this paper, we present a reactive, resource-aware scheduling optimization algorithm for crew ingress in deep space habitats, validated through the Human-Centered Autonomous Resilient Space Habitat (HARSH) cyber-physical testbed. Unlike static preplanned schedules, the proposed scheduler dynamically adjusts action sequences in response to evolving mission conditions, employing a genetic algorithm to evaluate candidate schedules based on energy consumption and precedence feasibility, with an adaptive constraint relaxation mechanism that defers non-critical actions when energy budgets are exceeded. Experimental validation demonstrates that habitats executing fixed ingress sequences consistently fail under limited energy availability, depleting stored energy at approximately 6.48 hours into the transition following a simulated micrometeorite impact. When the scheduler is activated, essential loads are prioritized over non-critical high-energy actions, yielding an approximately 5% reduction in cumulative energy demand. The results show how adaptive scheduling may be used to advance transition planning, contributing to the safety and success of future Lunar and Martian missions.

  PublisherDOI

• Unsupervised anomaly detection based on deep autoencoders, information fusion, and active sensing

  Structural Health Monitoring · 2026-01-23 · 1 citations

  articleSenior author

  Structural health monitoring plays a crucial role in ensuring the safety and resilience of engineering structures. Detecting structural anomalies is essential for maintaining the safety of citizens and the normal operation of civil infrastructures. In this study, a novel anomaly detection framework is proposed based on deep autoencoders (DAEs), information fusion, and active sensing. The framework involves exciting the structure at specific locations and collecting acceleration data. The data collected from multiple excitation and sensor locations are analyzed and fused to enhance anomaly detection performance. More specifically, an unsupervised anomaly detection framework using DAEs has been developed. Continuous wavelet transforms (CWTs) of acceleration signals are utilized to train DAEs. Information fusion strategies are proposed to enhance the robustness of the approach to both structural uncertainty and measurement noise. A comprehensive evaluation is performed to compare the performance of fully connected AEs, convolutional AEs (CAEs), variational AEs, transformer AEs, and one-class support vector machines, each trained separately on either CWTs or raw acceleration signals, to investigate the effect of input representation on anomaly detection performance. The results show that CAEs outperform other DAE-based approaches in detecting structural anomalies, achieving higher F1 scores and lower computational costs. Additionally, training DAEs with CWTs yields better performance than using acceleration time series. Numerical studies based on the ASCE benchmark structure and experimental studies based on a geodesic dome structure have been carried out to study the capabilities as well as limitations of the proposed approach. The effects of sensor and damage locations on anomaly detection performance are analyzed through damage identifiability. Solutions are proposed for the practical cases of limited sensors and insufficient data. The framework’s ability to extract information from multiple sources allows it to identify anomalies that traditional detection methods might have missed.

  PublisherDOI

• Physics-Informed Machine Learning for Advanced StructuralDamage Detection and Localization

  2025-01-01

  articleOpen accessSenior author

  Detection and identification of nonlinearity is a task of high importance for structural dynamics.On the one hand, identifying nonlinearity in a structure would allow one to build more accurate models of the structure.On the other hand, detecting nonlinearity in a structure, which has been designed to operate in its linear region, might indicate the existence of damage within the structure.Common damage cases which cause nonlinear behaviour are breathing cracks and points where some material may have reached its plastic region.Therefore, it is important, even for safety reasons, to detect when a structure exhibits nonlinear behaviour.In the current work, a method to detect nonlinearity is proposed, based on the distribution of the gradients of a data-driven model, which is fitted on data acquired from the structure of interest.The data-driven model selected for the current application is a neural network.The selection of such a type of model was done in order to not allow the user to decide how linear or nonlinear the model shall be, but to let the training algorithm of the neural network shape the level of nonlinearity according to the training data.The neural network is trained to predict the accelerations of the structure for a time-instant using as input accelerations of previous time-instants, i.e. one-step-ahead predictions.Afterwards, the gradients of the output of the neural network with respect to its inputs are calculated.Given that the structure is linear, the distribution of the aforementioned gradients should be unimodal and quite peaked, while in the case of a structure with nonlinearities, the distribution of the gradients shall be more spread and, potentially, multimodal.To test the above assumption, data from an experimental structure are considered.The structure is tested under different scenarios, some of which are linear and some of which are nonlinear.More specifically, the nonlinearity is introduced as a column-bumper nonlinearity, aimed at simulating the effects of a breathing crack and at different levels, i.e. different values of the initial gap between the bumper and the column.Following the proposed method, the statistics of the distributions of the gradients for the different scenarios can indeed be used to identify cases where nonlinearity is present.Moreover, via the proposed method one is able to quantify the nonlinearity by observing higher values of standard deviation of the distribution of the gradients for lower values of the initial column-bumper gap, i.e. for "more nonlinear" scenarios.

  PublisherOA PDFDOI

• Computational Framework for Assessing Mission Outcomes with Humans and Robots

  AIAA Journal · 2025-03-13 · 3 citations

  article

  Space exploration is progressing toward long-term missions that involve both human (HAs) and robotic agents (RAs) in operations in lunar space habitats, the Gateway space station, and the moon-to-Mars program. These missions require high-level intelligence and a sustained performance over extended periods. Analyzing agent performance solely at the task level is insufficient for such complex applications because the resources consumed by agents are coupled with the utility they provide under various conditions. Additionally, factors such as the availability of agents to respond to hazardous events, impacted by factors including human sleep cycles and robot charging times, must be considered. Understanding how resources, utility, and availability are interrelated is crucial for early-phase decision making, assessing logistics, and steering investments in promising directions. In this study, the rapid simulation capabilities of control-oriented dynamic computational modeling (CDCM) were used to explore the trade space involving an HA and an RA tasked with maintaining a smart space habitat. This approach was used to model two independent parallel scenarios as systems of systems that use stochastic methods to account for mission variabilities. A human scientist (HS) was included to quantify the mission’s research outcomes. The outcomes generated by the HS served as a metric to compare the performance of the agents along with the costs associated with engaging the HA and RA.

  PublisherDOI

• Experimental validation of thermomechanical cyber-physical testing for the assessment of lunar habitats

  Acta Astronautica · 2025-05-14 · 1 citations

  articleSenior author
  PublisherDOI

• Assessment of the Mechanical Properties of Low-Cost Seismic Isolators Exposed to Environmental Conditions

  Applied Sciences · 2025-03-21 · 4 citations

  articleOpen access

  In Colombia, low-cost unbonded fiber-reinforced elastomeric isolators made from natural rubber (UN-FREI) and recycled rubber (UR-FREI) have emerged as a solution to mitigate damage in low-rise structures during earthquakes. However, their performance under environmental degradation caused by factors such as carbon dioxide, saltwater, relative humidity, and UV radiation has not been sufficiently studied. These agents can compromise the mechanical properties of rubber, affecting its ability to dissipate energy. This study evaluates the performance of these isolators under different environmental conditions through the initial characterization of rubber, mechanical testing of small-scale prototypes exposed to controlled environments, and seismic analysis of an isolated structure. Modification factors (λ(ae,max) and λ(ae,min)) were determined to quantify the impact of degradation on structural behavior. The results indicate that UN-FREI specimens are more sensitive to environmental conditions than UR-FREI specimens, whereas the mechanical properties of UN-FREI small-scale prototypes remain more stable compared to those of UR-FREI. This leads to increased drift, base shear, and demand-to-capacity ratios (DCRs) in the structural analysis. The findings emphasize the need for experimental testing of isolators to establish modification factors that accurately reflect the effects of environmental conditions on structures throughout their service life.

  PublisherOA PDFDOI

Recent grants

• Collaborative Research: SI2-SSI: Empowering the Scientific Community with Streaming Data Middleware: Software Integration into Complex Science Environments

  NSF · $42k · 2012–2016

• EAGER: Active Citizen Engagement to Enable Lifecycle Management of Infrastructure Systems

  NSF · $100k · 2016–2018

• Nonlinear Model Updating Using Ambient Responses For Damage Diagnosis In Concrete Structures

  NSF · $40k · 2009–2011

• CDS&E: Enabling Time-critical Decision-support for Disaster Response and Structural Engineering through Automated Visual Data Analytics

  NSF · $300k · 2016–2019

• RCN: Research Network in Hybrid Simulation for Multi-Hazard Engineering

  NSF · $500k · 2017–2023

Frequent coauthors

• Amin Maghareh

  46 shared

• Wei Song

  State Grid Corporation of China (China)

  32 shared

• Jongseong Choi

  28 shared

• Daniel Gómez

  Universidad del Valle

  28 shared

• Billie F. Spencer

  University of Illinois Urbana-Champaign

  26 shared

• Chul Min Yeum

  25 shared

• Juan M. Caicedo

  American Institute of Steel Construction

  24 shared

• Ilias Bilionis

  24 shared

Labs

• Intelligent Infrastructure Systems LabPI

Education

• PhD, Civil Engineering

  University of Notre Dame

  1996

• BS, Aeronautical and Astronautical Engineering

  University of Illinois Urbana-Champaign

  1991

Awards & honors

• National Science Foundation PECASE Award (1998)
• National Science Foundation CAREER Award (1998)
• Purdue University College of Engineering Team Excellence Awa…
• ANCRISST Young Investigator Award (2007)
• Outstanding Alumna Award for the Department of Aeronautical…