Joyce Chai

Joyce Chai

By |

My research interests are in the area of natural language processing, situated dialogue agents, and artificial intelligence. I’m particularly interested in language processing that is sensorimotor-grounded, pragmatically-rich, and cognitively-motivated. My current work explores the intersection of language, vision, and robotics to facilitate situated communication with embodied agents and applies different types of data (e.g., capturing human behaviors in communication, perception, and, action) to advance core intelligence of AI.

Anthony Bloch

Anthony Bloch

By |

My research interests include : Hamiltonian and Lagrangian mechanics, gradient flows on manifolds, integrable systems stability, the motion of mechanical systems with constraints, the relationship between continuous and discrete flows, nonlinear and optimal control and the control of quantum systems. I also interested in data-guided control and in particular the dynamics and control
of networks and systems arising from large sets, particularly in biological applications.

Katie Skinner

Katie Skinner

By |

My research spans robotics, computer vision, and machine learning with a focus on enabling autonomy in dynamic, unstructured, or remote environments across field robotics applications (air, land, sea, and space). In particular, my group focuses on problems that rely on limited labeled data.

Majdi Radaideh

Majdi Radaideh

By |

Prof. Majdi Radaideh leads the Artificial Intelligence and Multiphysics Simulations lab (AIMS), which focuses on the intersection between nuclear reactor design, nuclear multiphysics modeling and simulation, advanced computational methods, and machine learning algorithms to drive advanced reactor research and improve the sustainability of the current reactor fleet. AIMS extensively employs data science and machine learning methods for various goals including but not limited to:
1- Development of surrogate models for expensive nuclear reactor simulations in steady-state and time-dependent modes using convolutional and recurrent neural networks.
2- Large-scale combinatorial optimization to improve the performance of the nuclear fuel inside nuclear power plants using physics-informed reinforcement learning and neuroevolution algorithms.
3- Long-short term memory and ensemble methods for anomaly detection and fault prognosis to monitor the health of the nuclear power plant components.
4- Uncertainty quantification of data-driven models with Bayesian inference and Gaussian processes.
5- Natural language processing methods to process nuclear plant maintenance and burnup records.

AIMS lab aims on bridging the gap between nuclear reactor design, nuclear multiphysics modeling and simulation, advanced computational methods, and machine learning algorithms to drive advanced nuclear reactor research and improve the sustainability of the current reactor fleet to promote nuclear power as a carbon-free energy source in order to achieve net-zero carbon emission.

Wei Hu

Wei Hu

By |

Wei Hu is broadly interested in the theoretical and scientific foundations of modern machine learning, especially deep learning. His research aims to obtain a solid, rigorous, and practically relevant theoretical understanding of machine learning pipelines, as well as to develop principles to make them more reliable and efficient.

Y Z (Yang Zhang)

By | | No Comments

YZ’s research can be summarized into two words: Matter and Machine. On the Matter side, his group studies far-from-equilibrium physics. They synergistically combine and push the boundaries of statistical and stochastic thermodynamic theories, accelerated molecular simulations, understandable AI/ML/DS methods, and neutron scattering experiments, with the goal of significantly extending our understanding of a wide range of long timescale phenomena and rare events. Particular emphasis is given to the physics and chemistry of liquids and complex fluids, especially at interfaces, driven away from equilibrium, or under extreme conditions. On the Machine side, leveraging their expertise in materials and modeling, his group advances the development of soft robots and human-compatible machines, swarm robots and collective intelligence, and robots in extreme environments, which can lead to immediate societal impact.
• Matter
o Far-from-equilibrium physics, long timescale phenomena, and rare events (statistical and stochastic thermodynamic theories, accelerated molecular simulations, understandable AI/ML/DS methods)
o Neutron scattering, sources, and instrumentation
o Physics and chemistry of liquids and complex fluids, especially under interfacial/non-equilibrium/extreme conditions (water, metallic liquids, molten salts, ionic liquids, electrolyte solutions)
• Machine
o Soft robots and human-compatible machines
o Swarm robots and collective intelligence
o Robots in extreme environments

Dawn Tilbury

Dawn Tilbury

By | | No Comments

One branch of my research considers how to use data and models to improve manufacturing productivity. We look at how to capture streaming data from the factory floor, put it together with models encapsulated in “digital twins”, and use the predictions to better schedule production, plan maintenance, detect anomalies (including cyber-attacks), reconfigure the manufacturing system or supply chain to react to disturbances, and improve quality. We use models based on physics, subject-matter expertise, and historical data. We partner with various companies and industries to understand the needs and opportunities.

Another branch of my research considers how to improve the performance of human-robot teams through modeling and feedback. We run user studies with semi-automated vehicles and robots, and ask humans to work together with the automation to accomplish tasks. We collect physiological and behavioral data, build models of trust and situation awareness, and adapt the automation to enhance the overall team performance. Models are built using historical data and are informed by existing literature.


Accomplishments and Awards

 


Research Highlights

Vikram Gavini

By | | No Comments

The Computational Materials Physics group under the direction of Vikram Gavini focuses on developing computational frameworks and algorithms to enable large-scale ab-initio calculations for predictive materials simulations. The two main present thrusts of the group since 2020 are: (i) Developing fast and accurate large-scale density functional theory (DFT) calculations; (ii) Development of a computational framework for improving the accuracy of DFT calculations by improving the exchange-correlation description with an aim to advance the accuracy towards quantum accuracy. DFT-FE a massively parallel and GPU ported real-space DFT code based on finite-element discretization originated from this group. The group is focused on data-driven approaches to improve the exchange-correlation description in DFT as well as accelerate predictive atomistic calculations.

Matias del Campo

By |

The goal of this project is the creation of a crucial building block of the research on AI and Architecture – a database of 3D models necessary to successfully run Artificial Neural Networks in 3D. This database is part of the first stepping-stones for the research at the AR2IL (Architecture and Artificial Intelligence Laboratory), an interdisciplinary Laboratory between Architecture (represented by Taubman College of Architecture of Urban Planning), Michigan Robotics, and the CS Department of the University of Michigan. A Laboratory dedicated to research specializing in the development of applications of Artificial Intelligence in the field of Architecture and Urban Planning. This area of inquiry has experienced an explosive growth in recent years (triggered in part by research conducted at UoM), as evidenced for example by the growth in papers dedicated to AI applications in architecture, as well as in the investment of the industry in this area. The research funded by this proposal would secure the leading position of Taubman College and the University of Michigan in the field of AI and Architecture. This proposal would also address the current lack of 3D databases that are specifically designed for Architecture applications.

The project “Generali Center’ presents itself as an experiment in the combination of Machine Learning processes capable of learning the salient features of a specific architecture style – in this case, Brutalism- in order to generatively perform interpolations between the data points of the provided dataset. These images serve as the basis of a pixel projection approach that results in a 3D model.


Accomplishments and Awards