Study & analyze the structure, performance, manufacturability, scalability, cost & testability of various Wave-Energy Converters (WECs) and choose the best candidate for design & implementation.

Create a high-performance WEC through

• physical experiments by designing & implementing WECs & test them in the wave tank, and

• computational experiments by running WEC simulations.

Study the AI framework of Reservoir Computing (RC) to perform pattern recognition of temporal data such as speech, ECG signals, etc. This type of pattern recognition traditionally was carried out using a type of neural network called Recurrent Neural Network (RNN). However, training RNN takes a long time and consumes huge amount of computing power. The RC utilizes the physical properties of the medium such as water to reduce the training time dramatically. The first stage of the project is to design & implement a “Liquid Brain” by using a ripple tank & actuators. 

The domain knowledge & experiences learned from this first-stage project will be used to explore the design & implementation of of RC with more compact & efficient materials & mechanism. 

Use AI techniques such as Large Language Models (LLMs) and Reinforcement Learning (RL) to optimize the design & manufacturing process of F1 in Schools race cars. It might integrate the 3D modeling with Computer-Aided Design (CAD) tools, aerodynamic simulation with Computational Fluid Dynamics (CFD) simulators, structure strength analysis with Finite-Element Analysis (FEA) software, 3D printing & Computer Numeric Control (CNC) processes with Computer-Aided Manufacturing (CAM) packages & virtual track testing with Race Time Calculator. The ultimate goal is to automatically turn the F1 in Schools Technical Regulations into the high- performance F1 in Schools race car design. This project also incorporates the theoretical & experimental study of the implementation of the physical cars such as the choices of assembly configurations, finishing processes, manufacturing methods, vendors, machines, materials, parts, etc.. 

Twelve million blind and visually impaired (BVI) people in US alone face challenges traveling outdoor independently, particularly in a complex urban environment. There are two possible elements for this project:

1. Extracting Street Information Using Google Maps

2. Smart Glass Navigation App for the Blind

The first part is focusing on the static information we can acquire & integrate into the existing Google Maps. The BVI people can potentially use it to learn about the trip details in advance or get real-time support on the street. 

The second part is focusing on the dynamic information (cars, pedestrians, obstacles, etc.) they may encounter while navigating in the city. The ultimate goal is to integrate these two parts together and create an open-source app on mobile phones such that the BVI people can have a powerful, yet low-cost solution to support their safe navigation around the city. 

This project is support by an NSF-founded project "Training a Virtual Guide Dog for Visually Impaired People to Learn Safe Routes Using Crowdsourcing Multimodal Data" (Award # 2131186)   

The Large Language Model (LLM) such as ChatGPT has caused a disruption to the industry and ignited a global AI race. Since the LLMs were trained using huge amount of human-created data. We might be able to view a LLM as a condensed entity of collective human intelligence & culture. We might be able to study LLM itself as we are study a human being. This project is aiming to study LLMs from a psychological point of view. For example, we will study the "personality traits" of LLMs. We might be able to use another LLM to play the role of a psychologist and make the study process automatic.        

Study the performance of Triply Periodic Minimal Surface (TPMS) breakwaters & seawalls and find the best TPMS type and design parameters through design & executing

• physical experiments by using 3D-printed TPMS models & the wave tank, and

• computational experiments by running Computational Fluid Dynamics (CFD) simulations.