AI-driven earth observation and protection

Numerical Simulation · Big Data · Artificial Intelligence · Environmental Sensing · Renewable Energy · Cybersecurity

Research Vision

My research goals are enabling the next-generation AI-driven earth observation technology to fur-ther understand the health of the planet and inventing engineering solutions to mitigate climate change’s impacts on the environment and society. I am focusing on addressing emerging critical is-sues involving 5G, cyber-attacks, dynamical system theory, bionic 3D printing, space-based earth sensing, and AI.

Themes

AI-driven Earth Observation

The emerging space- and IoT-based sensing technologies provide an unprecedented variety and resolution of earth surface sensing, but critical challenges emerge when the data is used for applica-tions: 1) Insurficient spatial and temporal resolution. 2) Vulnerable data transmission and collection from the site to the data center, and 3) High costs of device deployment and mantenance.

To address these challenges, my research group developed a series of research projects and formulate visions for future research directions, including 1) 5G disruption resolution, 2) data failure-proof flood forecasting, and 3) optical-fiber and hyperspectral sensing.

Innovative Monitoring and Forecasting for Urban and Coastal Floods

Coastal and urban areas are experiencing increasing risk of flooding. A remarkable example is the increase of high tide flooding (also called “blue sky” flooding or “sunny day” flooding), which is shallow (several centimeters) but widely spread. NOAA reported that 75% of the US East and Gulf Coast’s monitored locations witnessed an increasing trend of high tide flooding. High tide flooding disrupts transportation, sewage, and other infrastructure systems, devaluates real estate, reduces income and jobs, exposes health hazards, increases public health risks, salinizes groundwater, and deteriorates coastal ecosystems. These relatively more frequent, smaller floods, at some locations, may prove to be more costly than large, infrequent extreme events.

We are developing AI-based solutions to monitor and forecast these floods: 1) A social media based flood observation platform, 2) Automatic monoplotting to mine visual data, 3) A computer vision-based real-time flood monitoring.

Environmental Co-Design of Renewable Energy

Renewable energy, as a potential energy source to offset carbon emission, is becoming a major solution to climate change. However, the conflict between the environmental requirement and energy production efficiency is difficult to resolve – the system is first designed to achieve the maximum power production efficiency, and then the final design experiences redesigning to sacrifice the system performance to meet the environmental requirements, which leads to sub-optimal performance and unnecessary workload. My research group received funding from USDOE and NOAA to study the environmental and ecological impacts of hydrokinetic and offshore wind farms, and we discovered that a co-design framework that involves the environmental requirement in the early design phase can largely improve the optimum system performance. An ongoing project is to assess the environmental impact of offshore wind farms on the New Jersey's summer flounder habitats.

Fluid-Structure-Particle Interaction in Microfluidics

Drug-resistant bacteria, or super-bugs, are a major public health concern. Globally, at least 700,000 people die each year as a result of drug-resistant infections, including 230,000 deaths from multidrug-resistant tuberculosis. That number could soar to 10 million deaths a year by 2050 if no action is taken, according to a 2019 report.

Tiny new devices called microfluidics can rapidly isolate, retrieve and concentrate target bacteria from bodily fluids. It efficiently filters particles and bacteria, capturing about 86 percent of them. Several fluid-structure interaction problems prevent researchers designing more efficient devices. We are developing advanced theoretical fluid dynamics and structure deformation models to compare with lab experiments and successfully improved various designs. The theoretical development could also lead to new biological detection method and improved bionic 3D printing technology.

WHIRLab