ALIVE

The overarching theme for our research is the development of Computational Intelligence (Autonomy, Learning, Infomation and Vision) for Earth applications (ALIVE), interpreted broadly. The key insights of our work are at the interface of data and models, physics and statistics, crowds and computing,  features and flows, among others, where new algorithms and systems emerge in a number of projects that we are working on: 

  • Cooperative Autonomous Observing Systems (caos.mit.edu): Small, unmanned aircraft systems in a Dynamic Data-driven  (DDDAS) paradigm for mapping and monitoring. We specifically propose an autonomous system to simultaneously localize and map coherent structures in the large-scale background by tightly coupling reduced modeling, uncertainty quantification, data assimilation, adaptive observation and planning in a closed loop.
  • Animal Biometrics (sloop.mit.edu):  Building an accurate ecological information system depends on an keep accurate inventory of individuals. We bring vision-based methods to individual animal identification, or Animal Biometrics. Sloop is one of the earliest, and remains one of the largest systems for individual identification. Among many firsts: adaptation of generic visual features into specific animal identification strategies; boosted retrieval and learning from relevance feedback using experts and crowds; large-scale conservation application. Other features include RESTful architecture, rapid-distributed closure, randomized representations, hybrid shape-context, and scale-cascaded deformation invariance. The techniques prompted by this work also have broad and exciting applications in Vision.
  • Coastal Risk and Environmental Sustainability (crises.mit.edu) Our interdisciplinary project develops methods for risk assessment (including uncertainty quantification), and mitigation. The approaches we take draw from statistical signal processing, physically-based models, machine learning and economics and have many other applications in flood hazard predictionand site or basic specific risk assessment (windrisktech.com).
  • Statistical Theory of Inference for Coherent Structures (stics.mit.edu):   Fluids have features. Lots of them. We call them "coherent structures in turbulence." These features have always been used to describe fluids but they are rarely used to solve inference problems. We've wondered why that is. This area considers patterns emergent in coherent fluids as a means to efficient Geophysical Inference including Data Assimilation, Uncertainty Quantification, Downscaling (Super-resolution), Nowcasting, Velocimetry, and Mapping, among others.
    • Field Alignment System and Testbed: Publicly released, patented codes that exploit pattern information in data assimilation.
       
  • FLuid imaging eXperiments (flux.mit.edu):   Discover new techniques to image and understand fluid behavior in the laboratory and in the field.
    • Turbulence Imaging.
    • Lightfield Imaging of fluid interfaces and ground mapping.
    • 3D-Imaging of bottom topography.
    • Synthetic Aperture Imaging of Bubbly Flows. Check out Lightfield imaging of bubbles! 
    • Planet-in-a-Bottle Project: The use of coupled physical-numerical systems to study geophysical fluids in the laboratory, demonstrated in this differentially-heated, rotating fluid experiment. 
    • Particle Tracking: The classic multi-subject tracking problem extended to particles in a fluid.