Thermodynamic State Variable Estimation on Distributed Networks of Computational Sensor Devices

Speaker: Ben M. Jordan, Department of Organismic and Evolutionary Biology at Harvard University; CEO of Sense Ai, Inc.

Date & Time:

Tuesday, April 4, 2017
12:00 PM - 1:00 PM

Location:

OWS 250

Abstract: Embedded, mobile, and traditional computational devices are outfitted with a variety of sensors capable of measuring physical and statistical quantities in their own local neighborhood at relatively high frequency and precision. There are nearly 20 billion such devices currently connected across private and public networks with myriad applications. Excess computational capacity and near-universal C-language compliance enable these devices to perform operations on the time series data provided by the sensing hardware. These operations typically include filtering, serializing, storing, or transmitting the data, but recent work has expanded the role of local computation to include prediction and estimation of quantities for which there are no direct sensor measurements available; a process dubbed "sensor fusion" in commercial settings.  The focus of the talk will be the physics, math, and computing issues involved in continuous prediction of quantities of interest on the local device. A parameterized compartmental thermodynamic model for a generic device and its surroundings will be discussed, with particular attention paid to the relationship between heat transfer, battery chemistry and electronics.  Experimental methods for determining device-specific parameters will be covered as well. The majority of the talk will focus on the numerical solution of the resulting system of differential equations, regression on the sensor time series data, and the estimation of free parameters and variables using numerical minimization techniques. Results from 10000 Internet-connected devices from across the planet collecting 65 sensor variables over 6 months will be reviewed. Demonstration devices will be available during the talk. Links to Kalman filtering will be discussed if time permits.

Biographical Information: Ben M. Jordan, Department of Organismic and Evolutionary Biology at Harvard University; CEO of Sense Ai, Inc. Accurate prediction of the physics of biological systems and their surroundings requires both a mathematical description whose assumptions are clear, and experimental techniques capable of measuring the parameters involved. My research uses thermodynamic mixture theory as a starting point to derive models for understanding how state variables such as temperature, pressure, concentration, and electrical potential contribute to environmental interactions through heat, mass and momentum transfer, as well as shape change. Custom experimental devices for measuring fiber reinforcement, intracellular pressure, strain rates, geometry, and other parameters and variables have been constructed to verify model predictions, often using regression techniques. Codes for the analysis of measurement data, solutions of systems of equations, and custom software libraries and apps have been developed along the way. Successful projects include the development of chicken limbs, patterning of zebrafish embryos, metabolic regulation in hydrothermal vent worms, shape change in cannabinoid-stimulated mouse neurons and giant-celled algae, and environmental sensing by mobile and IoT devices. I received my BS from the U of M in computer science and math, and my PhD from Harvard in biomechanics. I am the founder and CEO at Sense Ai, and the chief scientist at ABV Technology. My other interests and activities include triathalons, woodworking, cosmology, and my family. 

All programs offered by the University of St. Thomas shall be readily accessible to individuals with disabilities. For details, call (651) 962-6315.