Aerospace and Mechanical Engineering
Adjust Font Size: Normal Large X-Large

AEM spotlight:

Recent grad Vibhor Bageshwar to join Honeywell

The objective of state vector estimation problems is to estimate the statistics of a state vector governed by a system model. Errors from sources such as model uncertainty and sensor measurements are inevitable. Part of the challenge in these problems is to estimate or bound the effects of these error sources using a filter. The most popular filter is the industry standard Kalman filter. Vibhor Bageshwar, a recent Ph.D. graduate in Aerospace Systems, worked on quantifying the performance limitations of the Kalman filter both in theory and in practical applications such as attitude determination systems. Bageshwar is set to start at Honeywell this February, where he will continue to work on concepts he developed during his Ph.D.

In framing an explanation of his thesis, Bageshwar states: “theory is not enough; the theory must be applied to practical applications with the associated errors.”

This idea permeates Bageshwar’s graduate career, which included research on adaptive cruise control systems, near Earth orbit and low Earth orbit satellite pointing systems, the Kalman filter, and attitude determination (AD) systems using low-cost sensors. Attitude refers to a vehicle's three-dimensional angular orientation.

A Kalman filter, under various assumptions on the system model, is a recursive algorithm that estimates the mean and variance of a state vector. The variance can be interpreted as the estimation errors of the state mean vector. The design question in state vector estimation problems is how should the system model be selected so that the Kalman filter provides unbiased, minimum variance estimates of the state mean vector. A significant component of Bageshwar's doctoral research was to develop a test to characterize the variance of the estimated state mean vector for general system models.

“This test allows the user to predict the performance of the Kalman filter during the design and selection process of the system model,” he explains. “Therefore, the user can adjust the system model or tune the filter before the filter is used on the real system.”

As part of his doctoral research, Bageshwar also designed the AD system for the University’s entry into the Air Force/NASA University Nanosat-4 competition, a national program where select universities build a satellite from conceptual design to prototype and compete to have it launched into space.

For the University’s entry into the competition – Minnesat – Bageshwar designed an AD system using low cost sensors and a Kalman filter to blend measurements from an inertial sensor and a magnetometer. A magnetometer measures the local magnetic field of the Earth. He demonstrated that this sensor set could be used for satellite AD systems and identified the quality of the sensors required to determine Minnesat’s attitude in support of Minnesat’s scientific mission. This technology helped Minnesat garner a fifth-place finish in its first entry into the competition.

“Typically, an AD system uses inertial sensors, however, the sensor measurement errors cause the estimation error of attitude to grow without bound,” he explains. “So the challenge is to estimate the sensor measurement errors and minimize their effect on the attitude estimates. The standard approach is to use additional sensors to aid the inertial sensors.”

Normally, two aiding sensors are used in AD systems. However, design trade-offs including Minnesat’s mission, size, weight, and cost constraints dictated that only one aiding sensor could be used.

“We wondered if we could design an AD system using one aiding sensor and how accurate the attitude estimates would be,” he recalls.

Demoz Gebre-Egziabher, a McKnight Land-Grant professor in AEM and adviser to Nanosat-4, says Bageshwar effectively designed the entire navigation, guidance and control system for Minnesat and trained the undergraduate engineering students on those aspects of the satellite.

“He developed an algorithm that will allow low-cost or off-the-shelf sensors to be used in small satellites for attitude determination,” Gebre-Egziabher explains. “His work will be used as legacy designs for all future small satellites the department builds.”

Last Modified: Thursday, 29-Jun-2017 14:16:06 CDT -- this is in International Standard Date and Time Notation