Title: Real-Time Iterative Software for EM-Based Motion Compensation

Date: Tuesday, April 12th, 2022

Time: 3:30 PM  - 5:00 PM EDT

Locations: Klaus Conference Room 2100

Virtual: https://gatech.zoom.us/j/93254665356

Alan Nussbaum

Computer Science Ph.D. Student

School of Computer Science

College of Computing

Georgia Institute of Technology

Committee:

Dr. Umakishore Ramachandran (advisor), School of Computer Science, Georgia Institute of Technology

Dr. Dale Blair (advisor), School of Electrical and Computer Engineering, Georgia Institute of Technology

Dr. Vivek Sartar, School of Computer Science, Georgia Institute of Technology

Dr. Ada Gavrilovska, School of Computer Science, Georgia Institute of Technolgy

Dr. Tushar Krishna, School of Electrical and Computer Engineering, Georgia Institute of Technology

Abstract:

A moving/maneuvering target (aircraft, missile, car, bus, etc.) exhibits range walk (RW) over a radar’s coherent processing interval (CPI). If the range walk and nonlinear slow-time phase changes are not compensated prior to coherent processing over slow-time, then a signal-to-noise ratio (SNR) loss is observed. Compensating for RW requires knowledge or an estimate of the target’s range rate and radial acceleration.   Measurement precision is inversely related to SNR in a tracking radar and directly impacts tracking performance. In addition, accurate estimates of abrupt changes in the target’s velocity and acceleration are needed to maintain the tracking over short time intervals.

To obtain the knowledge required to compensate for RW, a real-time sensor architecture and robust algorithms are needed to estimate the target’s radial motion and apply the appropriate corrections. At the same time, maintaining a stringent timeline and adhering to latency requirements are essential for real-time sensor processing. This research aims to examine existing methods and explore new approaches and technologies to mitigate the harmful effects of RW in tracking radar systems and implement the new methods in a real-time computing environment.

This dissertation builds upon existing Expectation-Maximization (EM) algorithm approaches that provide iterative processing to obtain maximum likelihood (ML) estimates of unobserved/unknown variables. The EM algorithms used in other applications are documented in the literature review. Prior work notes that the EM algorithm may, in some instances, have difficulty converging to an accurate ML estimate in a timely manner. This research aims to develop an iterative EM algorithm for RW correction that reliably converges and is implemented in real-time software architecture. The EM algorithm requires a seed obtained from the radar measurements and updated via a Tracker filter to ensure real-time convergence. Computational steering is required to activate the E-Step and M-Step of the iterative algorithm. The following steps list the expectation and maximization enhancements developed under this work to compensate for RW.