Dr. Christopher M. Kreucher's Home Page Welcome to Chris Kreucher's Home Page

A Brief Biosketch
My Academic genealogy


Current Work
I am currently doing research in the fields of multisensor fusion, nonlinear filtering, and sensor management. Past and current collaborators in this work include Drs. Mark Stuff and Web Stayman (Michigan Tech Research Institute), Dr. Keith Kastella (SRI International), Prof. Al Hero (The University of Michigan), Dr. Ben Shapo (Integrity Applications Incorporated), Prof. Edwin Chong (Colorado State University) and Dr. Mark Morelande (The University of Melbourne).

You may browse and download publications related to our work here. I can be reached via email as ckreuche at umich.edu.

My work involves the following interrelated topics.
  • Bayesian Multitarget Tracking. Our group is interested in multi-target tracking via non-linear filtering, specifically using the Monte Carlo technique known as particle filtering. We formulate the problem of tracking multiple moving targets in a Bayesian setting and recursively estimate the joint multitarget probability density (JMPD) as time evolves and new measurements become available. Implementationally, we use a unique particle filtering approach to recursively estimate the JMPD that emcompasses both the unknown number of targets and their unknown states. The key to computationally tractability in this high dimensional space is a novel adaptive sampling scheme that automatically factorizes the JMPD when permissible and provides a measurement directed bias for target addition and removal. A detailed exposition of this work is given in this paper published in the IEEE Transactions on AES.

  • Information-based Sensor Resource Allocation. We are also interested in sensor management using information theoretic measures. Sensor management is the problem of determining how to direct agile sensors, where each sensor may have many modes and many pointing directions available. We use the JMPD estimated by the tracking algorithm and formulate the sensor management problem as one of tasking the sensor to make the measurement that maximizes the expected gain in information, as measured by the Renyi divergence. A detailed exposition of this work is present in this Signal Processing paper.

  • Multisensor Distributed Sensor Management. We are interested in decentralized, limited communication methods of performing sensor management in a multisensor (hundreds or thousands of sensors) environment. Our work in this area is documented in this paper that recently appeared in The Proceedings of the IEEE.

  • Multistage Sensor Scheduling. We are interested in extending the sensor management scheme to long-term (nonmyopic) sensor scheduling. We have investigated two approximate methods for long-term sensor scheduling as the exact solution is computationally intractable. First, we have developed an information-directed search algorithm which focusses the Monte Carlo evaluations on action sequences that are most informative. Second, we have developed an approximate method of solving the Bellman equation which replaces the value-to-go with an easily computed function that approximates the long term value of the current action. Some preliminary work is available in this paper.


Course Information

From September 1998 until August 2002, I was an adjunct lecturer in the Electrical and Computer Engineering Department at the University of Michigan - Dearborn, where I taught ECE 210, ECE 273, ECE 365, ECE 460, and ECE 500 at one time or another. Students are directed to these links which contain solutions to old quizzes, exams, homeworks, and lab assignments.

ECE 210 Course Materials
ECE 273 Course Materials
ECE 460 / ECE 365 / ME 442 Course Materials
ECE 500 Course Materials


Courses Taught:
Fall 1998: ECE 210
Winter 1999: ECE 460 & ECE 365 (ME 442)
Spring/Summer 1999: ECE 210
Fall 1999: ECE 500
Winter 2000: ECE 460 & ECE 365 (ME 442)
Spring/Summer 2000: ECE 365 (ME 442)
Fall 2000: ECE 210
Winter 2001: ECE 460
Winter 2002: ECE 273
Spring/Summer 2002: ECE 273