Graduation Semester and Year




Document Type


Degree Name

Doctor of Philosophy in Industrial Engineering


Industrial and Manufacturing Systems Engineering

First Advisor

Brian Huff


This dissertation presents an organic approach to the cooperative localization problem by sequentially solving the problems of sensor calibration, multi-sensor fusion filtering, and cooperative localization. Successful navigation of unmanned ground vehicles requires accurate localization. Localization refers to the determination of the pose (position and orientation) of an unmanned ground vehicle with respect to a local or a global frame of reference. Cooperative localization is suited to multi-vehicle systems where vehicles with better accuracy in localization can assist those with poor accuracy through communication and relative pose sensing. A parametric modeling approach is presented for sensor calibration. Designed experiments are conducted with the objective of building parametric models and mass assignment tables. An evidence theoretic adaptive fusion filter, the Eta-Filter, is proposed for multi-sensor fusion filtering. The Eta-Filter leverages the Dempster-Shafer theory of evidence to make a Kalman filter adaptive to operating scenarios and sensor goodness while accounting for the ignorance component of uncertainty. It is composed of an adaptive pre-processing unit, an evidence extraction and combination unit, and a Kalman filter. The evidence extraction and combination unit uses fuzzy-type techniques or rule-based mass assignment tables to compute the mass function. Then, the Dempster's rule for combination is used to combine the disparate evidences for a proposition. Based on the combined evidence, decisions on switching between pre-processing models and between corresponding input noise covariance matrices in the adaptive pre-processing unit are made. Also, the measurement noise covariance matrix of the Kalman filter is varied depending upon the evidence that the sensor is good. Experiments that demonstrate the validity of the Eta-Filter using empirical data are presented. A range-only cooperative localization system that resembles a "star" arrangement is presented. Combination is performed in the minimum variance sense under the assumption of independence of errors between the individual estimates. A statistically designed experiment that demonstrates the merits of the range-only cooperative localization system is presented. An ANOVA F-test, conducted at the one percent significance level, reveals that the range-only cooperative localization system has a significantly lower mean final position error when compared to a non-cooperative localization system.


Engineering | Operations Research, Systems Engineering and Industrial Engineering


Degree granted by The University of Texas at Arlington