Professor Ferrari's research aims at developing novel techniques for adaptive control of complex systems. One aspect of this research consists of guaranteeing the robustness and stability of nonlinear adaptive controllers that optimize performance over time. Another aspect is the development of practical modeling and control techniques for heterogeneous and distributed systems, such as, sensor surveillance systems.

Recent efforts have focused on the development of a flight control system that learns in real time how to provide improved safety and performance in the presence of unforeseen conditions, such as, emergency maneuvers, control failures, and aircraft parameter variations. This control system is characterized by the same robustness and stability characteristics as classical controllers, and utilizes adaptive neural networks to adapt to the actual aircraft dynamics in real time. One of her main contributions was the development of several algebraic learning techniques that exhibit unsurpassed speed and provide an innovative framework for investigating neural approximation properties. Following this work, Professor Ferrari combined these learning techniques with stability analysis tools based on Linear Matrix Inequalities (LMIs) to design a neural network controller with proven closed-loop stability. Most recently, she and her students introduced the use of Bayesian networks for multiple, heterogeneous sensor modeling and fusion. This technique was successfully applied to Ground Penetrating Radar (GPR), InfraRed (IR), and Electromagnetic Induction (EMI) sensors, for the detection and classification of buried mines. In an on-going collaboration with the International Crime Analysis Association, Professor Ferrari and her students are also working on developing computational tools for criminal profiling.