About

Mazen Farhood is a professor in the Kevin T. Crofton Department of Aerospace and Ocean Engineering at Virginia Tech. He holds a Ph.D. in Mechanical Engineering from the University of Illinois at Urbana-Champaign, obtained in 2005, along with a Master's degree in Mechanical Engineering from the same university and a Bachelor's degree from the American University of Beirut, Lebanon. His research expertise focuses on dynamics and control, with particular emphasis on autonomous systems, control, and estimation. Dr. Farhood has been a faculty member at Virginia Tech since 2008, progressing from Assistant Professor to Professor by 2022. His professional service includes memberships in the Institute of Electrical and Electronics Engineers (IEEE), the American Institute of Aeronautics and Astronautics (AIAA), and the American Society of Mechanical Engineers (ASME). His contributions to the field have been recognized through awards such as the NSF CAREER Grant in 2014. His academic and research activities are centered around advancing knowledge in dynamics and control, with a focus on autonomous systems, and he actively participates in professional service and leadership within his field.

Research topics

• Computer Science
• Mathematics
• Artificial Intelligence
• Mathematical optimization
• Distributed computing
• Algorithm
• Discrete mathematics
• Computer network

Selected publications

• Quadratic Characterizations for Reachability Analysis of Neural Networks

  ArXiv.org · 2026-05-19

  articleOpen access

  Quadratic constraints (QCs) are widely used to characterize nonlinearities and uncertainties, but generic analytical characterizations can be conservative on bounded domains. This paper develops a framework for constructing verified quadratic characterizations of scalar relations in the two-dimensional real plane. Candidate quadratic inequalities are locally generated by solving convex quadratic programs using samples from the relation and exterior sample points. They are then verified globally using sum-of-squares certificates over an exact semialgebraic description or, in the case of nonpolynomial relations, over relaxed polynomial descriptions. The resulting verified constraints define a sound overapproximation of the scalar relations over the considered domains. These constraints are directly compatible with existing analysis frameworks based on QCs and pointwise integral quadratic constraints (IQCs) for static nonlinearities and uncertainties, and they can also be embedded in QC-based semidefinite programs for reachability and safety analysis of feedforward neural networks. For smooth activations such as $\tanh$, the method yields domain-dependent quadratic characterizations that constitute an alternative to generic sector- or slope-based descriptions. For ReLU networks, we give methods to reduce conservatism in QC-based reachability analysis of feedforward networks by exploiting dependencies between neurons and tighter local bounds. Numerical examples demonstrate improved reachability results for smooth activations, reduced conservatism for ReLU networks, and applicability beyond neural networks through an example involving saturation.

  PublisherOA PDF

• Hypernetwork-Conditioned Reinforcement Learning for Robust Control of Fixed-Wing Aircraft under Actuator Failures

  arXiv (Cornell University) · 2026-04-03

  preprintOpen accessSenior author

  This paper presents a reinforcement learning-based path-following controller for a fixed-wing small uncrewed aircraft system (sUAS) that is robust to certain actuator failures. The controller is conditioned on a parameterization of actuator faults using hypernetwork-based adaptation. We consider parameter-efficient formulations based on Feature-wise Linear Modulation (FiLM) and Low-Rank Adaptation (LoRA), trained using proximal policy optimization. We demonstrate that hypernetwork-conditioned policies can improve robustness compared to standard multilayer perceptron policies. In particular, hypernetwork-conditioned policies generalize effectively to time-varying actuator failure modes not encountered during training. The approach is validated through high-fidelity simulations, using a realistic six-degree-of-freedom fixed-wing aircraft model.

  PublisherDOI

• Hypernetwork-Conditioned Reinforcement Learning for Robust Control of Fixed-Wing Aircraft under Actuator Failures

  arXiv (Cornell University) · 2026-04-03

  articleOpen accessSenior author

  This paper presents a reinforcement learning-based path-following controller for a fixed-wing small uncrewed aircraft system (sUAS) that is robust to certain actuator failures. The controller is conditioned on a parameterization of actuator faults using hypernetwork-based adaptation. We consider parameter-efficient formulations based on Feature-wise Linear Modulation (FiLM) and Low-Rank Adaptation (LoRA), trained using proximal policy optimization. We demonstrate that hypernetwork-conditioned policies can improve robustness compared to standard multilayer perceptron policies. In particular, hypernetwork-conditioned policies generalize effectively to time-varying actuator failure modes not encountered during training. The approach is validated through high-fidelity simulations, using a realistic six-degree-of-freedom fixed-wing aircraft model.

  PublisherOA PDF

• Quadratic Characterizations for Reachability Analysis of Neural Networks

  arXiv (Cornell University) · 2026-05-19

  preprintOpen access

  Quadratic constraints (QCs) are widely used to characterize nonlinearities and uncertainties, but generic analytical characterizations can be conservative on bounded domains. This paper develops a framework for constructing verified quadratic characterizations of scalar relations in the two-dimensional real plane. Candidate quadratic inequalities are locally generated by solving convex quadratic programs using samples from the relation and exterior sample points. They are then verified globally using sum-of-squares certificates over an exact semialgebraic description or, in the case of nonpolynomial relations, over relaxed polynomial descriptions. The resulting verified constraints define a sound overapproximation of the scalar relations over the considered domains. These constraints are directly compatible with existing analysis frameworks based on QCs and pointwise integral quadratic constraints (IQCs) for static nonlinearities and uncertainties, and they can also be embedded in QC-based semidefinite programs for reachability and safety analysis of feedforward neural networks. For smooth activations such as $\tanh$, the method yields domain-dependent quadratic characterizations that constitute an alternative to generic sector- or slope-based descriptions. For ReLU networks, we give methods to reduce conservatism in QC-based reachability analysis of feedforward networks by exploiting dependencies between neurons and tighter local bounds. Numerical examples demonstrate improved reachability results for smooth activations, reduced conservatism for ReLU networks, and applicability beyond neural networks through an example involving saturation.

  PublisherDOI

• Robust Control Design and Analysis Based on Lifting Linearization of Nonlinear Systems Under Uncertain Initial Conditions

  International Journal of Robust and Nonlinear Control · 2025-12-09

  articleOpen accessSenior author

  ABSTRACT This paper presents a robust control synthesis and analysis framework for nonlinear systems with uncertain initial conditions. First, a deep learning‐based lifting approach is proposed to approximate nonlinear dynamical systems with linear parameter‐varying (LPV) state‐space models in higher‐dimensional spaces while simultaneously characterizing the uncertain initial states within the lifted state space. Then, convex synthesis conditions are provided to generate full‐state feedback nonstationary LPV (NSLPV) controllers for the lifted LPV system. A performance measure similar to the ‐induced norm is used to provide robust performance guarantees in the presence of exogenous disturbances and uncertain initial conditions. The paper also includes results for synthesizing full‐state feedback linear time‐invariant controllers and output feedback NSLPV controllers. Additionally, a robustness analysis approach based on integral quadratic constraint (IQC) theory is developed to analyze and tune the synthesized controllers while accounting for noise associated with state measurements. This analysis approach characterizes model parameters and disturbance inputs using IQCs to reduce conservatism. Finally, the effectiveness of the proposed framework is demonstrated through two illustrative examples.

  PublisherDOI

• Automatically Stopping Bayesian Optimization under Uncertain Surrogate Model Means

  2025-12-09

  article

  Stopping criteria for Bayesian optimization (BO) automatically terminate the optimization algorithm when a near-optimal solution has likely been reached, avoiding unnecessary expenditure of computational resources. Existing criteria, however, only guarantee accuracy given a well-specified Gaussian process surrogate model, an assumption that often does not hold in practice. We propose a stopping criterion for Bayesian optimization when the mean function of the Gaussian process surrogate model is uncertain, but modeled with a prior. The prior induces a probability distribution over surrogate models, which our criterion uses to evaluate the probability that a model will be stopped under an existing stopping criterion. We demonstrate that our criterion will eventually terminate BO, with high probability of achieving a desired accuracy to the optimal value. Empirical analysis on a controller tuning problem for an autonomous underwater vehicle suggests that our method can terminate BO with high accuracy solutions, particularly for low dimensional problems.

  PublisherDOI

• Primitive-Based State Invariants for Nonstationary LPV Systems

  2025-12-09

  articleSenior author

  This paper presents an approach based on recently developed reachability analysis tools to validate primitive-based, switched, linear control laws. The controlled system is formulated as a switched system comprising a family of dynamic subsystems, each represented by a nonstationary linear parameter-varying (NSLPV) model. We derive convex conditions to generate ellipsoidal invariant sets that bound the initial and final states of NSLPV models subject to finite-energy or pointwise-bounded exogenous inputs. Additionally, we provide a routine to construct a common invariant set that remains valid across all dynamic subsystems, establishing stability and performance guarantees for the switched control law. The approach’s effectiveness is demonstrated in a four-thruster hovercraft example.

  PublisherDOI

• Formally Proving Invariant Systemic Properties of Control Programs Using Ghost Code and Integral Quadratic Constraints

  Lecture notes in computer science · 2025-01-01

  book-chapterSenior author
  PublisherDOI

• Adversarial Reinforcement Learning for Robust Control of Fixed-Wing Aircraft under Model Uncertainty

  ArXiv.org · 2025-10-18

  preprintOpen accessSenior author

  This paper presents a reinforcement learning-based path-following controller for a fixed-wing small uncrewed aircraft system (sUAS) that is robust to uncertainties in the aerodynamic model of the sUAS. The controller is trained using the Robust Adversarial Reinforcement Learning framework, where an adversary perturbs the environment (aerodynamic model) to expose the agent (sUAS) to demanding scenarios. In our formulation, the adversary introduces rate-bounded perturbations to the aerodynamic model coefficients. We demonstrate that adversarial training improves robustness compared to controllers trained using stochastic model uncertainty. The learned controller is also benchmarked against a switched uncertain initial condition controller. The effectiveness of the approach is validated through high-fidelity simulations using a realistic six-degree-of-freedom fixed-wing aircraft model, showing accurate and robust path-following performance under a variety of uncertain aerodynamic conditions.

  PublisherOA PDFDOI

• Development and application of a dynamic obstacle avoidance algorithm for small fixed-wing aircraft with safety guarantees

  Control Engineering Practice · 2025-12-23

  articleSenior author
  PublisherDOI

Recent grants

• CAREER: Provably Correct Control Software for Autonomous High-Performance Agents

  NSF · $400k · 2014–2020

• Systematic Control Algorithms and Computational Tools for Complex Distributed Systems

  NSF · $275k · 2013–2016

Frequent coauthors

• Dany Abou Jaoude

  American University of Beirut

  23 shared

• Geir E. Dullerud

  University of Illinois Urbana-Champaign

  19 shared

• Devaprakash Muniraj

  Indian Institute of Technology Madras

  14 shared

• J. Micah Fry

  MIT Lincoln Laboratory

  8 shared

• Mark C. Palframan

  Virginia Tech

  7 shared

• Sourav Sinha

  6 shared

• Daniel J. Stilwell

  6 shared

• Carolyn L. Beck

  5 shared

Education

• Ph.D., Mechanical Engineering

  University of Illinois at Urbana-Champaign

  2005

• M.S., Mechanical Engineering

  University of Illinois at Urbana-Champaign

  2001

Awards & honors

• 2014 NSF CAREER Grant Recipient