Fabrizio Sergi

This study validates dynamic Bayesian optimization (DBO) for human-in-the-loop exoskeleton control, showing it improves walking speed and personalization compared to standard Bayesian optimization (BO).

The paper shows that task instructions modulate stretch-evoked motor responses in the human brainstem, as evidenced by changes in reticulospinal activation during fMRI.

This study investigates how different TMS intensities affect stretch-evoked long-latency responses in the flexor carpi radialis, finding that only suprathreshold TMS inhibits these responses via a cortical silent period.

This paper develops a multi-objective optimization framework for designing a compliant gravity balancing shoulder orthosis, validated through mechanical testing and in-vivo muscle activity reduction in healthy individuals.

This study compares dynamic Bayesian optimization (DBO) and Bayesian optimization (BO) in a human-in-the-loop system, finding DBO better handles non-stationary behavior in later training stages.

This study simulates human-in-the-loop optimization using Bayesian Optimization (BO) and Dynamic Bayesian Optimization (DBO) to assess DBO's effectiveness in scenarios involving motor learning, finding that DBO generally outperforms BO after sufficient iterations.

This study uses fMRI and a robotic wrist task to identify neural networks involved in force and impedance control during movement, revealing the cortico-thalamic-cerebellar network's role in co-contraction.

This paper presents a simulation-optimization framework to design gravity-balancing orthoses that balance compensation accuracy and device protrusion.

This study develops an experimental protocol to decouple force and impedance control in wrist movements using robotics and fMRI, showing distinct torque and stiffness responses to different force conditions.

This paper develops a novel MRE-based muscle force estimator that separates shear wave speed into active and passive components, improving offset parameter estimation without affecting slope parameter accuracy.

This paper validates existing models of motor learning in gait training and proposes a modified use-dependent learning model that achieves better fit with fewer parameters.

The paper develops models to estimate anterior ground reaction force (AGRF) using kinematic and kinetic metrics in the absence of direct measurement.

The paper develops an MRI-compatible robotic system to study brainstem contributions to motor responses during different tasks, showing task-dependent changes in muscle activity and brain activation.

This paper introduces StretchfMRI, a technique to measure RST function in humans using fMRI and robotic perturbations, revealing muscle-specific brainstem activity patterns during stretch-evoked LLRs.

This paper introduces StretchfMRI, a technique to measure reticular formation function via fMRI by analyzing long-latency responses to robotic perturbations.

This study introduces a novel estimator to improve muscle force estimation from incomplete muscle activity measurements, showing a 25% reduction in mean estimation bias in simulations.