Embedded Files
Computational Approaches to Neural and Physiological Systems
Computational Approaches to Neural and Physiological Systems
Our research combines computational neuroscience, dynamical systems, control theory, and machine learning to understand complex biological systems and develop predictive approaches for neural and physiological stimulation.
We study how system states can be estimated from measured signals, how computational models can predict future behavior, and how feedback can be used to design adaptive interventions.
Model-based closed-loop framework connecting neural simulations, measured signals, reduced-order dynamics, and optimal feedback control.
Closed-Loop Neurostimulation
Closed-Loop Neurostimulation
Conventional stimulation therapies often use fixed parameters that require repeated manual adjustment. Our research explores model-based and feedback-control approaches that adapt stimulation in response to changing neural activity.
We combine large-scale neural simulations, electrophysiological signals, reduced-order models, state estimation, and optimal control to investigate how pathological neural oscillations and synchronization may be regulated.
Research directions
Research directions
Computational modeling of neural populations
Neural-state estimation
Reduced-order dynamical models
Feedback and optimal control
Adaptive stimulation strategies
Selected Published Work
Selected Published Work
Selected publications illustrate the progression from open-loop neural stimulation strategies toward feedback-controlled regulation of network synchrony.
PI-controller-based closed-loop FTSTS framework for regulating synchronization in an excitatory-inhibitory neural network.
Source: Olumuyiwa and Kumar, 2025 American Control Conference, Figure 1.
Published Framework: PI-Controlled FTSTS
Published Framework: PI-Controlled FTSTS
This framework extends Forced Temporal Spike-Time Stimulation into a closed-loop strategy. The mean firing rate of inhibitory neurons provides the feedback signal, while a proportional-integral controller adjusts stimulation amplitude to regulate synchronization in an excitatory-inhibitory neural network.
View publication →Published Result: Closed-Loop Desynchronization
Published Result: Closed-Loop Desynchronization
In simulation, the PI-controlled FTSTS strategy adjusts stimulation amplitude to drive inhibitory-neuron firing rate toward its target and move the E–I network from a synchronized state toward an asynchronous regime.
Closed-loop FTSTS result showing changes in firing rate, network spectra, spiking activity, stimulation amplitude, and E-to-I synaptic weight during desynchronization. View publication →
Source: Olumuyiwa and Kumar, 2025 American Control Conference, Figure 7.
Auditory Stimulation and Affective Dynamics
Auditory Stimulation and Affective Dynamics
Ongoing research
Ongoing research
We are investigating how noninvasive auditory stimulation may influence emotional states and how EEG signals can characterize the associated neural dynamics.
This work combines controlled auditory stimuli, EEG measurements, signal processing, and computational modeling to examine changes in neural and affective states over time.
Research directions
Research directions
Controlled auditory stimulation
EEG acquisition and signal processing
Neural and affective-state characterization
Low-dimensional representations
Computational modeling
Future adaptive-stimulation strategies
Controlled auditory stimulation, EEG measurements, affective-state characterization, and computational analysis.
Experimental Setup
Experimental Setup
We use noninvasive electrophysiological measurements during controlled auditory-stimulation experiments to study changes in neural and affective dynamics.
Auditory-stimulation experiment
EEG preparation and recording
Noninvasive EEG sensor array
Computational framework for predicting gait events from neural signals recorded during deep brain stimulation.
Prediction of Gait Events Using Deep Learning
Prediction of Gait Events Using Deep Learning
Ongoing collaborative research
Ongoing collaborative research
Gait impairment can substantially affect mobility and quality of life in people with Parkinson’s disease. In collaboration with Professor Doris Wang at UCSF, we are investigating deep-learning approaches that use neural data recorded during deep brain stimulation to predict the onset of gait events.
This work investigates whether computational models can characterize gait-related neural dynamics across stimulation settings. Such models may ultimately help inform patient-specific stimulation strategies and the development of adaptive closed-loop approaches for gait support.
Research directions
Research directions
Gait-event detection and prediction
Neural-signal processing
Deep-learning model development
Evaluation across stimulation settings
Patient-specific computational modeling
Adaptive closed-loop stimulation concepts
Computational Modeling and Control
Computational Modeling and Control
We develop computational methods for understanding complex biological systems from measured and simulated data. These methods help identify important system states, characterize nonlinear dynamics, and support prediction and control.
Our work combines mechanistic models with data-driven techniques to study systems that are high-dimensional, partially observed, and dynamically changing.
Methods
Methods
Dynamical-systems analysis
State-space modeling
Dimensionality reduction
System identification
Machine learning
Prediction and forecasting
Feedback and optimal control
Framework linking neural activity, low-dimensional state estimation, computational modeling, prediction, and feedback control.
From Signals to Models to Intervention
From Signals to Models to Intervention
Across our research, we follow a common framework: measure biological activity, estimate the underlying system state, develop predictive models, and use those models to inform feedback-control strategies.
Current Support
Current Support
This work is supported by the National Institutes of Health (Grant Number: R16NS140310) through research focused on computational modeling and closed-loop approaches to neural stimulation.
Page updated
Report abuse