Rohan Thorat

About | Portfolio | Blogs | Videos | GitHub | Publications | Teaching | Extra-activities

Research & Projects

Research at the intersection of mechanics, machine learning, and control.

Position PhD Researcher, IIT Delhi
Advisor Prof. Rajdip Nayek
Focus Control & System Identification
Publications 1 Journal · 2 Conference
Reinforcement Learning Structural Control System Identification Variational Bayes Modal Reduction Hyperelastic Materials Safe RL Digital Twins Policy Gradient Bayesian Filtering Vibration Control Actor-Critic Reinforcement Learning Structural Control System Identification Variational Bayes Modal Reduction Hyperelastic Materials Safe RL Digital Twins Policy Gradient Bayesian Filtering Vibration Control Actor-Critic
0
Years Research
0
Journal Paper
0
Conference Papers
Curiosity

Featured

Research, in the round.

001 Preprint

ML-Assisted State Prediction under Model-Form Error

SIMPLY SUPPORTED BEAM · 1ST MODE
Read Paper →
002 Accepted · ICCMS 2025

Safe RL for Structural Vibration Control via LQR Guidance

🧠 RL AGENT u(t) x(t) LQR SAFETY SAFE RL VIBRATION CONTROL
View Project →
003 Published

System ID & Reliability of Hyperelastic Materials via Variational Bayes

HYPERELASTIC STRESS-STRAIN ID
Read Paper →
004 Blog

Deriving Actor-Critic Methods with Intuition

ENVIRONMENT s , r ACTOR π(a|s;θ) CRITIC V(s;w) a(t) s(t) A(s,a) r(t) ∇J(θ) = 𝔼[ A · ∇log π ]
Read Blog →

Hover (or touch & hold) the ring to pause · full details below

Filter
001
Preprint

ML-Assisted State Prediction under Model-Form Error

Structural models always carry modelling assumptions — wrong beam theory, misspecified damping, incorrect boundary conditions. This framework uses a Gaussian Process Latent Force Model (GPLFM) in the reduced modal domain to learn and correct these discrepancies on-the-fly. A mesh-invariant neural network maps modal states to error estimates, validated across five realistic error scenarios including unmodelled nonlinearity and local damage.

GPLFM Modal Reduction Bayesian Filtering Digital Twins
Read Paper
x=0 x=L/2 x=L True response Misspecified model Model-form error SIMPLY SUPPORTED BEAM · 1ST MODE
002
Accepted · ICCMS 2025

Safe RL for Structural Vibration Control via LQR Guidance

Training RL directly on a physical structure is risky — random exploration can cause damage. This work proposes a hybrid LQR+RL framework where an LQR controller (derived from arbitrarily chosen, potentially wrong parameters) linearly mixes with the RL policy during training to keep exploration safe. After training, the RL policy takes over and surpasses LQR in vibration suppression with lower control effort — all without ever identifying the true system model.

Safe RL LQR Guidance Vibration Control
View Project
🧠 RL AGENT u(t) x(t) u(t) LQR SAFETY SAFE RL VIBRATION CONTROL
003
Published

System ID & Reliability of Hyperelastic Materials via Variational Bayes

Hyperelastic materials like rubber exhibit highly nonlinear stress-strain behaviour that is hard to model from first principles. This work uses a sparsity-promoting variational Bayesian framework to automatically select the right constitutive terms from a large candidate library — recovering the governing law with full uncertainty quantification on each coefficient, enabling downstream reliability assessment.

Variational Bayes Sparse Regression Constitutive Law Reliability
Read Paper
σ = C₁ε + C₃ε³ Strain ε → Stress σ HYPERELASTIC STRESS-STRAIN ID
004
Blog

Deriving Actor-Critic Methods with Intuition

Starting from the REINFORCE policy gradient, this blog rigorously derives the full Actor-Critic algorithm step by step — reward-to-go, baseline subtraction (with no-bias and variance-reduction proofs), and the advantage function A(s,a). Each step is motivated by the problem it solves, so the reader builds genuine intuition rather than just following equations.

REINFORCE Policy Gradient Advantage Function Variance Reduction
Read Blog
ENV s, r ACTOR π(a|s;θ) CRITIC V(s;w) a(t) s(t) A(s,a) r(t) ∇J(θ) = 𝔼[A·∇log π] ACTOR-CRITIC POLICY GRADIENT