|
Siddharth Prabhu, Srinivas Rangarajan, and Mayuresh Kothare. Bi-level optimization for parameter estimation of differential equations using interpolation, 2025. Paper. Code
- Designed a bilevel optimization framework for parameter estimation that leverages the convexity of linear parameters in dynamical systems.
- Demonstrated superior performance of the proposed method over shooting-based techniques across multiple complex dynamical systems.
|
|
Siddharth Prabhu, Sulman Haque, Dan Gurr, Loren Coley, Jim Beilstein, Srinivas Rangarajan, and Mayuresh Kothare. An event-based neural partial differential equation model of heat and mass transport in an industrial drying oven. Computers & Chemical Engineering, page 109171, 2025. Paper. Code.
- Developed a fully differentiable physics-based simulator for Universal Partial Differential Equations (UPDE) to model noisy, partially observed dynamical systems with events.
- Demonstrated that incorporating physics-based priors enables effective learning from partially observed states and improves generalization to unobserved regions.
|
|
Siddharth Prabhu, Srinivas Rangarajan, and Mayuresh Kothare. A condensing approach to multiple shooting neural ordinary differential equation, 2025. Paper. Code.
- Developed an optimization method to incorporate equality constraints from multiple shooting into the training of Neural Ordinary Differential Equation (NODE).
- Demonstrated that the proposed approach achieves better performance on highly nonlinear, complex and oscillatory dynamics using fewer parameters compared to single shooting NODE.
|
|
Siddharth Prabhu, Srinivas Rangarajan and Mayuresh Kothare, "Differential Dynamic Programming with Stagewise Equality and Inequality Constraints Using Interior Point Method," 2025 American Control Conference (ACC), Denver, CO, USA, 2025, pp. 2255-2261, doi: 10.23919/ACC63710.2025.11108083. Paper. Code.
- Developed and Interior Point Differential Dynamic Programming algorithm to incorporate stagewise arbitrary equality and inequality constraints in an optimal control problem.
- The algorithm outperforms nonlinear programming based methods on several benchmark problems such as car parking, obstacle avoidance, quadcopter, robot arm, quadcopter, continuously stirred tank reactor.
|
|
Siddharth Prabhu, Nick Kosir, Mayuresh Kothare, and Srinivas Rangarajan. Derivative-free domain-informed data-driven discovery of sparse kinetic models. Industrial & Engineering Chemistry Research, 2025. Paper. Code.
- Developed an integration-based technique for sparse identification of dynamical equations compatible with domain knowledge such as mass balance and chemical constraints.
- Demonstrated that the proposed method achieves lower error rates and accurately recovers kinetic models for complex reaction networks across varying noise levels, sampling frequencies, and numbers of experiments.
|
|
Siddharth Prabhu, Srinivas Rangarajan, and Mayuresh Kothare. Data-driven discovery of sparse dynamical model of cardiovascular system for model predictive control. Computers in biology and medicine, 166:107513, 2023. Paper.
- Developed a data-driven, physics-constrained symbolic reduced-order model for model predictive control (MPC) of the cardiovascular system.
|
|
Ou Yang, Siddharth Prabhu, and Marianthi Ierapetritou. Comparison between Batch and Continuous Monoclonal Antibody Production and Economic Analysis. Industrial & Engineering Chemistry Research, 2019. Paper.
|
