PSALM Lab

PSALM (Predictive Stochastic Adaptive Learning Machines) Lab at IIT Kharagpur is focused on research at the intersection of control theory and Artificial Intelligence (AI). The real world applications of AI require safety guarantee and explainability of the AI module. In addition, several modules such as sensing and perception, planning and action need to interact to complete the feedback loop. Our main focus is in planning and decision making but we also consider sensing and actuation along with the effects of communication channels whenever required.

Deep MPC

This paper presents a deep learning-based model predictive control algorithm for control affine nonlinear discrete-time systems with matched and bounded state-dependent uncertainties of unknown structure. Since the structure of uncertainties is not known, a deep neural network (DNN) is employed to approximate them. In order to avoid any unwanted behavior during the learning phase, a tube-based nonlinear model predictive controller is employed, which ensures satisfaction of constraints and input-to-state stability of the closed-loop states.

Algorithmic contributions

• We have developed several computationally tractable classes of policies.
• A moving horizon estimator developed by us exploits control-estimation duality.
• Our current focus is on theoretical deevelopments based on physics informed neural networks, neural operators, polynomial chaos theory.