Omanshu Thapliyal

Projects

• Data-Driven Reachability We present a data-driven method for computing reachable sets for unknown nonlinear dynamical systems using a Koopman operator based approach. We find mixed-monotone decompositions for a class of Koopman lifted dynamics. The mixed-monotone system can be further embedded to a higher-dimensional dynamical model which is propagated in time deterministically. This allows us to find over-approximations of forward reachable sets that do not suffer from the curse of dimensionality. The proposed method can account for unknown nonlinear dynamics and allow a calculation of the conservative approximations of the reachable sets to a prefixed degree of accuracy in a computationally inexpensive manner. We demonstrate the efficacy of the proposed algorithm using an illustrative example with an unknown, nonlinear dynamical model, and compare it to CORA, a well-known existing method.
• Identifying mode confusion in aircraft-pilot interactions Most human-machine interaction systems suffer from the dreaded failure of automation surprise. This project aims at predicting automation surprises in safety critical systems such as aircraft-autopilot system; and then proposing possible resolution methods to get rid of the mode mismatch events thus identified. | Collaborators : Prof. Meeko Oishi, Vignesh Sivaramakrishnan
• State estimation subject to packet losses

  The project aimed at studying state estimation subject to random packet dropouts. Addressing different ways to incorporate this randomness led to two different subproblems:

  (i) packet losses are Markovian in nature: such packet losses are common in communication channels (such as ethernet).

  (ii) packet losses that occur within specific regions, where the regions are themselves distributed according to known priors (such as radar jammers).

  Systematic inclusion of known information about packet losses (Markovian, or state dependent distribution of jammers) leads to an improved accuracy in the baseline Kalman filter, with a comparable computation complexity. | Collaborator : J. S. Nandiganahalli

• Consensus based hybrid state estiamtion for networked control systems (NCS) [code] This project aimed at extending the idea of consensus based state estimation (such as consensus-Kalman filters) to hybrid state estimators (such as IMM, GPB-I, GPB-II, etc.). The key idea is to allow consensus at each agent of the NCS and include it in a manner that allows for time-changing topologies of the connecting network. The state estimator developed outperforms all consensus based estimators for hybrid systems (as of 2018). | Collaborators : Raj Deshmukh, Cheolhyeon Kwon

Publications

• Kalman filtering with state-dependent packet losses, IET Control Theory & Applications. O Thapliyal, JS Nandiganahalli, I Hwang
• Distributed state estimation for a stochastic linear hybrid system over a sensor network, IET Control Theory & Applications. R Deshmukh, O Thapliyal, C Kwon, I Hwang
• Approximate Reachability for Koopman Systems Using Mixed Monotonicity, IEEE Access. O Thapliyal, I Hwang

• Path Planning for a Network of Robots with Distributed Multi-Objective Linear Programming, American Control Conference, 2021. O Thapliyal, I Hwang
• Distributed Fast-Tracking Alternating Direction Method of Multipliers (ADMM) Algorithm with Optimal Convergence Rate, IEEE International Conference on Systems, Man, and Cybernetics (SMC), 2021. S Shethia, A Gupta, O Thapliyal, I Hwang
• Learning Based Cyberattack Design and Defense for Supervisory Control Systems, European Control Conference, 2021. O Thapliyal, I Hwang
• Predicting Mode Confusion Through Mixed Integer Linear Programming, IEEE 58th Annual Conference on Decision and Control, 2019. V Sivaramakrishnan, O Thapliyal, A Vinod, M Oishi, I Hwang
• Optimal state estimation for LTI systems with imperfect observations, IEEE 56th Annual Conference on Decision and Control, 2017. O Thapliyal, JS Nandiganahalli, I Hwang