Krishna Chaitanya Kosaraju

Krishna Chaitanya Kosaraju, Michele Cucuzzella, Jacquelien M. A. Scherpen et al.

This paper deals with a class of Resistive-Inductive-Capacitive (RLC) circuits and switched RLC (s-RLC) circuits modeled in Brayton Moser framework. For this class of systems, new passivity properties using a Krasovskii's type Lyapunov function as storage function are presented. Consequently, the supply-rate is a function of the system states, inputs and their first time-derivatives. Moreover, after showing the integrability property of the port-variables, two simple control methodologies called output shaping and input shaping are proposed for regulating the voltage in RLC and s-RLC circuits. Global asymptotic convergence to the desired operating point is theoretically proved for both proposed control methodologies. Moreover, robustness with respect to load uncertainty is ensured by the input shaping methodology. The applicability of the proposed methodologies is illustrated by designing voltage controllers for DC-DC converters and DC networks.

Michele Cucuzzella, Riccardo Lazzari, Yu Kawano et al.

This work deals with the design of a robust and decentralized passivity-based control scheme for regulating the voltage of a DC microgrid through boost converters. A Krasovskii-type storage function is proposed and a (local) passivity property for DC microgrids comprising unknown 'ZIP' (constant impedance 'Z', constant current 'I' and constant power 'P') loads is established. More precisely, the input port-variable of the corresponding passive map is equal to the first-time derivative of the control input. Then, the integrated input port-variable is used to shape the closed loop storage function such that it has a minimum at the desired equilibrium point. Convergence to the desired equilibrium is theoretically analyzed and the proposed control scheme is validated through experiments on a real DC microgrid.

Krishna Chaitanya Kosaraju, Yu Kawano, Jacquelien M. A. Scherpen

In this paper we introduce a new notion of passivity which we call Krasovskii's passivity and provide a sufficient condition for a system to be Krasovskii's passive. Based on this condition, we investigate classes of port-Hamiltonian and gradient systems which are Krasovskii's passive. Moreover, we provide a new interconnection based control technique based on Krasovskii's passivity. Our proposed control technique can be used even in the case when it is not clear how to construct the standard passivity based controller, which is demonstrated by examples of a Boost converter and a parallel RLC circuit.

Krishna Chaitanya Kosaraju, Shravan Mohan, Ramkrishna Pasumarthy

The aim of this paper is to study the convergence of the primal-dual dynamics pertaining to Support Vector Machines (SVM). The optimization routine, used for determining an SVM for classification, is first formulated as a dynamical system. The dynamical system is constructed such that its equilibrium point is the solution to the SVM optimization problem. It is then shown, using passivity theory, that the dynamical system is global asymptotically stable. In other words, the dynamical system converges onto the optimal solution asymptotically, irrespective of the initial condition. Simulations and computations are provided for corroboration.

Krishna Chaitanya Kosaraju, Venkatesh Chinde, Ramkrishna Pasumarthy et al.

In this paper, we derive new passive maps akin to incremental passive maps, for a class of nonlinear systems using dynamic feedback and Krasovskii's method. Further using the passive maps we present a control methodology for stabilization to a desired operating point. This work is illustrated by designing a controller for a nonlinear building heating ventilating and air conditioning (HVAC) subsystem.

Krishna Chaitanya Kosaraju, Arun D. Mahindrakar, Vijay Muralidharan et al.

In this paper we present a geometric control law for position and line-of-sight stabilization of the nonholonomic spherical robot actuated by three independent actuators. A simple configuration error function with an appropriately defined transport map is proposed to extract feedforward and proportional-derivative control law. Simulations are provided to validate the controller performance.

Krishna Chaitanya Kosaraju

In the recent years, passivity theory has gained renewed attention because of its advantages and practicality in modeling of multi-domain systems and constructive control techniques. Unlike Lyapunov theory, passivity theory takes a behavioral approach in its control design methodologies. Hence, it provides solutions, which not only achieve the control objectives, but are also easily interpretable in the standard engineering parlance. The fundamental idea in passivity based control (PBC) methodologies is to find a controller that renders the closed-loop system passive. It is well known that, the PBC methodologies that rely on power-conjugate port-variables do not work for control objectives that require bounded power and unbounded energy. This is commonly known as the dissipation obstacle. One possible alternative that has been well explored, in the case of finite dimensional systems, is Brayton-Moser formulation. However, designing controllers in this framework leads to various difficulties, such as, solving for partial differential equations and finding storage functions satisfying a gradient structure.

Wesley A. Suttle, Vipul K. Sharma, Krishna C. Kosaraju et al.

We develop provably safe and convergent reinforcement learning (RL) algorithms for control of nonlinear dynamical systems, bridging the gap between the hard safety guarantees of control theory and the convergence guarantees of RL theory. Recent advances at the intersection of control and RL follow a two-stage, safety filter approach to enforcing hard safety constraints: model-free RL is used to learn a potentially unsafe controller, whose actions are projected onto safe sets prescribed, for example, by a control barrier function. Though safe, such approaches lose any convergence guarantees enjoyed by the underlying RL methods. In this paper, we develop a single-stage, sampling-based approach to hard constraint satisfaction that learns RL controllers enjoying classical convergence guarantees while satisfying hard safety constraints throughout training and deployment. We validate the efficacy of our approach in simulation, including safe control of a quadcopter in a challenging obstacle avoidance problem, and demonstrate that it outperforms existing benchmarks.

Martin Figura, Krishna Chaitanya Kosaraju, Vijay Gupta

Recently, many cooperative distributed multi-agent reinforcement learning (MARL) algorithms have been proposed in the literature. In this work, we study the effect of adversarial attacks on a network that employs a consensus-based MARL algorithm. We show that an adversarial agent can persuade all the other agents in the network to implement policies that optimize an objective that it desires. In this sense, the standard consensus-based MARL algorithms are fragile to attacks.