arXiv subjects

Vijay Gupta

Publications and source records attributed to Vijay Gupta.

At least 37 records · Page 2Linked to original sources

arXiv2025

End-to-End Learning Framework for Solving Non-Markovian Optimal Control

Integer-order calculus often falls short in capturing the long-range dependencies and memory effects found in many real-world processes. Fractional calculus addresses these gaps via fractional-order integrals and derivatives, but fractional-order dynamical systems pose substantial challenges in system identification and optimal control due to the lack of standard control methodologies. In this paper, we theoretically derive the optimal control via linear quadratic regulator (LQR) for fractional-order linear time-invariant (FOLTI) systems and develop an end-to-end deep learning framework based on this theoretical foundation. Our approach establishes a rigorous mathematical model, derives analytical solutions, and incorporates deep learning to achieve data-driven optimal control of FOLTI systems. Our key contributions include: (i) proposing an innovative system identification method control strategy for FOLTI systems, (ii) developing the first end-to-end data-driven learning framework, Fractional-Order Learning for Optimal Control (FOLOC), that learns control policies from observed trajectories, and (iii) deriving a theoretical analysis of sample complexity to quantify the number of samples required for accurate optimal control in complex real-world problems. Experimental results indicate that our method accurately approximates fractional-order system behaviors without relying on Gaussian noise assumptions, pointing to promising avenues for advanced optimal control.

eess.SY↗

arXiv2025

Parameter-Free Federated TD Learning with Markov Noise in Heterogeneous Environments

Federated learning (FL) can dramatically speed up reinforcement learning by distributing exploration and training across multiple agents. It can guarantee an optimal convergence rate that scales linearly in the number of agents, i.e., a rate of $\tilde{O}(1/(NT)),$ where $T$ is the iteration index and $N$ is the number of agents. However, when the training samples arise from a Markov chain, existing results on TD learning achieving this rate require the algorithm to depend on unknown problem parameters. We close this gap by proposing a two-timescale Federated Temporal Difference (FTD) learning with Polyak-Ruppert averaging. Our method provably attains the optimal $\tilde{O}(1/NT)$ rate in both average-reward and discounted settings--offering a parameter-free FTD approach for Markovian data. Although our results are novel even in the single-agent setting, they apply to the more realistic and challenging scenario of FL with heterogeneous environments.

cs.LG↗

arXiv2025

Communication-aware Wide-Area Damping Control using Risk-Constrained Reinforcement Learning

Non-ideal communication links, especially delays, critically affect fast networked controls in power systems, such as the wide-area damping control (WADC). Traditionally, a delay estimation and compensation approach is adopted to address this cyber-physical coupling, but it demands very high accuracy for the fast WADC and cannot handle other cyber concerns like link failures or {cyber perturbations}. Hence, we propose a new risk-constrained framework that can target the communication delays, yet amenable to general uncertainty under the cyber-physical couplings. Our WADC model includes the synchronous generators (SGs), and also voltage source converters (VSCs) for additional damping capabilities. To mitigate uncertainty, a mean-variance risk constraint is introduced to the classical optimal control cost of the linear quadratic regulator (LQR). Unlike estimating delays, our approach can effectively mitigate large communication delays by improving the worst-case performance. A reinforcement learning (RL)-based algorithm, namely, stochastic gradient-descent with max-oracle (SGDmax), is developed to solve the risk-constrained problem. We further show its guaranteed convergence to stationarity at a high probability, even using the simple zero-order policy gradient (ZOPG). Numerical tests on the IEEE 68-bus system not only verify SGDmax's convergence and VSCs' damping capabilities, but also demonstrate that our approach outperforms conventional delay compensator-based methods under estimation error. While focusing on performance improvement under large delays, our proposed risk-constrained design can effectively mitigate the worst-case oscillations, making it equally effective for addressing other communication issues and cyber perturbations.

eess.SY↗

arXiv2025

The role of communication delays in the optimal control of spatially invariant systems

We study optimal proportional feedback controllers for spatially invariant systems when the controller has access to delayed state measurements received from different spatial locations. We analyze how delays affect the spatial locality of the optimal feedback gain leveraging the problem decoupling in the spatial frequency domain. For the cases of expensive control and small delay, we provide exact expressions of the optimal controllers in the limit for infinite control weight and vanishing delay, respectively. In the expensive control regime, the optimal feedback control law decomposes into a delay-aware filtering of the delayed state and the optimal controller in the delay-free setting. Under small delays, the optimal controller is a perturbation of the delay-free one which depends linearly on the delay. We illustrate our analytical findings with a reaction-diffusion process over the real line and a multi-agent system coupled through circulant matrices, showing that delays reduce the effectiveness of optimal feedback control and may require each subsystem within a distributed implementation to communicate with farther-away locations.

math.OC↗

arXiv2025

Learning Neural Koopman Operators with Dissipativity Guarantees

We address the problem of learning a neural Koopman operator model that provides dissipativity guarantees for an unknown nonlinear dynamical system that is known to be dissipative. We propose a two-stage approach. First, we learn an unconstrained neural Koopman model that closely approximates the system dynamics. Then, we minimally perturb the parameters to enforce strict dissipativity. Crucially, we establish theoretical guarantees that extend the dissipativity properties of the learned model back to the original nonlinear system. We realize this by deriving an exact relationship between the dissipativity of the learned model and the true system through careful characterization of the identification errors from the noisy data, Koopman operator truncation, and generalization to unseen data. We demonstrate our approach through simulation on a Duffing oscillator model.

eess.SY↗

arXiv2025

Power-Constrained Policy Gradient Methods for LQR

Consider a discrete-time Linear Quadratic Regulator (LQR) problem solved using policy gradient descent when the system matrices are unknown. The gradient is transmitted across a noisy channel over a finite time horizon using analog communication by a transmitter with an average power constraint. This is a simple setup at the intersection of reinforcement learning and networked control systems. We first consider a communication-constrained optimization framework, where gradient descent is applied to optimize a non-convex function under noisy gradient transmission. We provide an optimal power allocation algorithm that minimizes an upper bound on the expected optimality error at the final iteration and show that adaptive power allocation can lead to better convergence rate as compared to standard gradient descent with uniform power distribution. We then apply our results to the LQR setting.

math.OC↗

arXiv2025

Leveraging Offline Data from Similar Systems for Online Linear Quadratic Control

``Sim2real gap", in which the system learned in simulations is not the exact representation of the real system, can lead to loss of stability and performance when controllers learned using data from the simulated system are used on the real system. In this work, we address this challenge in the linear quadratic regulator (LQR) setting. Specifically, we consider an LQR problem for a system with unknown system matrices. Along with the state-action pairs from the system to be controlled, a trajectory of length $S$ of state-action pairs from a different unknown system is available. Our proposed algorithm is constructed upon Thompson sampling and utilizes the mean as well as the uncertainty of the dynamics of the system from which the trajectory of length $S$ is obtained. We establish that the algorithm achieves $\tilde{\mathcal{O}}({f(S,M_δ)\sqrt{T/S}})$ Bayes regret after $T$ time steps, where $M_δ$ characterizes the \emph{dissimilarity} between the two systems and $f(S,M_δ)$ is a function of $S$ and $M_δ$. When $M_δ$ is sufficiently small, the proposed algorithm achieves $\tilde{\mathcal{O}}({\sqrt{T/S}})$ Bayes regret and outperforms a naive strategy which does not utilize the available trajectory.

eess.SY↗

arXiv2025

Online Model Order Reduction of Linear Systems via $(γ,δ)$-Similarity

Model order reduction aims to determine a low-order approximation of high-order models with least possible approximation errors. For application to physical systems, it is crucial that the reduced order model (ROM) is robust to any disturbance that acts on the full order model (FOM) -- in the sense that the output of the ROM remains a good approximation of that of the FOM, even in the presence of such disturbances. In this work, we present a framework for online model order reduction for a class of continuous-time linear systems that ensures this property for any $\mathcal{L}_2$ disturbance. Apart from robustness to disturbances in this sense, the proposed framework also displays other desirable properties for model order reduction: (1) a provable bound on the error defined as the $L_2$ norm of the difference between the output of the ROM and FOM, (2) preservation of stability, (3) compositionality properties and a provable error bound for arbitrary interconnected systems, (4) a provable bound on the output of the FOM when the controller designed for the ROM is used with the FOM, and finally, (5) compatibility with existing approaches such as balanced truncation and moment matching. Property (4) does not require computation of any gap metric and property (5) is beneficial as existing approaches can also be equipped with some of the preceding properties. The theoretical results are corroborated on numerical case studies, including on a building model.

eess.SY↗

arXiv2025

A Sampling Complexity-aware Framework for Discrete-time Fractional-Order Dynamical System Identification

A variety of complex biological, natural and man-made systems exhibit non-Markovian dynamics that can be modeled through fractional order differential equations, yet, we lack sample comlexity aware system identification strategies. Towards this end, we propose an affine discrete-time fractional order dynamical system (FoDS) identification algorithm and provide a detailed sample complexity analysis. The algorithm effectively addresses the challenges of FoDS identification in the presence of noisy data. The proposed algorithm consists of two key steps. Firstly, it avoids solving higher-order polynomial equations, which would otherwise result in multiple potential solutions for the fractional orders. Secondly, the identification problem is reformulated as a least squares estimation, allowing us to infer the system parameters. We derive the expectation and probabilistic bounds for the FoDS parameter estimation error, assuming prior knowledge of the functions $ f $ and $ g $ in the FoDS model. The error decays at a rate of $ N = O\left( \frac{d}ε \right) $, where $ N $ is the number of samples, $ d $ is the dimension of the state variable, and $ ε$ represents the desired estimation accuracy. Simulation results demonstrate that our theoretical bounds are tight, validating the accuracy and robustness of this algorithm.

eess.SY↗

arXiv2024

Model-Free Learning for the Linear Quadratic Regulator over Rate-Limited Channels

Consider a linear quadratic regulator (LQR) problem being solved in a model-free manner using the policy gradient approach. If the gradient of the quadratic cost is being transmitted across a rate-limited channel, both the convergence and the rate of convergence of the resulting controller may be affected by the bit-rate permitted by the channel. We first pose this problem in a communication-constrained optimization framework and propose a new adaptive quantization algorithm titled Adaptively Quantized Gradient Descent (AQGD). This algorithm guarantees exponentially fast convergence to the globally optimal policy, with no deterioration of the exponent relative to the unquantized setting, above a certain finite threshold bit-rate allowed by the communication channel. We then propose a variant of AQGD that provides similar performance guarantees when applied to solve the model-free LQR problem. Our approach reveals the benefits of adaptive quantization in preserving fast linear convergence rates, and, as such, may be of independent interest to the literature on compressed optimization. Our work also marks a first step towards a more general bridge between the fields of model-free control design and networked control systems.

math.OC↗

arXiv2024

A Scalable Game Theoretic Approach for Coordination of Multiple Dynamic Systems

Learning in games provides a powerful framework to design control policies for self-interested agents that may be coupled through their dynamics, costs, or constraints. We consider the case where the dynamics of the coupled system can be modeled as a Markov potential game. In this case, distributed learning by the agents ensures that their control policies converge to a Nash equilibrium of this game. However, typical learning algorithms such as natural policy gradient require knowledge of the entire global state and actions of all the other agents, and may not be scalable as the number of agents grows. We show that by limiting the information flow to a local neighborhood of agents in the natural policy gradient algorithm, we can converge to a neighborhood of optimal policies. If the game can be designed through decomposing a global cost function of interest to a designer into local costs for the agents such that their policies at equilibrium optimize the global cost, this approach can be of interest to team coordination problems as well. We illustrate our approach through a sensor coverage problem.

eess.SY↗

arXiv2024

Online Actuator Selection and Controller Design for Linear Quadratic Regulation with Unknown System Model

We study the simultaneous actuator selection and controller design problem for linear quadratic regulation with Gaussian noise over a finite horizon of length $T$ and unknown system model. We consider both episodic and non-episodic settings of the problem and propose online algorithms that specify both the sets of actuators to be utilized under a cardinality constraint and the controls corresponding to the sets of selected actuators. In the episodic setting, the interaction with the system breaks into $N$ episodes, each of which restarts from a given initial condition and has length $T$. In the non-episodic setting, the interaction goes on continuously. Our online algorithms leverage a multiarmed bandit algorithm to select the sets of actuators and a certainty equivalence approach to design the corresponding controls. We show that our online algorithms yield $\sqrt{N}$-regret for the episodic setting and $T^{2/3}$-regret for the non-episodic setting. We extend our algorithm design and analysis to show scalability with respect to both the total number of candidate actuators and the cardinality constraint. We numerically validate our theoretical results.

math.OC↗

arXiv2024

Online Mixed Discrete and Continuous Optimization: Algorithms, Regret Analysis and Applications

We study an online mixed discrete and continuous optimization problem where a decision maker interacts with an unknown environment for a number of $T$ rounds. At each round, the decision maker needs to first jointly choose a discrete and a continuous actions and then receives a reward associated with the chosen actions. The goal for the decision maker is to maximize the accumulative reward after $T$ rounds. We propose algorithms to solve the online mixed discrete and continuous optimization problem and prove that the algorithms yield sublinear regret in $T$. We show that a wide range of applications in practice fit into the framework of the online mixed discrete and continuous optimization problem, and apply the proposed algorithms to solve these applications with regret guarantees. We validate our theoretical results with numerical experiments.

math.OC↗

arXiv2024

Learning Decentralized Linear Quadratic Regulators with $\sqrt{T}$ Regret

We propose an online learning algorithm that adaptively designs a decentralized linear quadratic regulator when the system model is unknown a priori and new data samples from a single system trajectory become progressively available. The algorithm uses a disturbance-feedback representation of state-feedback controllers coupled with online convex optimization with memory and delayed feedback. Under the assumption that the system is stable or given a known stabilizing controller, we show that our controller enjoys an expected regret that scales as $\sqrt{T}$ with the time horizon $T$ for the case of partially nested information pattern. For more general information patterns, the optimal controller is unknown even if the system model is known. In this case, the regret of our controller is shown with respect to a linear sub-optimal controller. We validate our theoretical findings using numerical experiments.

math.OC↗

arXiv2024

New Exponential operators connected with a^2+x^2: a generalization of Post-Widder and Ismail May

The present study offers a general exponential operator connected with a^2+x^2; for positive real "a". We estimate the asymptotic formula for simultaneous and ordinary approximation of the constructed operator. In the last section, we graphically interpret the created operator's convergence to two periodic functions "x sin(x)" and "-x/2*cos(pi*x)". We also consider the limiting case a tends to 0; which provides Post-Widder operator. In addition, we analyze each particular case of the defined operator and determine the optimal value of "a", that would yield the greatest approximation; this facilitates us to contrast the well-known operators existing in the literature, especially the Post-Widder operator and the operator due to Ismail-May.

math.FA↗

arXiv2024

Sampling-based Safe Reinforcement Learning for Nonlinear Dynamical Systems

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.

cs.LG↗

arXiv2024

Rationality of Learning Algorithms in Repeated Normal-Form Games

Many learning algorithms are known to converge to an equilibrium for specific classes of games if the same learning algorithm is adopted by all agents. However, when the agents are self-interested, a natural question is whether agents have a strong incentive to adopt an alternative learning algorithm that yields them greater individual utility. We capture such incentives as an algorithm's rationality ratio, which is the ratio of the highest payoff an agent can obtain by deviating from a learning algorithm to its payoff from following it. We define a learning algorithm to be $c$-rational if its rationality ratio is at most $c$ irrespective of the game. We first establish that popular learning algorithms such as fictitious play and regret matching are not $c$-rational for any constant $c\geq 1$. We then propose and analyze two algorithms that are provably $1$-rational under mild assumptions, and have the same properties as (a generalized version of) fictitious play and regret matching, respectively, if all agents follow them. Finally, we show that if an assumption of perfect monitoring is not satisfied, there are games for which $c$-rational algorithms do not exist, and illustrate our results with numerical case studies.

cs.GT↗

arXiv2023

Finite-Time Error Bounds for Distributed Linear Stochastic Approximation

This paper considers a novel multi-agent linear stochastic approximation algorithm driven by Markovian noise and general consensus-type interaction, in which each agent evolves according to its local stochastic approximation process which depends on the information from its neighbors. The interconnection structure among the agents is described by a time-varying directed graph. While the convergence of consensus-based stochastic approximation algorithms when the interconnection among the agents is described by doubly stochastic matrices (at least in expectation) has been studied, less is known about the case when the interconnection matrix is simply stochastic. For any uniformly strongly connected graph sequences whose associated interaction matrices are stochastic, the paper derives finite-time bounds on the mean-square error, defined as the deviation of the output of the algorithm from the unique equilibrium point of the associated ordinary differential equation. For the case of interconnection matrices being stochastic, the equilibrium point can be any unspecified convex combination of the local equilibria of all the agents in the absence of communication. Both the cases with constant and time-varying step-sizes are considered. In the case when the convex combination is required to be a straight average and interaction between any pair of neighboring agents may be uni-directional, so that doubly stochastic matrices cannot be implemented in a distributed manner, the paper proposes a push-sum-type distributed stochastic approximation algorithm and provides its finite-time bound for the time-varying step-size case by leveraging the analysis for the consensus-type algorithm with stochastic matrices and developing novel properties of the push-sum algorithm. Distributed temporal difference learning is discussed as an illustrative application.

cs.LG↗