Data-driven Control Research Articles

Periodic event-triggered data-driven control for networked control systems with time-varying delays

This article focuses on data-driven analysis and controller design for networked control systems (NCSs) with network-induced delays. The study considers a linear time-invariant (LTI) system controlled through a periodic event-triggering mechanism. First, by leveraging data-based representations, we establish data-based stability conditions for NCSs with time-varying delays. Furthermore, we propose the data-based method for co-designing the controller and the periodic event-triggering scheme. In addition, we present novel data-based conditions for verifying dissipativity properties of NCSs. The effectiveness of our proposed methods is validated through a simulation and a turbofan engine hardware-in-the-loop (HIL) experiment.

Data-Driven Control of Unknown Switched Linear Systems Using Scenario Optimization

Data-Driven Control of Unknown Switched Linear Systems Using Scenario Optimization

An Overview of Systems-Theoretic Guarantees in Data-Driven Model Predictive Control

The development of control methods based on data has seen a surge of interest in recent years. When applying data-driven controllers in real-world applications, providing theoretical guarantees for the closed-loop system is of crucial importance to ensure reliable operation. In this review, we provide an overview of data-driven model predictive control (MPC) methods for controlling unknown systems with guarantees on systems-theoretic properties such as stability, robustness, and constraint satisfaction. The considered approaches rely on the fundamental lemma from behavioral theory in order to predict input–output trajectories directly from data. We cover various setups, ranging from linear systems and noise-free data to more realistic formulations with noise and nonlinearities, and we provide an overview of different techniques to ensure guarantees for the closed-loop system. Moreover, we discuss avenues for future research that may further improve the theoretical understanding and practical applicability of data-driven MPC.

On-line data-driven control for uncertain systems based on greedy algorithm

Considering a result that persistently exciting data can be used to replace the linear system model, this paper is devoted to applying this result in the field of data-driven control of nonlinear systems. An on-line iteration based on greedy algorithm to stabilize uncertain discrete-time systems is proposed. The method tends to obtain approximate optimal control through solving a series of programming problems. Every programming problem is linear for the convenience of solving. Besides, in particular, the method requires few prior conditions, as long as the system is controllable and observable and the equilibrium state of the system is known. First, we prove that under certain circumstances, the solution to our linear matrix inequality can stabilize the system. Next, a multi-objective programming problem is proposed to deal with situations where the required conditions are unknown. Finally, an on-line iteration is used to enhance robustness as well as real-time evaluation. The method is illustrated to be effective through a simulation under repeated experiments.

Hand It to Me Formally! Data-Driven Control for Human-Robot Handovers With Signal Temporal Logic

Hand It to Me Formally! Data-Driven Control for Human-Robot Handovers With Signal Temporal Logic

A robust data-driven model predictive thermal control for rack-based data center

Optimizing and controlling of air-cooled data centers cooling systems are essential for reducing energy consumption. However, traditional cooling system control strategies have been proven inadequate due to the thermodynamic processess complexity and the challenge for accuratly modeling the thermal behavior of data center (DC). Therefore, a data-driven control method whithout explicit modeling process is urgently needed. In this paper, a robust data-driven model predictive control (MPC) method for air-cooled rack-based DCs, based on Willems fundamental lemma, is proposed. The methodology is executed in three phases: Firstly, we show that the system is approximately linear time-invariant (LTI) by zonal model method and experiment. Secondly, we propose a data-driven MPC framework for air-cooled rack-based DC. Thirdly, considering the uncertainties caused by disturbances and the nonlinear dynamics, the data-driven MPC approach is tailored as a robust one. Finally, the proposed approach is verified by computational fluid dynamics simulation experiments. This approach ensures the rack-based DC cold aisle temperature to the reference temperature as well as within acceptable limits. The experiment shows that the steady-state error is guaranteed to be less than 2%, with a control accuracy increased by 2.4% compared with the regular MPC. Thus, this method provides an alternative solution for maintaining the reference temperature in air-cooled rack-based DCs by utilizing only input–output data.

Data-Driven Control of COVID-19 in Buildings: A Reinforcement-Learning Approach

Data-Driven Control of COVID-19 in Buildings: A Reinforcement-Learning Approach

Distributed Data-Driven Control for a Connected Autonomous Vehicle Platoon Subjected to False Data Injection Attacks

Distributed Data-Driven Control for a Connected Autonomous Vehicle Platoon Subjected to False Data Injection Attacks

Data-Driven Control for Nonlinear Networked Systems on Basis of Two Description Coding Scheme

Data-Driven Control for Nonlinear Networked Systems on Basis of Two Description Coding Scheme

Anti Wind‐Up and Robust Data‐Driven Model‐Free Adaptive Control for MIMO Nonlinear Discrete‐Time Systems

ABSTRACTThis article addresses a solution to one of the main challenges of online data‐driven control (DDC) methods: reducing the sensitivity of the model‐free adaptive control (MFAC) method to initial conditions and control parameters with the new control cost function and added the output error rate and integral along with a new anti‐wind up strategy for multi‐input multi‐output (MIMO) systems. The parameters introduced to the new control law have been validated using the boundary‐input boundary‐output (BIBO) approach to design and converge the controller. The simulation findings on a nonlinear auto‐regressive moving average model with exogenous inputs (NARMAX) system with triangular control input demonstrate that the proposed control rule will outperform to prototype MFAC. Furthermore, to analyze the sensitivity of the controller to the initial conditions and the uncertainties of the control parameters, 30 Monte Carlo simulations were performed with random initial conditions in the presence of disturbance in the control input, and output noise, and the results were compared with the prototype MFAC and conventional PID controller using standard criteria such as integral time absolute error, standard deviation, steady‐state error, and mean maximum error, which shows a noticeable superiority of proposed controller relative to the prototype MFAC.

Data Reconstruction Using Smart Sensor Placement.

This paper deals with spatio-temporal field estimation with efficient sensor placement based on the QR decomposition. The proposed method also identifies the optimal number of sensors needed for field estimation that captures the most relevant features of the field of interest. To address the uncertainties inherent in spatio-temporal field estimation, a robust data-driven control method is utilized, providing resilience against unpredictable environmental and model changes. In particular, the approach uses the Kriged Kalman Filter (KKF) for uncertainty-aware field reconstruction. Unlike other reconstruction methods, the positional uncertainty originating from the data acquisition platform is integrated into the KKF estimator. Numerical results are presented to show the efficacy of the proposed dynamic sensor placement strategy together with the KKF field estimator.

Control of combustion phasing using accelerometer-based non-intrusive sensing

Abstract Measuring the combustion phasing of an engine using in-cylinder pressure sensors is well established. However, pressure sensors need to be directly exposed to the in-cylinder environment, requiring changes to the cylinder head. Several methods have been proposed for sensing combustion phasing non-intrusively by mounting an accelerometer on the engine block. This paper presents real-time control of combustion phasing in a compression-ignition engine using non-intrusive accelerometer-based sensing during a dynamic fuel switch. A systematic data-driven control framework capable of handling fuel switching in real time is used. The control is designed based on the pressure data, and the real-time implementation is performed using the accelerometer signal. Results from combustion phasing tracking experiments performed on a compression-ignition engine are presented.

Data-Driven Control Design for Power Converters Approximated as Switched Affine Systems and Experimental Validation

Data-Driven Control Design for Power Converters Approximated as Switched Affine Systems and Experimental Validation

Moment-Based Reinforcement Learning for Ensemble Control.

Problems involving controlling the collective behavior of a population of structurally similar dynamical systems, the so-called ensemble control, arise in diverse emerging applications and pose a grand challenge in systems science and control engineering. Owing to the severely under-actuated nature and the difficulty of placing large-scale sensor networks, ensemble systems are limited to being actuated and monitored at the population level. Moreover, mathematical models describing the dynamics of ensemble systems are often elusive. Therefore, it is essential to design broadcast controls that excite the entire population in such a way that the heterogeneity in system dynamics is robustly compensated. In this article, we propose a reinforcement learning (RL)-based data-driven control framework incorporating population-level aggregated measurement data to learn a global control signal for steering a dynamic population in the desired manner. In particular, we introduce the notion of ensemble moments induced by aggregated measurements and derive the associated moment system to the original ensemble system. Then, using the moment system, we learn an approximation of optimal value functions and the associated policies in terms of ensemble moments through RL. We illustrate the feasibility and scalability of the proposed moment-based approach via numerical experiments using a population of linear, bilinear, and nonlinear dynamic ensemble systems. We report that the proposed method achieves the desired control objectives of various ensemble control tasks and obtains significantly better averaged-reward when compared with three existing methods.

Utilizing a optimized method for evaluating vapor recovery equipment control efficiency and estimating evaporative VOC emissions from urban oil depots via an extensive survey

As an important intermediary between upstream refineries and downstream urban gas stations, volatile organic compound (VOC) emissions from urban oil depots were often disregarded, underestimating their environmental and health implications. An extensive investigation of urban depots' fuel composition and operational dynamics was conducted nationwide. We developed a novel approach that integrates theoretical models with easily measurable operational data from the depots to evaluate the efficiency of post-treatment devices in actual situations. Even in well-managed oil depots, the actual control efficiency of vapor recovery units fluctuates between 63 % and 85 %, depending on the concentration of hydrocarbon vapors in the intake of the device. The national emission factors for gasoline, diesel, and aviation kerosene at a national level were 6.64 ± 1.16, 2.07 ± 0.42, and 6.17 ± 1.05 tons per 10,000 tons, respectively. In 2019, China's urban oil depots emitted 165 thousand tons of VOC. Enhancing control strategies by optimizing the physical and chemical parameters of refined oil, improving storage capacity and turnover efficiency, and upgrading storage tanks had the potential to reduce emissions by more than 60 %. However, a 30 % failure rate in these systems could negate the benefits of these improved strategies.

Multi-sensor fault detection and correction for automated IAQ monitoring in smart buildings through attention-aware autoencoders with spatial prediction module

The integration of interconnected monitoring and control devices has significantly enhanced the management of indoor building environments, notably in underground subway stations. In these sophisticated settings, the accuracy of monitoring systems is critical, as data-driven control systems rely extensively on sensor feedback. However, sensor faults arising from hardware damage, wear, and cyberattacks compromise the reliability of these complex systems. Despite extensive efforts, existing methods still face challenges in accurately detecting and reconstructing faults, particularly when dealing with multiple sensor failures. This study introduces a hybrid sensor validation framework that combines a multi-head attention-based denoising autoencoder (AE) with a deep learning prediction module. The performance assessment of the proposed gated recurrent unit (GRU) integrated attention-based fault detection, and reconstruction network (GADRN) shows a significant 33 % and 21 % improvement in critical success index compared to Independent Component Analysis and GRU-AE frameworks. GADRN consistently identified faulty sensors across various fault periods, achieving the lowest false alarm rate of 11 % against several conventional and machine learning-based models. A ventilation control analysis highlights the significance of incorporating a sensor validation framework in modern buildings. The outputs by the GADRN led to a 22.5 % decrease in energy consumption by preventing unnecessary resource use attributed to sensor faults.

Data-driven control of wind turbine under online power strategy via deep learning and reinforcement learning

This study proposes a data-driven wind turbine (WT) model predictive control (MPC) enhanced by a deep-learning (DL) radial basis function network (RBFN) and a reinforcement-learning (RL) deep Q-learning network (DQN). The RBFN provides comprehensive aerodynamic predictions, including thrust, torque, and power. Besides, the MPC linearization relies on the RBFN prediction to estimate force sensitivities. The DQN achieves an online power strategy (OPS) that solves the 2-degree-of-freedom (2-DOF) optimization of rotor speed and pitch angle, which can actively adjust power capture to meet different power requirements. The DQN adopts a novel bisection algorithm with a first-in-first-out (FIFO) queue for high-precision 2-DOF results. The MPC coordinates the permanent magnet synchronous generator (PMSG) and pitch servo, considering shaft rotation and tower movement. Compared with the maximum power point tracking (MPPT) and power reference point tracking (PRPT) based controls, the proposed RBFN-DQN-MPC reduces power fluctuation and ensures constant output. This study also compares the DQN with the categorical DQN (C51), which indicates that the DQN is more effective in the 2-DOF optimization. Hence, WTs enhanced by the DL-RL-MPC are intelligent and reliable for flexible wind generation.

Model-free stabilization via Extremum Seeking using a cost neural estimator

In this paper, a fully model-free architecture for vertical stabilization of thermonuclear plasmas in tokamak experimental reactors is presented. For the first time, an Extremum Seeking control algorithm is combined with neural networks to estimate the Lyapunov function to be minimized, resulting in a fully data-driven control architecture. The performance of different neural networks are compared. Specifically, Multilayer Perceptrons and Extreme Learning Machines are considered. The proposed architecture is tested in simulation to show that it can counteract relevant plasma disturbances, resulting in a significant improvement in terms of the achievable operative space compared to the Extremum Seeking algorithm, which still relies on model-based cost estimator.

Thermal analysis of fuel cells in renewable energy systems using Generative Adversarial Networks (GANs) and Reinforcement learning

This research introduces a novel in silico thermal management strategy for fuel cell systems that combines Generative Adversarial Networks (GANs) and Reinforcement Learning (RL). The framework utilises the unconditional generation ability of GANs to produce realistic thermal management, which can be used to train an RL agent to control system parameters dynamically. A GAN can be trained on historical data or simulation results to generate realistic and diverse scenarios of the complex relationship between the working point and the thermodynamics of a fuel cell. Once the adversarial network is trained to output realistic data, it can feed an RL agent that tunes the parameters influencing cooling, such as coolant flow rates, air stoichiometry or reactant humidification in response to a generated working point. We conduct experiments and simulations to show that GANs can improve the performance and resilience of RL agents to uncertainty. This approach can be applied to a real-world system, demonstrating significant thermal efficiency, durability and robustness improvements. Adopting such a strategy for thermal management in a fuel cell system is just one example of a new class of intelligent, data-driven control approaches that will enable improved understanding and behaviour of systems, leading to better performance and reliability.

Robust output group formation tracking control of heterogeneous multi-agent systems with multiple leaders using reinforcement learning

This paper studies the distributed output formation tracking problem of grouped heterogeneous multi-agent systems under multiple leaders and uncertainties using reinforcement learning (RL). The outputs of followers are supposed to achieve robust tracking to the respective convex point of group leaders while generating an expected time-varying formation configuration. First, a distributed adaptive observer is designed under a directed graph to coordinate the multiple group leaders while estimating the leaders’ dynamics in finite-time. The adaptive mechanism avoids global information of the graph. Second, an optimal tracking problem with respect to the observer is formulated for each follower, while the feedback tracking controller is derived using an action-dependent RL algorithm. An extended learning process for essential dynamics is constructed using the same data, while the output regulation equations are solved equivalently. Third, the robust formation controller and feasibility condition are further proposed based on previous learning results. Stability of the synthetical data-driven controller is analyzed under internal uncertainties and external disturbances. Finally, simulation results are provided to demonstrate the effectiveness of the hierarchical control framework.