This paper introduces a novel framework that bridges advanced reinforcement learning (RL) with traditional PID control by converting model-based RL policies into interpretable PID gains. By combining inverse reinforcement learning (IRL) with Kullback–Leibler divergence minimization, our method aligns sophisticated control strategies with the simplicity and robustness of PID controllers. In doing so, the proposed approach maintains the transparency and simplicity of PID controllers while incorporating the adaptability, data-driven optimization, and long-horizon planning capabilities of RL. Compatible with both model-based and model-free RL algorithms, the approach has been validated through extensive simulations on benchmark systems and real-world experiments on the Robotarium platform, demonstrating resilience against disturbances, parameter uncertainties, and noise. By blending the strengths of reinforcement learning with the practical familiarity of PID control, the proposed framework offers a data-efficient, scalable, and transparent solution for enhancing PID controller design in complex and dynamic environments.

This paper investigates the challenging task of identifying switched Boolean networks (SBNs) with unknown initial subnetworks, leveraging the semi-tensor product tool. The inherent complexity arising from the lack of knowledge regarding the initial subnetwork necessitates a rigorous approach to identification. Initially, the observability of SBNs is discussed using a matrix-based method, providing foundational insights for subsequent identification analysis. We find the accurate one-to-one correspondence between states and outputs by taking time into account. Building on this correspondence and the observability property, the identification of state and output evolution rules is systematically addressed, leading to the proposition of a necessary and sufficient condition for successful SBN identification. Subsequently, an efficient algorithm is developed to implement the proposed identification approach. Finally, the theoretical findings are validated through an illustrative example.

The folding and transport of proteins in the Endoplasmic Reticulum (ER) of mammalian cells exhibit similarities to industrial manufacturing processes, in that they are complex systems regulated by control mechanisms. Recently, two such control systems have been identified: the Unfolded Protein Response (UPR) and AutoRegulation of ER eXport (AREX), which allow the ER to adapt to fluctuations and stress. However, the challenges of modeling their activities arise from the lack of data and the complexity of the signaling pathways that activate them. In this study, we utilize a simulation tool commonly employed in manufacturing plants to develop a model that replicates the protein production process in the ER and the actions of the UPR and AREX in mitigating stress conditions. Our simulations provide insights into the behavior of the cell and represent the first attempt to integrate the entire protein production process and the control activity in the ER. The simulation results demonstrate the potential of regarding the ER as a manufacturing process and provide a novel approach to understanding the complex regulation of the ER.

On the verge of technology, manufacturing industries revolutionize into smart industries, which create a large amount of multivariate time-series data. However, due to sensors’ failure, extreme environment, etc., the collected data are incomplete and have missing values at several instances that result in an erroneous analysis of the data. The key to resolving this problem is data imputation, i.e., replacing the missing values with synthetic values. In this paper, we introduce a generative adversarial network (GAN) framework to generate the synthetic data pertaining to the data imputation. Over the last decade, GANs have presented excellent results to generate synthetic data for images. By following this stream of research, we consider multivariate time-series data from a steel manufacturing industry and propose a GAN-based data imputation technique. We perform several computer simulations to validate and compare the performance of the proposed GAN method with state-of-the-art data imputation techniques.

In this paper, we propose a novel random forest (RF) Q-learning hybrid with experience replay for feedback stabilization of probabilistic Boolean control networks (PBCNs). In particular, by resorting to a model-free reinforcement learning (RL) framework, we present a random forest Q-learning (QLRF) algorithm to design optimal state feedback controllers, thereby stabilizing PBCNs to a given equilibrium point. In reference to better the process of learning the Q-table by replacing it with a function approximator, we substitute the existent neural network (NN) architecture by a RF. We provide insights on the overall computational complexity between the two ways of solving the same problem, proving RF better than its NN counterparts for such applications. The simulations performed on some of the standard examples in the literature demonstrates the effectiveness of the proposed idea.

Utility companies are an integral part of the smart grid, providing consumers with a broad range of energy management programs. The quality of service is based on the measurements obtained from smart metering infrastructures, which can further be improved by sensing at finer resolutions. However, sensing at higher resolutions poses serious challenges both in terms of storage and communication overload due to overgrowing traffic. Compressive sensing is a data compression technique that accounts for the sparsity of electricity consumption pattern in a transformation basis and achieves subNyquist compression. To the best of the authors' knowledge, this is the first study to use the semi-tensor product (STP) for compressed sensing (CS) of power consumption data in the smart grid. In contrast to the conventional CS, the proposed approach has the advantage of reducing the dimension of the sensing matrix needed to sense the signal, thereby significantly lowering the storage requirements. In this regard, we present a comparative study highlighting the difference in compression performance with the conventional CS and STP based CS, where the transformation basis used is Haar and Hankel. We present the results on three publicly available datasets at different sampling rates and outline the key findings of the study.

Probabilistic Boolean control network (PBCN) is a discrete-time dynamical system comprised of a collection of Boolean control networks (BCNs) and switching among them in a stochastic manner. In this paper, the output tracking control of PBCNs is investigated via state feedback and output feedback control. By resorting to the algebraic state-space representation of BCNs, necessary and sufficient conditions for the solvability of the output tracking control problem are presented. A constructive procedure is given to obtain all possible state feedback and output feedback controllers such that the output of PBCNs tracks a constant reference signal. Finally, a PBCN model of a simple manufacturing system is considered to illustrate the effectiveness of the proposed results.

Operating heating power plant (DHPP) with fluctuating load is a complex problem. Thermal energy storage (TES), flexible loads, and operating constraints compound this complexity further. This investigation focuses on the design of a model predictive controller (MPC) that reduces the operating and maintenance cost in a DHPP, considering TES and flexible loads. The MPC accomplishes this task by scheduling boilers, TES units, and flexible loads. To handle the fluctuating demand, the MPC uses forecasts and combines it with a constrained optimization problem. The objective function reflects the cost, whereas the generator limits, TES dynamics, thermal loads, including supply temperature, power plant layout, and reliability, are the constraints. The resulting optimization problem is modeled as a mixed-integer linear program with both continuous and logic variables. Here the logic variables model the operating modes of the boiler and storage units. The use of receding horizon approach enhances the robustness to the forecast errors. The constraints modeling plant layout, supply temperature, and grid reliability lead to a more realistic solution. The MPC is illustrated using simulation on historical data and experiments on a DHPP at Ylivieska, Finland. Our results demonstrate the cost benefits of the proposed approach.