1,721,055 research outputs found
Recurrent Neural Network-based Internal Model Control design for stable nonlinear systems
Full text linkOwing to their superior modeling capabilities, gated Recurrent Neural Networks, such as Gated Recurrent Units (GRUs) and Long Short-Term Memory networks (LSTMs), have become popular tools for learning dynamical systems. This paper aims to discuss how these networks can be adopted for the synthesis of Internal Model Control (IMC) architectures. To this end, first a gated recurrent network is used to learn a model of the unknown input-output stable plant. Then, a controller gated recurrent network is trained to approximate the model inverse. The stability of these networks, ensured by means of a suitable training procedure, allows to guarantee the input-output closed-loop stability. The proposed scheme is able to cope with the saturation of the control variables, and can be deployed on low-power embedded controllers, as it requires limited online computations. The approach is then tested on the Quadruple Tank benchmark system and compared to alternative control laws, resulting in remarkable closed-loop performances
On the stability properties of Gated Recurrent Units neural networks
Full text linkThe goal of this paper is to provide sufficient conditions for guaranteeing the Input-to-State Stability (ISS) and the Incremental Input-to-State Stability (δISS) of Gated Recurrent Units (GRUs) neural networks. These conditions, devised for both single-layer and multi-layer architectures, consist of nonlinear inequalities on network's weights. They can be employed to check the stability of trained networks, or can be enforced as constraints during the training procedure of a GRU. The resulting training procedure is tested on a Quadruple Tank nonlinear benchmark system, showing remarkable modeling performances
Stability of discrete-time feed-forward neural networks in NARX configuration
No full textThe idea of using Feed-Forward Neural Networks (FFNNs) as regression functions for Nonlinear AutoRegressive eXogenous (NARX) models, leading to models herein named Neural NARXs (NNARXs), has been quite popular in the early days of machine learning applied to nonlinear system identification, owing to their simple structure and ease of application to control design. Nonetheless, few theoretical results are available concerning the stability properties of these models. In this paper we address this problem, providing a sufficient condition under which NNARX models are guaranteed to enjoy the Input-to-State Stability (ISS) and the Incremental Input-to-State Stability (δISS) properties. This condition, which is an inequality on the weights of the underlying FFNN, can be enforced during the training procedure to ensure the stability of the model. The proposed model, along with this stability condition, are tested on the pH neutralization process benchmark, showing satisfactory results. Copyright (C) 2021 The Authors
Performance improvement of an air-to-water heat pump through linear time-varying MPC with adaptive COP predictor
No full textAir-to-water heat pumps are one of the most common and energy efficient heating systems for buildings, particularly floor-heating plants. One way to further improve their effectiveness is to control the heat pump exploiting the dependence of its coefficient of performance (COP) on the external temperature and temperature of the return water from the load. In particular, it is possible to exploit the heat pump when its efficiency is higher, so optimizing its performance in a predictive manner, anticipating the impact of external conditions. For the case of an air-to-water heat pump, the optimization problem is nonlinear due to the load dependence of the heat pump COP and variable supply water flow rate. This may pose implementation problems. If we address a standard control hardware, simplified optimal control formulations are more effective. In this paper, we specifically address this issue, and a reduced-order, linear, but adaptive time-varying predictive model of the heat pump COP is designed. Our solution takes into account the variation of the heat pump efficiency based on the external temperature and the load profile, which are changing within the control horizon. The proposed COP model is then used within a linear time-varying model predictive controller formulation which provides a prediction of the heat pump dynamical behavior based on the load dependence of the heat pump COP, while tackling the nonlinearities of the system imposed by the variable water flow rate in the hot water tank and also by the load dependence of the heat pump COP. The proposed approach has been implemented and in detail tested on a reference model based on a real case study from the Denmark Technical University, Risø Campus, SYSLAB. An intensive simulation analysis and complements the testing, showing the accuracy and the potential of the method, also in the perspective of practical implementation
An Offset-Free Nonlinear MPC scheme for systems learned by Neural NARX models
No full textThis paper deals with the design of nonlinear MPC controllers that provide offset-free setpoint tracking for models described by Neural Nonlinear AutoRegressive eXogenous (NNARX) networks. The NNARX model is identified from input-output data collected from the plant, and can be given a state-space representation with known measurable states made by past input and output variables, so that a state observer is not required. In the training phase, the Incremental Input-to-State Stability (δISS) property can be forced when consistent with the behavior of the plant. The δISS property is then leveraged to augment the model with an explicit integral action on the output tracking error, which allows to achieve offset-free tracking capabilities to the designed control scheme. The proposed control architecture is numerically tested on a water heating system and the achieved results are compared to those scored by another popular offset-free MPC method, showing that the proposed scheme attains remarkable performances even in presence of disturbances acting on the plant
Control design for nonlinear systems: trading roubustness and performance with the MPC approach
No full textSafeguarded optimal policy learning for a smart discrete manufacturing plant
No full textAn approach to safely learn and deploy, at fast rate, a given optimization-based controller for the routing problem in smart manufacturing is presented. The considered application features a large number of integer decision variables, combined with nonlinear dynamics, temporal-logic constraints, and hard safety constraints. The approach employs a neural network as feedback controller, trained using a data-set of state-input pairs collected with the optimization-based controller. A safeguard mechanism checks whether the input computed by the neural network is feasible or not, in the latter case the optimization-based controller is called. Results on a high-fidelity simulation suite indicate a strong decrease of average computational time combined with a negligible loss of plant performance
- …
