Robust Control Framework Research Articles

Learning-Based Neural Dynamic Surface Predictive Control for MMC

Reinforcement learning technique was developed recently as an interesting topic in designing adaptive optimal controllers. This technique explicitly provided a feasible solution to circumvent the “curse of dimensionality” and requiring a system model inherent in the classical dynamic programming algorithm. By virtue of this property, in our work, by introducing this technique into a predictor-based online adaptive neural dynamic surface predictive control architecture, we concentrate on a novel robust predictive control framework subject to system uncertainties. To be specific, in this presented framework, an adaptive dynamic programming control strategy utilizing a critic neural network point of view is developed to learn the optimal control policy. Our modification is able to facilitate the alleviation of performance deterioration caused by system uncertainties and enable the smooth and fast learning, while keeping the merits of the finite control-set model predictive control. Finally, the interest and applicability of the proposed control methodology are verified by performance evaluation.

Robust Hierarchical Tracking Control of Vehicle-Manipulator Systems

For vehicle-manipulator systems (VMSs) to perform precise operations, a robust tracking control framework is required. Moreover, to exploit the platform's redundancy, it is desirable that the framework allows several tasks to be completed simultaneously. This work presents a robust hierarchical tracking controller for floating-base robots. In addition to providing uniform global asymptotic stability (UGAS) in free motion and allowing a total task dimension greater than the number of degrees of freedom (DoFs), the proposed controller includes a sliding mode effect and is proved to achieve UGAS even in the presence of bounded disturbances. Moreover, it is proved that uniform global ultimate boundedness (UGUB) is obtained if a continuous approximation of the sliding mode term is used. The stability results are proved mathematically, and are validated through simulations of an articulated intervention autonomous underwater vehicle (AIAUV).

Persistently Feasible Robust Safe Control by Safety Index Synthesis and Convex Semi-Infinite Programming

Model mismatches prevail in real-world applications. Ensuring safety for systems with uncertain dynamic models is critical. However, existing robust safe controllers may not be realizable when control limits exist. And existing methods use loose over-approximation of uncertainties, leading to conservative safe controls. To address these challenges, we propose a control-limits aware robust safe control framework for bounded state-dependent uncertainties. We propose safety index synthesis to find a robust safe controller guaranteed to be realizable under control limits. And we solve for robust safe control via Convex Semi-Infinite Programming, which is the tightest formulation for convex bounded uncertainties and leads to the least conservative control. In addition, we analyze when and how safety can be preserved under unmodeled uncertainties. Experiment results show that our robust safe controller is always realizable under control limits and is much less conservative than strong baselines.

Recursively Feasible Data-Driven Distributionally Robust Model Predictive Control With Additive Disturbances

In this letter we propose a data-driven distributionally robust Model Predictive Control framework for constrained stochastic systems with unbounded additive disturbances. Recursive feasibility is ensured by optimizing over a linearly interpolated initial state constraint in combination with a simplified affine disturbance feedback policy. We consider a moment-based ambiguity set with data-driven radius for the second moment of the disturbance, where we derive a minimum number of samples in order to ensure user-given confidence bounds on the chance constraints and closed-loop performance. This letter closes with a numerical example, highlighting the performance gain and chance constraint satisfaction based on different sample sizes.

Distributionally robust model predictive control for wind farms

In this paper, we develop a distributionally robust model predictive control framework for the control of wind farms with the goal of power tracking and mechanical stress reduction of the individual wind turbines. We introduce an ARMA model to predict the turbulent wind speed, where we merely assume that the residuals are sub-Gaussian noise with statistics contained in an moment-based ambiguity set. We employ a recently developed distributionally model predictive control scheme to ensure constraint satisfaction and recursive feasibility of the control algorithm. The effectiveness of the approach is demonstrated on a practical example of five wind turbines in a row.

Data-driven nonparametric robust control under dependence uncertainty

We consider a multi-period stochastic control problem where the multivariate driving stochastic factor of the system has known marginal distributions but uncertain dependence structure. To solve the problem, we propose to implement the nonparametric adaptive robust control framework. We aim to find the optimal control against the worst-case copulae in a sequence of shrinking uncertainty sets which are generated from continuously observing the data. Then, we use a stochastic gradient descent ascent algorithm to numerically handle the corresponding high dimensional dynamic inf-sup optimization problem. We present the numerical results in the context of utility maximization and show that the controller benefits from knowing more information about the uncertain model.

Multi-agent hierarchical reinforcement learning for energy management

The increasingly complex energy systems are turning the attention towards model-free control approaches such as reinforcement learning (RL). This work proposes novel RL-based energy management approaches for scheduling the operation of controllable devices within an electric network. The proposed approaches provide a tool for efficiently solving multi-dimensional, multi-objective and partially observable power system problems. The novelty in this work is threefold: We implement a hierarchical RL-based control strategy to solve a typical energy scheduling problem. Second, multi-agent reinforcement learning (MARL) is put forward to efficiently coordinate different units with no communication burden. Third, a control strategy that merges hierarchical RL and MARL theory is proposed for a robust control framework that can handle complex power system problems. A comparative performance evaluation of various RL-based and model-based control approaches is also presented. Experimental results of three typical energy dispatch scenarios show the effectiveness of the proposed control framework.

A multi-model multi-objective robust damping control of GCSC for hybrid power system with offshore/onshore wind farm

In the recent decade, the power network has experienced a remarkable energy transition due to the large-scale integration of wind energy resources, especially converter-interfaced modern wind turbines. The increasing/decreasing wind penetration, on the other hand, unexpectedly affects the low frequency oscillations of the modern power grid. In this context, the proposed work adopts a multi-model multi-objective robust control framework to design a supplementary damping controller for the gate-controlled series capacitor (GCSC) to stabilize the hybrid power system with a permanent magnet synchronous generator (PMSG)-based off-shore wind farm (OSWF) and a doubly-fed induction generator (DFIG)-based onshore wind farm (ONWF). A bilinear matrix inequality (BMI) optimization problem, formulated as multi-objective synthesis considering H2/H∞ performance along with pre-defined pole placement, is presented for the GCSC control design, which is solved by a two-step linear matrix inequality (LMI) approach. In addition, all LMI constraints are constructed based on the multi-model control framework to incorporate multiple operating conditions. Afterward, the significant improvement in the damping characteristics of the closed-loop system, covering a wide operating range, is confirmed using eigenvalue analysis. The effectiveness of the scheme is validated using two case studies based on the hybrid power system, subject to various disturbances and uncertainties. The simulation results show the robustness and higher damping performance of the proposed multi-model strategy, compared to a conventional and a robust damping controller, for mitigating power system oscillations alongside voltage fluctuations of the wind farms.

Robust tracking control of a robot arm actuated by permanent magnet synchronous motors

AbstractIn this article, we show the use of the recently developed enhanced extended observer design technique in tracking time‐dependent references by a PMSM‐actuated robot arm. It is proven that this robust control framework guarantees achievement of semiglobal practical stability and makes trajectories of the system arbitrarily close to those that would had been obtained by the non‐robust nominal feedback‐linearizing control if there were no plant uncertainties.

Predictive control of uncertain systems over networks with redundant link

Networked systems using redundant channels to transmit data can effectively reduce the probability of data loss and improve system reliability and control margin. However, the structural complexity and economic cost of the system are also increased. To balance the redundancy and feasibility, the ratio of attraction domain to packet loss rate is defined as a balanced feasibility index. In this paper, single-channel packet loss is considered as Bernoulli distribution and a bounded packet loss network system control model is constructed as the arbitrary bounded packet loss control problem for redundant channel transmission network system. Therefore, the robust conditions of the closed-loop system and the constraints of the input and state are established under the framework of robust predictive control to construct the linear matrix inequality (LMI) optimization problem. Finally, to verify the effectiveness of the design method proposed in this paper, the discrete time-varying linear system and the main steam control system with redundant channels are used as study cases.

Adaptive robot climbing with magnetic feet in unknown slippery structure.

Firm foot contact is the top priority of climbing robots to avoid catastrophic events, especially when working at height. This study proposes a robust planning and control framework for climbing robots that provides robustness to slippage in unknown environments. The framework includes 1) a center of mass (CoM) trajectory optimization under the estimated contact condition, 2) Kalman filter–like approach for uncertain environment parameter estimation and subsequent CoM trajectory re-planing, and 3) an online weight adaptation approach for whole-body control (WBC) framework that can adjust the ground reaction force (GRF) distribution in real time. Though the friction and adhesion characteristics are often assumed to be known, the presence of several factors that lead to a reduction in adhesion may cause critical problems for climbing robots. To address this issue safely and effectively, this study suggests estimating unknown contact parameters in real time and using the evaluated contact information to optimize climbing motion. Since slippage is a crucial behavior and requires instant recovery, the computation time for motion re-planning is also critical. The proposed CoM trajectory optimization algorithm achieved state-of-art fast computation via trajectory parameterization with several reasonable assumptions and linear algebra tricks. Last, an online weight adaptation approach is presented in the study to stabilize slippery motions within the WBC framework. This can help a robot to manage the slippage at the very last control step by redistributing the desired GRF. In order to verify the effectiveness of our method, we have tested our algorithm and provided benchmarks in simulation using a magnetic-legged climbing robot Manegto.

Renewable energy-powered semi-closed greenhouse for sustainable crop production using model predictive control and machine learning for energy management

Renewable energy consumption in agriculture is ascending, catering to the food needs of the rising population and protecting the environment. Maximizing renewable energy usage efficiency is essential for achieving sustainable development goals. In this work, a nonlinear integrated controlled environment agriculture model is constructed to correlate the impact of weather disturbances, temperature control actuators, humidity control actuators, fertilization, and irrigation to the states of crop production facilities and crop growing conditions. Linearization of the model is performed to reduce the computation time while retaining the accuracy of the nonlinear model. A robust model predictive control framework is developed to maximize renewable energy power usage efficiency and maintain a hospitable sustainable cultivation environment. To improve the robustness of control to hedge against the forecast uncertainties, the disjunctive data-driven uncertainty sets built upon the historical forecast errors are constructed by using machine learning methods, including principal component analysis and kernel density estimation. This work also presents the result of the simulation of controlling a renewable energy-powered semi-closed greenhouse growing tomatoes located in Ithaca, New York. Compared to other model predictive control frameworks, which do not leverage the machine learning approaches, the proposed control framework enables a 0.7%–66.9% reduction in renewable energy usage and a 0.7 wt% to 16.1 wt% increase in crop production. Further analysis of this work reveals that the integrated controlled environment agriculture model can help in increasing renewable energy usage efficiency from 4.7% to 127.5%.

A Fuzzy Approximation for FCS-MPC in Power Converters

Standard model predictive control is an optimization-based control strategy that can handle multiple control objectives and system nonlinear constraints. However, it typically suffers from the limitation of the uncertainties in practical systems, such as external unknown disturbances and parametric uncertainties. Motivated by aforementioned limitation, in this article, a novel robust model predictive control framework, endowed with the merits of fuzzy logic system and finite control-set model predictive control solution, is proposed. The main objective of this article is to enhance the system robustness while guaranteeing adaptability to different conditions. More specifically, a fuzzy approximation point of view, which has a good potential to approximate the unknown nonlinear functions, is deployed and incorporated into the proposed design, which allows one to explicitly take the system nonlinear dynamics and uncertainties into account. The novelty of the proposed methodology relies on the fact that any prior knowledge and explicit information of system model parameters are not required, thereby resulting in considerable enhancement of robustness. Furthermore, the input-to-state stability of the approximation error system is proven through Lyapunov analysis, and it demonstrates that the estimated errors are uniformly ultimately bounded. Finally, the interest of the proposal is experimentally confirmed for modular multilevel converter.

A Visual Predictive Control Framework for Robust and Constrained Multi-Agent Formation Control

A Visual Predictive Control Framework for Robust and Constrained Multi-Agent Formation Control

Novel Compliant Control of a Pneumatic Artificial Muscle Driven by Hydrogen Pressure Under a Varying Environment

In this article, a pneumatic artificial muscle (PAM) based on a metal hydride (MH) is considered for a compact compliant actuator. It is suitable for broad applications of the human–robot interaction (HRI). To address the problem of the HRI represented by a varying environment, a compliant control is introduced. In fact, the bottlenecks of improving the performance in the compliant control of the PAM actuator are: an inherent nonlinear dynamics of a PAM, the parametric and nonlinear uncertainties influenced by a varying environment, and an additional high dimension introduced by an MH employed as a driving force for the PAM. We propose a learning-based adaptive robust control (LARC) framework to tackle these challenges. A Bayesian learning technique deals with the parameter adaptation for the adaptive control. The effectiveness of the LARC has been examined in extensive experiments of tracking control.

HD-RMPC: A Hierarchical Distributed and Robust Model Predictive Control Framework for Urban Traffic Signal Timing

Due to the nonlinearity and dynamics of transportation systems, traffic signal control (TSC) in urban traffic networks has always been an important challenge. In recent years, model predictive control (MPC) has shown extraordinary potential in TSC due to its outstanding ability to model dynamic systems. However, the relatively complex online computing, lack of reasonable setpoints for target solving, and uncertainty of traffic network hinder MPC from being further applied. To address these problems, we propose a hierarchical, distributed, and robust model predictive control (HD-RMPC) framework for urban TSC. At the slow-update layer, the road network is dynamically divided into several subareas according to regional attributes and real-time traffic demand. Meanwhile, the volume is coordinated in a robust way for the purpose of traffic equilibrium and overflow prevention. Then, the set-point matrix of each subarea is calculated to equalize the flow in the subarea. This distributed framework guarantees the real-time performance of MPC in urban traffic networks. At the fast-update layer, we adopt an improved prediction model by explicit modeling of the disturbance and reduce the prediction error. Finally, the objective function is reconstructed and solved at the control layer to obtain the optimal control law. Through continuous and asynchronous optimization of the set point and prediction model, the framework significantly improves the control effect. Simulation evaluation based on a real-world road network demonstrates that the proposed HD-RMPC method outperforms all baselines and maintains excellent real-time performance.

Semiclosed Greenhouse Climate Control Under Uncertainty via Machine Learning and Data-Driven Robust Model Predictive Control

This work proposes a novel data-driven robust model predictive control (DDRMPC) framework for automatic control of greenhouse in-door climate. The framework integrates dynamic control models of greenhouse temperature, humidity, and CO₂ concentration level with data-driven robust optimization models that accurately and rigorously capture uncertainty in weather forecast error. Data-driven uncertainty sets for ambient temperature, solar radiation, and humidity are constructed from historical data by leveraging a machine learning approach, namely, support vector clustering with weighted generalized intersection kernel. A training-calibration procedure that tunes the size of uncertainty sets is implemented to ensure that data-driven uncertainty sets attain an appropriate performance guarantee. In order to solve the optimization problem in DDRMPC, an affine disturbance feedback policy is utilized to obtain tractable approximations of optimal control. A case study of controlling temperature, humidity, and CO₂ concentration of a semiclosed greenhouse in New York City is presented. The results show that the DDRMPC approach ends up with 14% and 4% lower total cost than rule-based control and robust model predictive control with L₁-norm-based uncertainty set, respectively. The constraint violation probability, which is the percentage of time that the greenhouse system states violate the constraint throughout the whole growing period, for DDRMPC is only 0.39%. Hence, the proposed DDRMPC framework can prevent the greenhouse climate from becoming harmful to plants and fruits. In conclusion, the proposed DDRMPC approach can improve the greenhouse climate control performance and reduce cost compared with other control strategies.

Hamilton–Jacobi–Bellman–Isaacs equation for rational inattention in the long-run management of river environments under uncertainty

A new stochastic control model for the long-run environmental management of rivers is mathematically and numerically analyzed, focusing on a modern sediment replenishment problem with unique nonsmooth and nonlinear properties. Rational inattention as a novel adaptive strategy to collect information and intervene against the target system is modeled using Erlangization. The system dynamics containing the river discharge following a continuous-state branching with an immigration-type process and the controlled sediment storage dynamics lead to a nonsmooth and nonlocal infinitesimal generator. Modeling uncertainty, which is ubiquitous in certain applications, is considered in a robust control framework in which deviations between the benchmark and distorted models are penalized through relative entropy. The partial integro-differential Hamilton–Jacobi–Bellman–Isaacs (HJBI) equation as an optimality equation is derived, and its uniqueness, existence, and optimality are discussed. A monotone finite difference scheme guaranteeing the boundedness and uniqueness of numerical solutions is proposed to discretize the HJBI equation and is verified based on manufactured solutions. Model applications are also conducted with the parameter values identified from the available data and physical formulae. The computational results suggest that environmental management should be rationally inattentive in a state-dependent and adaptive manner.

Reinforcement learning-based saturated adaptive robust neural-network control of underactuated autonomous underwater vehicles

This paper studies a high-performance intelligent online adaptive robust saturated dynamic surface control framework for underactuated autonomous underwater vehicles by engaging Actor–Critic neural networks in the presence of unmodeled dynamics, uncertainties, ocean disturbances and actuator saturation. The proposed controller is designed based on reinforcement learning method to compensate the effects of unmodeled dynamics and uncertainties more accurately that can lead to a better performance for the controller. The Actor–Critic neural networks are trained real-time by designing online training laws creatively and a new critic function is proposed to supervise the closed-loop performance with the aid of the Critic neural network. The proposed structure for the reinforcement learning method benefits from a model-free algorithm and only relies on the measurable variables of the closed-loop control system. This independency from system dynamics results in considerable low computational burden for the controller and, hence, the proposed control algorithm is efficient computationally. The hyperbolic tangent function is ingeniously used as a saturated stabilizer term to bound the control signals that results in low-amplitude control action and the probability of actuator saturation phenomenon is minimized by learning and compensating the actuator saturation nonlinearity as well. Finally, the stability of the proposed closed-loop system is investigated by the Lyapunov’s direct methodology and simulations along a comparative study with some quantitative assessments certify the contributions of this paper.

An sEMG based adaptive method for human-exoskeleton collaboration in variable walking environments

In this paper, a novel long short memory network – adaptive robust iterative learning control (LSTM-ARILC) framework is proposed to achieve accurate continuous motion estimation and adaptive following control for the exoskeleton robot, which is applied to offer assistance to human walking in variable environments. The LSTM network is established to estimate the human lower limb motion through the processed surface Electromyography (sEMG) signals, and the human-exoskeleton coupling fuzzy dynamic model is further developed. Then the closed-loop ARILC controller is designed to compensate the estimated errors and realize the adaptive and robust following control, and its asymptotic stability is rigorously proved via Lyapunov theory. The performance of the proposed method is evaluated with numerical experiments and compared with adaptive PID controller and adaptive fuzzy sliding mode controller (SMC). The maximum following errors of ARILC after 6 iterations can be reduced by more than 99% compared to the initial iteration, and the controller is quite less sensitive to the environment changes than the other two controllers, which proves that the proposed control framework is much more robust and effective in variable walking environments.