Adaptive power system control with neural networks

A New Approach to the Design of an Adaptive Control System

SummaryIn this study, we present a reinforcement learning (RL)‐based flight control system design method to improve the transient response performance of a closed‐loop reference model (CRM) adaptive control system. The methodology, known as RL‐CRM, relies on the generation of a dynamic adaption strategy by implementing RL on the variable factor in the feedback path gain matrix of the reference model. An actor‐critic RL agent is designed using the performance‐driven reward functions and tracking error observations from the environment. In the training phase, a deep deterministic policy gradient algorithm is utilized to learn the time‐varying adaptation strategy of the design parameter in the reference model feedback gain matrix. The proposed control structure provides the possibility to learn numerous adaptation strategies across a wide range of flight and vehicle conditions instead of being driven by high‐fidelity simulators or flight testing and real flight operations. The performance of the proposed system was evaluated on an identified and verified mathematical model of an agile quadrotor platform. Monte‐Carlo simulations and worst case analysis were also performed over a benchmark helicopter example model. In comparison to the classical model reference adaptive control and CRM‐adaptive control system designs, the proposed RL‐CRM adaptive flight control system design improves the transient response performance on all associated metrics and provides the capability to operate over a wide range of parametric uncertainties.

Adaptive Output Feedback Control System Design with Adaptive PFC for Combustion Control of Diesel Engine

Important class of adaptive control systems with non-stochastic description of the uncertainty and disturbances is considered. New approach to the design of the non-stochastic adaptive dual control algorithms is suggested. The elaborated approach is based on the introducing of two new cost functions for optimisation of the control signal and exiting signal added to the control signal. The ellipsoid bounded estimation of the plant parameters is used for the plant parameter estimation and indirect adaptation. This estimation method was also named set-membership estimation. The approach is introduced and described comprehensive using simple controller of self-tuning type. The designed controller, which is used as an example to demonstrate the new suggested method, can be applied to control of nonminimumphase plants. The method can be extended to adaptive controllers of various other types. The violent contrast is attributed to the suggested design method as compared with other dual control approaches for non-stochastic systems with set-membership estimation. The method allows designing of non-stochastic dual adaptive controllers that are very simple computational and consist of the cautious controller with the optimal excitations. Non-stochastic adaptive dual controllers, which are designed using suggested approach, have superiorities over the similar controllers based on the certainty equivalence approach because of the properties of caution and optimal excitation.

Robust Adaptive Control

The paper deals with an adaptive control system design problem for a four rotor helicopter or quadrotor. A simple adaptive control design scheme with a parallel feedforward compensator (PFC) in the internal loop of the considered quadrotor will be proposed based on the backstepping strategy. As is well known, the backstepping control strategy is one of the advanced control strategy for nonlinear systems. However, the control algorithm will become complex if the system has higher order relative degrees. We will show that one can skip some design steps of the backstepping method by introducing a PFC in the inner loop of the considered quadrotor, so that the structure of the obtained controller will be simplified and a high gain based adaptive feedback control system will be designed. The effectiveness of the proposed method will be confirmed through numerical simulations.

Adaptive traffic control systems continuously monitor traffic conditions on an arterial and adjust traffic signal timings to accommodate variation in traffic volume and minimize stops and delays. This paper describes currently practical adaptive control systems which have been used in the United States and worldwide, which include Adaptive Control System-Lite (ACS-Lite), Optimization Policies for Adaptive Control (OPAC), Real-Time Hierarchical Optimized Distributed and Effective System (RHODES), Sydney Coordinated Adaptive Traffic System (SCATS) and Split Cycle Offset Optimization (SCOOT). The comparisons of the benefits from past experiences, cost and hardware requirements will help the traffic engineers and decision makers a guideline to select the optimized adaptive control system based on research objectives.

The paper addresses initial steps involved in the development and flight implementation of new metrics driven L1 adaptive flight control system. The work concentrates on (i) definition of appropriate control driven metrics that account for the control surface failures; (ii) tailoring recently developed L1 adaptive controller to the design of adaptive flight control systems that explicitly address these metrics in the presence of control surface failures and dynamic changes under adverse flight conditions; (iii) development of a flight control system for implementation of the resulting algorithms onboard of small UAV; and (iv) conducting a comprehensive flight test program that demonstrates performance of the developed adaptive control algorithms in the presence of failures. As the initial milestone the paper concentrates on the adaptive flight system setup and initial efforts addressing the ability of a commercial off-the-shelf autopilot with and without adaptive augmentation to recover from control surface failures.

Considered influence of nonlinear system and disturbance, a new adaptive fault-tolerant control method based on neural network model-following adaptive inversion control is introduced for flight control system in the presence of control surface damage. To restrain modeling uncertainties caused by fault system, neural network PID and inversion controllers are design for fault-tolerant flight control system. It is based on BP (back-propagation) neural network which is designed by applying improved algorithm of on-line training and parameter optimization. Moreover the output of fault aircraft can perfectly track the trajectory of the reference model. The scheme is demonstrated through simulations applying the flight control system of a fighter. The simulation results show fault-tolerant flight control system maintains stability and handling qualities.

Parametric study of nonlinear adaptive control algorithm with magneto-rheological suspension systems

A direct adaptive control system for nonparametric plants in the presence of disturbances is proposed. This adaptive control system has robust performance and several desirable features. Since both the proposed controller structure and the frequency-domain adaptation algorithm have parallel architectures, the proposed total adaptive control system is suitable for parallel processing procedures. The effectiveness of the proposed adaptive control system is confirmed by numerical simulations. >

Purpose. The purpose of the article is to develop a methodology for creating cone crusher adaptive control system based on the model-based design method for automated software generation of microprocessor controllers. Methodology. A method based on a block-oriented predictive model was used to generate control signals for the cone crusher. The parameters and structure of this model were identified in real time using measured data from the plant. A prototype of the control system was created in MATLAB/Simulink. Then, the model-based design method was used to generate software for digital signal processors. Mathematical statistics methods were employed to analyze the experimental results. Findings. A method of model-based design of an adaptive control system for a cone crusher has been developed. This system uses a predictive model of a block-oriented structure. This model adjusts the predictive controller’s structure and parameters directly during operation. This approach makes it possible to divide the functions of identifying the process model and generating controls between the two digital controllers. Consequently, the average computational time is reduced while ensuring the stabilization of the degree of homogeneity of ore crushing and separate output of the control size class, with respective standard deviation coefficients not exceeding 3.42 and 1.83 %, respectively. Originality. The regularity of the effect of the closed-side setting and the eccentric speed on the particle size distribution of crushed ore has been established. This shows that high homogeneity of the crushed product is ensured by simultaneously adjusting these input coordinates. We propose a new method for synthesizing an adaptive cone crusher control system based on a model-based design approach. This method provides automated real-time generation of software for microprocessor-based controllers, allowing the system to quickly adjust to changes in rock mass characteristics and other disturbances. Practical value. A hardware and software implementation of an adaptive control system for a cone crusher is proposed. This system is based on a block-oriented predictive model. The model ensures the stabilization of the required ore particle size distribution. This stabilization is achieved by adjusting the closed-side setting and the eccentric speed. The system is based on 16-bit, low-cost digital signal processors. A prototype of the system was tested in a crushing plant at a metallurgical enterprise.

Fuzzy adaptive scheduling and control systems

: Our research is focused on defining robust adaptive control architectures that can be completely characterized with their performance bounds and robustness/stability margins. Research in this direction over the past several years has led to a powerful set of tools, known as L1 adaptive control theory. This new paradigm for design of adaptive control systems embeds the robustness specification explicitly into the control problem formulation (control objective) and allows for decoupling adaptation from robustness, limiting the speed of adaptation only by available hardware (CPU). The performance bounds can be predicted a priori based on the conservative bounds of uncertainty, and the time-delay margin of it can be tuned systematically. The results of this research have been leveraged into different programs across the country, including WP AFRL's Certification of Advanced Flight Critical Systems: Challenge Problem Integration and NASA s Integrated Resilient Aircraft Control . We also revealed some similarities with the disturbance observer (DOB) controllers and identified the main features in the difference between them. The key feature of this difference is that the estimation loop of the L1 adaptive architectures reconstructs the inverse map of the DOB architecture by fast estimation and with that avoids the number of challenges of DOB related to the plant inversion. Insights from this comparison were used towards modification of DOB with improved transient performance.

Robust modification of nonlinear L1 adaptive flight control system via noise attenuation