Estimation-based disturbance adaptive model predictive control for wheeled biped robots

This paper presents the adaptive model predictive control approach for a two-wheeled robot manipulator with varying mass. The mass variation corresponds to the robot picking and placing objects or loads from one place to another. A linear parameter varying model of the system is derived consisting of local linear models of the system at different values of the varying parameter. An adaptive model predictive control controller is designed to control the fast-varying center of gravity angle in the inner loop. The reference for the inner loop is generated by a slower outer loop controlling the linear position using a linear quadratic Gaussian regulator. This adaptive model predictive control/linear quadratic Gaussian control system is simulated on the nonlinear model of the robot, and the closed-loop performance of the proposed scheme is compared with a system having inner/outer loop controllers as proportional integral derivative/proportional integral derivative, feedback linearization/linear quadratic Gaussian, and linear quadratic Gaussian/linear quadratic Gaussian. It is seen that adaptive model predictive control shows mostly superior and otherwise very good performance when compared to these benchmarks in terms of reference tracking and robustness to mass parameter variations.

To solve the horizontal vibration problem of high-speed elevators caused by unevenness of guide rails and excitation of guide rail joints, considering the uncertainty of high-speed elevator parameters and multivariable coupling, an adaptive explicit distributed model predictive control strategy based on a multi-agent car system of high-speed elevators is proposed. First, the coupling constraints of high-speed elevator car are analyzed, and the horizontal vibration model of multi-agent car system is established. Second, the topology relationship of car multi-agent communication based on graph theory is analyzed, and the optimal control law of agent distribution model prediction is realized offline by multi-parameter programming technology. Meanwhile, by introducing a correction term with parameter estimation error, the uncertain parameter terms of the multi-agent car system can be estimated online. Design an adaptive explicit distributed model predictive control strategy for multi-agent car systems by combining offline collaborative distributed optimization with online uncertain parameter adaptive estimation. Finally, the time-domain and frequency-domain responses of high-speed elevator multi-agent car systems under two typical guide excitations are analyzed by simulation calculation and compared with the numerical results under passive control, adaptive control, and model predictive control. The results show that after the proposed control strategy is adopted, the mean horizontal vibration acceleration of the car system decreases by more than 48.4%, which further verifies the effectiveness of the proposed control strategy.

Adaptive Model Predictive Control for Constrained Nonlinear Systems

Over the past decades, machine learning techniques, such as supervised learning, reinforcement learning, and unsupervised learning, have been increasingly used in the control engineering community. Various learning algorithms have been developed to achieve autonomous operation and intelligent decision making for many complex and challenging control problems. One of such problems is bipedal walking robot control. Although still in their early stages, learning techniques have demonstrated promising potential to build adaptive control systems for bipedal robots. This paper gives a review of recent advances on the state-of-the-art learning algorithms and their applications to bipedal robot control. The effects and limitations of different learning techniques are discussed through a representative selection of examples from the literature. Guidelines for future research on learning control of bipedal robots are provided in the end.

In this paper, an adaptive model predictive control (MPC) scheme with friction compensation, subject to incremental control input constraints and parameter uncertainties, is proposed for a three-wheeled omnidirectional mobile robot (OMR). The proposed control framework is in a cascaded structure, wherein the outer-loop is kinematic-based control and the inner-loop is designed based on adaptive linear MPC. First, a complex nonlinear dynamic model of the OMR in the world coordinate frame is transformed and partially linearized into a reduced nonlinear model in the moving coordinate system. The nonlinearity of the reduced model only arises from Coulomb friction. Then an estimated system is established for the reduced nonlinear system, with an adaptive update law estimating the system uncertain parameters. To facilitate the linear MPC design, part of the control efforts is derived by feedback compensation of the Coulomb friction forces, resulting in a linear estimated system. The other part is designed by a constrained linear MPC. Feasibility and stability analyses are given for the proposed adaptive MPC scheme. Finally, experimental comparisons with model-based MPC are carried out to verify the effectiveness of the proposed control scheme.

The main challenge of the well-celebrated Levenberg-Marquardt algorithm (LMA) is the selection of the searching direction and adaptation parameters. Secondly, the implementation of the LMA for online model identification has faced challenges as it is a batch optimization. As a third challenge, the solution of the Levenberg-Marquardt based on the full-Newton nonlinear optimization (FNNO) for online applications have been limited due to its unguaranteed positive definiteness. This paper presents two versions of the modified Levenberg-Marquardt algorithm (MLMA) for neural network model identification and adaptive predictive control for the online dynamic model identification and adaptive control of a self-balancing two-wheel LEGO Mindstorms NXTway-GS robot. The first version is the online-window-approach of the modified Levenberg-Marquardt algorithm (OWA-MLMA) based on approximate Guass-Newton algorithm (AGNA) for training neural network model predictor. The second version is a neural network-based adaptive predictive control (APC) algorithm based on the full-Newton nonlinear optimization of the modified Levenberg-Marquardt algorithm (FNNO-MLMA) for online adaptive control. A NNARMAX model predictor for the NXT robot is first trained and validated using the OWA-MLMA based on AGNA. The validated NNRAMAX model is then used for the design of the NN-APC based on FNNO-MLMA. Finally, the model identification based on OWA-MLMA and APC based on FNNO-MLMA schemes are simulated in closed-loop for online NNARMAX model identification and adaptive predictive control of the NXT robot. The comparison of the proposed OWA-MLMA based on AGNA shows superior performance over the recursive incremental back-propagation algorithm (INCBPA) while the proposed NN-APC based on FNNO-MLMA shows excellent control and good tracking performances over a discrete-time fixed-parameter proportional-integral-derivative (PID) controller for the NXT robot control. The simulation results show that the developed OWA-MLMA based on AGNA and the APC based on FNNO-MLMA can be adapted for the online dynamic modeling, automation, control and robotics applications.

Bipedal walk as an activity requires an excellent sensorial and movement integration to coordinate the motions of different joints, getting as a result an efficient navigation system for a changing environment. Main applications of the study of biped walking are in the field of medical technology, to diagnose gait pathologies, to take surgical decisions, to adequate prosthesis and orthesis design to supply natural deficiencies in people and for planning rehabilitation strategies for specific pathologies. The same principles can also be applied to develop biped machines; in daily situations, a biped robot would be the best configuration to interact with humans and to get through an environment difficult for navigation. If the biped robot is designed with human proportions, the robot could manage his way through spaces designed for humans, like stairs and elevators, and hopefully the interaction with the robot would be similar to interaction with a human being. The National University of Colombia has been working on the design and control of biped robots, supported by two research groups, Biomechanics and Mobile Robots. The joint effort of the groups has produced three biped robots with successful walks, based on a single idea: if an appropriate design methodology exists, the resulting hardware must have appropriate dynamical characteristics, making easier the control of the walking movements. The design process successfully merges two lines of research in bipedal walk, passive an active walks, by using gait patterns obtained thanks to the simulation of a kneed passive walker to create the trajectory followed by the control of an active biped robot. Our actual line of research in biped robots is to use biped robots reproducing the human gait pattern as engineering tools to test the behavior of below-knee prostheses, thus producing a biped robot with heterogeneous legs that allows the evaluation of how the prosthetics influence the normal gait of the robot while it is walking as a human.

In the research of renewable energy power generation, tubular grid-connected solid oxide fuel cells with the apparent advantage in voltage regulation have been widely applied in power systems. Recently, a model predictive control has been applied to consider the nonlinear constraints of tubular grid-connected solid oxide fuel cells, which cannot be considered by a proportional-integral-derivative controller. Both model predictive control and proportional-integral-derivative controller achieve only 80% fuel efficiency. An adaptive multistep model predictive control (AMMPC) is proposed to improve the fuel efficiency of tubular grid-connected solid oxide fuel cells and simultaneously consider systemic thermodynamics and electrochemistry constraints. The AMMPC contains the advantages of adaptive control and multistep model predictive control. Both adaptive two-step model predictive control and three-step model predictive control are designed for tubular grid-connected solid oxide fuel cells. With the more accurate prediction ability, the AMMPC improves the fuel efficiency of tubular grid-connected solid oxide fuel cells with higher fuel efficiency (86.5%) than model predictive control (80%) and proportional-integral-derivative (80%). Both feasibility and effectiveness of the AMMPC are verified with high fuel efficiency under simple and complex power demands cases.

Fundamentally, control system designs are concerned with the flow of signals in the closed loop. In this paper, we are to present the control technique at the next level of abstraction in control system design. We construct a control using implicit function with support vector regression-based data-driven model for the biped, in the presence of parametric and functional dynamics uncertainties. Based on Lyapunov synthesis, we develop decoupled adaptive control based on the model predictive and the data-driven techniques and construct the control directly from online or offline data. The adaptive predictive control mechanisms use the advantage of data-driven technique combined with online parameters estimation strategy in order to achieve an efficient approximation. Simulation results are presented to verify the effectiveness of the proposed control.

Analysis and design of control systems via parameter‐based approach

This paper considers adaptive model predictive control of an underwater vehicle with an uncertain discrete-time model that contains a singular regressor matrix complicating to estimate uncertain parameters online. As the first, immersion and invariance-based estimator is designed via a state augmentation approach to remove the singularity of the regressor matrix, and stability of the estimator is investigated. Then, using the information coming from the designed estimator, an adaptive linear model predictive control is established for reference trajectory tracking considering some hard and soft constraints. The performance of the proposed estimator and adaptive model predictive control is presented by simulation results.

Adaptive model predictive control with propulsion load estimation and prediction for all-electric ship energy management

This book provides a comprehensive study of nonlinear adaptive robust model predictive control (MPC). Chapters 2–5 present a framework for the analysis and synthesis of nonlinear robust MPC. This framework includes the treatment of robustness, computation methods, and performance improvement. Chapters 6–7 show how to develop the basic ideas for the design and analysis of the nonlinear adaptive robust MPC. One of the key techniques is the set-based approach, in which the internal model identifier allows the MPC to compensate for future changes in the parameter estimates and uncertainty associated with the unknown model parameters. Chapters 8–12 illustrate how to implement the synthesis approaches for nonlinear adaptive robust MPC, and a robust adaptive economic MPC is also proposed. This text also gives a finite-time identification method, which can be used to estimate the unknown parameters in finite time, provided a persistence of excitation (PE) condition is satisfied. This identification method is particularly effective in the online implementation of MPC. The early chapters study continuous-time systems, and Chapters 13–14 extend the set-based estimation and robust adaptive MPC to discrete-time problems. While adaptive robust MPC is an improvement on robust MPC, this book shows that feedback MPC can be used to improve the open-loop MPC. At each sampling instant, a sequence of parameter estimates can be performed/invoked to improve the control performance. Economic MPC is also incorporated so as to improve the control performance in a broader way. This book is intended for someone learning functions of a complex variable and who enjoys using Matlab. It will enhance the experience of learning complex-variable theory and will strengthen the knowledge of someone already trained in this branch of advanced calculus. Supplying students with a bridge between the functions of complex-variable theory and Matlab, this supplemental text enables instructors to easily add a Matlab component to their complex-variables courses. The book shows students how Matlab can be a powerful learning aid in such staples of complex-variable theory as conformal mapping, infinite series, contour integration, and Laplace and Fourier transforms. In addition to Matlab programming problems, the text includes many examples in each chapter along with Matlab code.

Underground coal gasification (UCG) is a potential technology that enables to mine coal without traditional mining equipment. The coal is gasified deep in underground and produced syngas is processed on the surface. The most important technical problem in UCG is unstable quality of syngas and control. This paper proposes advanced control based on an adaptive predictive controller. The maintaining of desired calorific value depends on flow rates of gasification agents injected to the underground geo-reactor and controlled exhaust. The paper proposes a physical model of UCG technology and applies a method of multivariate adaptive regression splines (MARS) to model the gasification process. This method satisfactorily approximates nonlinearity in the process variables. The paper proposes adaptive model predictive control (MPC) using online model estimation and applied it on the MARS model of UCG that imitates the real process. The results have shown that optimization of manipulation variables can replace manual control in UCG. Getting better quality of syngas depends on setpoints, optimized manipulation variables, and constraints used in MPC. In simulations, the adaptive MPC has shown better performance in comparison with manual and PI control.

In this paper, an adaptive model predictive controller for overheating steam temperature control of thermal power plants is designed, which is based on the control object with large delay, large inertia, nonlinearity and strong time-varying properties. Through the on-line identification and control of different models, compared with predictive controllers in a general model, in terms of adjusting the superheat steam temperature, it can shorten adjusting time drastically, reduce even eliminate the overshoot and improve the dynamic performance greatly when applying in adaptive model predictive controller. The results show that the adaptive model predictive controller, because of its simple implementation, can be used in power plants, and also can be applied to solve similar problems, which has a broad application prospects.