Unknown Dynamic Model Research Articles

Coagulo-Net: Enhancing the mathematical modeling of blood coagulation using physics-informed neural networks

Blood coagulation, which involves a group of complex biochemical reactions, is a crucial step in hemostasis to stop bleeding at the injury site of a blood vessel. Coagulation abnormalities, such as hypercoagulation and hypocoagulation, could either cause thrombosis or hemorrhage, resulting in severe clinical consequences. Mathematical models of blood coagulation have been widely used to improve the understanding of the pathophysiology of coagulation disorders, guide the design and testing of new anticoagulants or other therapeutic agents, and promote precision medicine. However, estimating the parameters in these coagulation models has been challenging as not all reaction rate constants and new parameters derived from model assumptions are measurable. Although various conventional methods have been employed for parameter estimation for coagulation models, the existing approaches have several shortcomings. Inspired by the physics-informed neural networks, we propose Coagulo-Net, which synergizes the strengths of deep neural networks with the mechanistic understanding of the blood coagulation processes to enhance the mathematical models of the blood coagulation cascade. We assess the performance of the Coagulo-Net using two existing coagulation models with different extents of complexity. Our simulation results illustrate that Coagulo-Net can efficiently infer the unknown model parameters and dynamics of species based on sparse measurement data and data contaminated with noise. In addition, we show that Coagulo-Net can process a mixture of synthetic and experimental data and refine the predictions of existing mathematical models of coagulation. These results demonstrate the promise of Coagulo-Net in enhancing current coagulation models and aiding the creation of novel models for physiological and pathological research. These results showcase the potential of Coagulo-Net to advance computational modeling in the study of blood coagulation, improving both research methodologies and the development of new therapies for treating patients with coagulation disorders.

Hierarchical temporal sequence convergence in prescribed-time control for quadrotor UAVs under unknown dynamic disturbances

This paper studies the trajectory tracking control problem for quadrotor unmanned aerial vehicles (UAVs) under unknown disturbances and model uncertainties. A priori information about disturbances is required for most existing prescribed time-extended state observer (PTESO) methods. For unknown disturbances, a barrier function-based adaptive prescribed-time extended state observer (BFAPTESO) is designed by utilizing the time-domain transformation and adaptive mechanism in the prescribed time and using the barrier function method in the rest time. To improve tracking performance amidst model uncertainty, a finite-gain prescribed-time state observer (FGPTSO) is applied in the position loop. Then, the novel prescribed-time controllers are designed in conjunction with the above observers in the attitude and position loops. The proposed control scheme mitigates unknown dynamic disturbances and model uncertainties, guaranteeing that the tracking error achieves and sustains the preset accuracy within the prescribed time. Finally, numerical results indicate that the proposed control method is effective and robust.

Cooperative Maritime Search of Multi-Ship Based on Improved Robust Line-of-Sight Guidance

In this paper, an improved robust line-of-sight (RLOS) guidance-based fuzzy sliding mode controller is presented to control underactuated ships to conduct the cooperative maritime search operation under the presented improved creeping line search method. First, considering that the ship cannot perform turning with corners, an improved creeping line search method is presented by integrating the Bezier method into the traditional creeping line search method to smooth the transition points with corners and employing the cubic spline interpolation method to generate continuous reference paths. Second, an improved RLOS guidance method is presented for the first time by exploring the idea of robust adaptive control to mitigate the chattering effect of the RLOS guidance. Third, the fuzzy logic system with approximate ability is integrated into the design of sliding mode controller to handle unknown nonlinear model dynamics and environmental disturbances. Finally, an improved RLOS guidance-based fuzzy sliding mode controller is presented. The closed-loop stability is guaranteed by the Lyapunov theorem. Comparative simulations are conducted to illustrate the advantages and verify the effectiveness of the presented method.

Data-Driven Model Predictive Control for Uncalibrated Visual Servoing

This paper addresses the image-based visual servoing (IBVS) control problem with an uncalibrated camera, unknown dynamics, and constraints. A novel data-driven uncalibrated IBVS (UIBVS) strategy is proposed, incorporated with the Koopman-based model predictive control (KMPC) algorithm and the adaptive robust Kalman filter (ARKF). First, to alleviate the need for calibration of the camera’s intrinsic and extrinsic parameters, the ARKF with an adaptive factor is utilized to estimate the image Jacobian matrix online, thereby eliminating the laborious camera calibration procedures and improving robustness against camera disturbances. Then, a data-driven MPC strategy is proposed, wherein the unknown nonlinear dynamic model is learned using the Koopman operator theory, resulting in a linear Koopman prediction model. Only input–output data are used to construct the prediction model, and hence, the proposed approach is robust against model uncertainties. Furthermore, with a symmetric quadratic cost function, the proposed approach solves the quadratic programming problem online, and visibility constraints as well as joint torque constraints are taken into account. As a result, the proposed KMPC scheme can be implemented in real time, and the UIBVS performance degradation which arises from the control torque constraints can be avoided. Simulations and comparisons for a 2-DOF robotic manipulator demonstrate the feasibility of the proposed approach. Simulation results further validate that the computation time of the proposed approach is comparable to the one of kinematic-based methods.

Human-in-the-Loop Adaptive Control of a Soft Exo-Suit With Actuator Dynamics and Ankle Impedance Adaptation.

Soft exo-suit could facilitate walking assistance activities (such as level walking, upslope, and downslope) for unimpaired individuals. In this article, a novel human-in-the-loop adaptive control scheme is presented for a soft exo-suit, which provides ankle plantarflexion assistance with unknown human-exosuit dynamic model parameters. First, the human-exosuit coupled dynamic model is formulated to express the mathematical relationship between the exo-suit actuation system and the human ankle joint. Then, a gait detection approach, including plantarflexion assistance timing and planning, is proposed. Inspired by the control strategy that is used by the human central nervous system (CNS) to handle interaction tasks, a human-in-the-loop adaptive controller is proposed to adapt the unknown exo-suit actuator dynamics and human ankle impedance. The proposed controller can emulate human CNS behaviors which adapt feedforward force and environment impedance in interaction tasks. The resulting adaptation of actuator dynamics and ankle impedance is demonstrated with five unimpaired subjects and implemented on a developed soft exo-suit. The human-like adaptivity is performed by the exo-suit in several human walking speeds and illustrates the promising potential of the novel controller.

Adaptive Neural Network Control of an Uncertain 2-DOF Helicopter With Unknown Backlash-Like Hysteresis and Output Constraints.

An adaptive neural network (NN) control is proposed for an unknown two-degree of freedom (2-DOF) helicopter system with unknown backlash-like hysteresis and output constraint in this study. A radial basis function NN is adopted to estimate the unknown dynamics model of the helicopter, adaptive variables are employed to eliminate the effect of unknown backlash-like hysteresis present in the system, and a barrier Lyapunov function is designed to deal with the output constraint. Through the Lyapunov stability analysis, the closed-loop system is proven to be semiglobally and uniformly bounded, and the asymptotic attitude adjustment and tracking of the desired set point and trajectory are achieved. Finally, numerical simulation and experiments on a Quanser's experimental platform verify that the control method is appropriate and effective.

Adaptive asymptotic tracking control of autonomous underwater vehicles based on Bernstein polynomial approximation

In this article, an adaptive asymptotic tracking control approach is first proposed for the autonomous underwater vehicle (AUV) on the horizontal plane by using the Bernstein polynomial approximation. Bernstein polynomials aim to compensate for the synthetical uncertainties including unknown model dynamics and environmental disturbances. However, they cannot be directly used due to the unknown polynomial coefficients. To avoid this trouble, the adaptive backstepping technique is applied to estimation. Considering that the adaptive backstepping can only procure the bounded steady tracking errors and suffer from the redundancy of adaptive parameters, we further integrate the σ-modification and the minimum learning parameter (MLP) technique with the Bernstein polynomial approximation, so as to guarantee the high-precision tracking performance of AUVs with the asymptotic convergent tracking errors and less computation burden. The direct Lyapunov’s method and the Barbalat’s lemma are introduced for proving the stability of the control system. Finally, simulation reveals the advantage of the proposed scheme.

Reinforcement Learning-based path tracking for underactuated UUV under intermittent communication

This paper studies the path control of a six-degree-of-freedom underactuated Unmanned Underwater Vehicle (UUV) under limited communication conditions. Considering the large number of coupling between six-degree-of-freedom underactuated UUV of unknown dynamic models, traditional model-based control methods are difficult to effectively solve the three-dimensional path control problem. A self-attention based soft actor and critic (A-SAC) algorithm is designed to learn effective control policy from random paths. The problem of limited target acquisition by UUV in the actual underwater environment is effectively overcome, which is mainly caused by the inability of UUV to consistently receive information about their expected path. A new state space is designed and a self-attention mechanism is introduced to improve the efficiency of using discontinuous path information. Furthermore, the validation experiment compares classical Reinforcement Learning methods such as DDPG, PPO, and etc. Compared to other existing methods, the proposed A-SAC algorithm can more quickly and effectively learn the path control policy for a six-degree-of-freedom UUV that operates in a complex environment.

Model-free finite-time formation containment control of underactuated UUVs subject to input delays and unknown interaction information

This paper put forwards a model-free adaptive neural finite-time formation-containment (FC) control scheme for time-delayed underactuated unmanned underwater vehicles (UUVs) with unknown dynamic models and some interaction information under switching topologies. The UUVs are steered by multiple child leaders moving along with a root leader, where all leaders are virtual and defined by a new generalized dynamic. A novel autoregressive model-based state predictor is created for each UUV, which can construct future motion states online relying on current and stored states. Considering the containment layer, a finite-time super-twisting observer is used to recover the unavailable interaction information. Moreover, observer-based finite-time kinematic guidance laws employing only 3D positions of the neighboring UUVs are devised, together with protocols for completing formation in finite time. Aiming to accurately identify unknown dynamics and control gains simultaneously, estimator-based adaptive neural networks are introduced exploiting both historical and real-time data. Finally, model-free anti-disturbance finite-time kinetic control protocols are designed based on the data-driven adaptive neural networks and nonlinear tracking differentiators. It is proven that all signals of the closed-loop FC system are bounded, the deployed FC tracking can be achieved in finite time. Simulation results verify the feasibility and superiority of the proposed FC control scheme.

Dynamic Event-Triggered Adaptive Formation With Disturbance Rejection for Marine Vehicles Under Unknown Model Dynamics

Dynamic Event-Triggered Adaptive Formation With Disturbance Rejection for Marine Vehicles Under Unknown Model Dynamics

Model-free kinematic control of redundant manipulators with simultaneous joint-physical-limit and joint-angular-drift handling

Model-free tracking control methods have been developed for redundant robotic manipulators with unknown kinematic or dynamic models. However, most existing works have not considered the problems of joint physical limits and joint angular drift. Joint angular drift refers to the phenomenon that the joint angles of a redundant manipulator are different from their initial state when the end-effector of the manipulator returns to its initial position. Therefore, this article proposes a model-free scheme for the tracking control of redundant manipulators as well as the avoidance of joint physical limits and joint angular drift. The proposed method aims to overcome these challenges by formulating the inverse kinematics problem as a constrained optimization problem, which incorporates the avoidance of joint angular drift into the optimization objective and takes the joint physical limits as a constraint. At the same time, an online Jacobian estimator is designed to observe the state of the manipulator. Specifically, it can estimate the Jacobian matrix of a redundant manipulator in real-time during the operation by fully exploiting the sensory data, without knowing the analytic robot model. Then, the optimization problem is integrated with the Jacobian estimator and solved by a discrete-time algorithm. Theoretical analysis is conducted to prove the stability and convergence of the proposed method. Moreover, the efficacy, practicability and superiority of the proposed method are supported by simulations and experiments based on different redundant manipulators.

Adaptive asymptotic tracking control for autonomous underwater vehicles with non-vanishing uncertainties and input saturation

This article provides an adaptive asymptotic backstepping-based tracking control algorithm for autonomous underwater vehicles (AUVs) under the situation of non-vanishing uncertainties and input saturation. Firstly, the adaptive asymptotic control algorithm is developed to compensate the non-vanishing uncertainties. And then, virtual control stabilizing functions are reconstructed to improve the system performance by virtue of variable substitutions. Next, the hyperbolic tangent function is applied to handle the unknown input saturation problem. Based on the Lyapunov stability theory, the asymptotic stability of AUVs is proven. Finally, simulation results illustrate that the control strategy can effectively inhibit the effect of unknown model dynamics on the system.

Optimal attitude takeover control for failed spacecraft based on multi-nanosatellites

Attitude takeover control of failed spacecraft, which is a key technology in on-orbit service, has received extensive attention in recent years. In the attitude takeover control mission, inertial parameters of the failed spacecraft are unknown or inaccurate. In the meantime, actuator consumption must be considered owing to the limited fuel or energy of the service spacecraft. Using a failed spacecraft takeover control mission executed by multiple nanosatellites as an example, an optimal attitude takeover control method is proposed in this paper to optimize actuator consumption while addressing model uncertainties. Firstly, an auxiliary nonlinear system is constructed and then a radial basis function neural network is employed to estimate the unknown nonlinear dynamics model. Secondly, an optimal control law is designed by combining the inverse optimal principle, adaptive technique, and backstepping theory. Finally, the Harris Hawks optimization (HHO) is adopted for the control allocation problem of multiple nanosatellites. Simulation results demonstrate the feasibility and effectiveness of the proposed method.

Adaptive anti‐disturbance sampling control of autonomous surface vehicles based on discrete‐time concurrent learning extended state observers

SummaryThis paper investigates the surge speed tracking control of an autonomous surface vehicle (ASV) subject to fully unknown internal dynamic, external disturbance, and unknown control input gain. An adaptive anti‐disturbance sampling control method is proposed for an ASV without using any model parameters. Specifically, by utilizing real‐time and historical data, discrete‐time reduced‐order and full‐order concurrent learning extended state observers (CLESOs) are designed to estimate the unknown ASV model parameters and ensure estimation convergence without requiring persistent excitation. Then, a surge speed tracking control law is designed based on the discrete‐time full‐order CLESO. Through Lyapunov stability analysis, the closed‐loop system is proven to be stable, and the estimation and tracking errors are bounded. Simulation and hardware‐in‐loop experimental results validate the effectiveness of the proposed discrete‐time CLESOs for the surge speed tracking of an ASV with fully unknown dynamic model.

Broad learning control of a two-link flexible manipulator with prescribed performance and actuator faults

AbstractIn this paper, we present a broad learning control method for a two-link flexible manipulator with prescribed performance (PP) and actuator faults. The trajectory tracking errors are processed through two consecutive error transformations to achieve the constraints in terms of the overshoot, transient error, and steady-state error. And the barrier Lyapunov function is employed to implement constraints on the transition state variable. Then, the improved radial basis function neural networks combined with broad learning theory are constructed to approximate the unknown model dynamics of flexible robotic manipulator. The proposed fault-tolerant PP control cannot only ensure tracking errors converge into a small region near zero within the preset finite time but also address the problem caused by actuator faults. All the closed-loop error signals are uniformly ultimately bounded via the Lyapunov stability theory. Finally, the feasibility of the proposed control is verified by the simulation results.

Improved dynamic event-triggered anti-unwinding control for autonomous underwater vehicles

This article investigates an improved adaptive dynamic event-triggered control scheme for autonomous underwater vehicles (AUVs) under the quaternion-based attitude representation. The main innovation in this paper is to improve the dynamic event-triggered (DET) algorithm, which makes the trade-off between the control performance of the AUV system and the communication frequency in the controller-to-actuator (CA) channel easier to be achieved. Primarily, to describe the attitude representation of the AUV in a unique form, a unified auxiliary variable confining the scalar quaternion's initial error value is developed to avoid the potential unwinding phenomenon. Subsequently, based on the unified auxiliary variable, an event-based adaptive control law is tactfully designed to mitigate the adverse effects of lumped external perturbations and unknown model dynamics. Moreover, by introducing the modified trigger adjustment algorithm, the trigger frequency can be adapted to the performance requirements of the AUV system at different evolutionary stages. Finally, theoretical analysis and numerical simulations are presented to demonstrate the stability and effectiveness of the provided method.

Safe and Stable RL (S2RL) Driving Policies Using Control Barrier and Control Lyapunov Functions

Deep Reinforcement Learning (DRL) has been successfully applied to learn policies for safety-critical systems with unknown model dynamics in simulation. DRL controllers though optimal in terms of reward, do not provide any safety and stability guarantees. With reliance on model information, safety conditions can be expressed as Control Barrier Functions (CBFs) and performance objectives can be expressed as Control Lyapunov Functions (CLFs) for real-time optimization-based controllers. In this work, we use an amalgamation of model-free RL and model-based controllers to establish safety and stability. We first design CLF, CBF Quadratic Programs (QPs) for different driving manoeuvres on nominal vehicle dynamics. Reinforcement Learning (RL) agents are trained to learn policies for the actual vehicle with enhanced dynamics. In order to incorporate safety and stability while retaining optimal behaviour we selectively guide the RL agents using CLF, CBF QPs. This results in both safe and stable (S2 RL) policies. We empirically validate the proposed methodology on different driving manoeuvres.

Event-Triggered Asymptotic Tracking Control of Underactuated Ships With Prescribed Performance

Adapting to the navigational high-precision tracking tasks, this paper develops an event-triggered adaptive neural asymptotic tracking control framework for underactuated ships. The prescribed performance control (PPC) technique is employed to address spatial constraints in navigation, where the positional tracking errors are transformed via the transformation functions. By using the approximation of radial basis function neural networks (RBF NNs) in the form of minimum learning parameters (MLPs), the uncertainties coming from the unknown model dynamics, the environmental disturbances and the derivation of virtual control laws are offset all together, and a succinct computation is guaranteed. Assisted by the property of neural basis function, the “algebraic loop” problem in the backstepping design is released. To achieve the asymptotic tracking performance, the hyperbolic tangent functions are incorporated into the control laws, which are characterized by the integral-bounded terms. The event-triggered control (ETC) is designed in the controller-to-actuator (CA) channel. Two separate triggering conditions are constructed for the surge and the yaw motions respectively, which are characterized by the compound thresholds composed of a variable and a constant. The “Zeno” behaviors can thereby be avoided. The proposed scheme has three notable characteristics: 1) the computational complexity is reduced by using the MLP technique with lumped uncertainties; 2) the high-precision tracking performance can be achieved through the asymptotic tracking control; 3) the practical problems of spatial constraints and communication burdens can be solved by fabricating a uniform control framework including PPC and ETC. With the aid of direct Lyapunov candidates and the Barbalat’s lemma, the asymptotic convergence of all the tracking errors is proved. Finally, a numerical experiment corroborates the feasibility of the proposed scheme.

Q-learning based model-free input-output feedback linearization control method⋆

Input-output feedback linearization is a nonlinear control method that relies on a precise dynamical model. Combining Q-learning techniques, an input-output feedback linearization correction framework is presented to accomplish model-free feedback linearization of affine nonlinear systems in order to tackle the problem caused by the unknown dynamics model. This framework formulates a model reference tracking control problem that guides the input-output relationship of the nonlinear system into a linear relationship. Due to the two Lie derivative terms present in the feedback linearized controller, the controller is designed as a dual network structure. To overcome the issue of coupling in the dual-network controller, a model-free Q-learning method is presented to solve the unknown controller network weights. The proposed method is experimentally validated on a single-link flexible joint manipulator system, and the resultant linearized system exhibits dynamics similar to the desired linear system in a new tracking task, proving the effectiveness of the proposed method.

Event-triggered adaptive disturbance rejection for marine surface vehicles with unknown dynamics and disturbances

This article presents an event-triggered adaptive disturbance rejection scheme for marine surface vehicles with unknown model dynamics and external disturbances. The ocean environmental disturbances are expressed as a series of sinusoidal superpositions with unknown frequencies, amplitudes and phases. By incorporating the adaptive internal model principle, the disturbance rejection is transformed into the adaptive control problem. The adaptive laws and disturbance filter achieve online disturbance estimations. The adaptive neural backstepping is employed to derive the trajectory tracking control law. The event-triggering mechanism is utilized to reduce the wear and tear of thrusters. The event-triggered adaptive disturbance rejection control does not require a priori knowledge of model dynamics and can be treated as a model-free disturbance estimation and rejection scheme. It is theoretically demonstrated that the closed-loop tracking control system is semi-globally stable and the tracking error can be adjusted to be small. Illustrative results on a 1:70 scaled model CyberShip II validate the control scheme.