Keywords: Adaptive control design
Abstract: This TUTORIAL paper introduces the basics of Deadzone-Adapted Disturbance Suppression (DADS) approach to robust adaptive control. Robustness to disturbances in adaptive control has heretofore always come with a price---either a bound on the unknown parameter needs to be known (when parameter projection is used for robustification) or the residual regulation error depends on unknown bounds of both the parameter and the disturbance and is therefore unassignable (when sigma-modification is used). With DADS, the residual regulation error becomes assignable, i.e., arbitrarily small, regardless of the unknown parameter and disturbance size. The DADS design incorporates a deadzone in the update law and nonlinear damping with updated gains in the control law---a combination of tools not previously used in robust adaptive control. Four stability properties are established, including a zero asymptotic gain relative to the disturbance and parameter, a form of ISS for the plant state transients, and the absence of the adaptation drift. These novel robustness features are proven with novel stability techniques, developed expressly for the purpose of robust adaptive control.

Keywords: Adaptive control design, Adaptive observers and estimators, Linear systems
Abstract: In this work, we present the design of a novel control strategy for the stabilization of an uncertain second-order multivariable linear system of which only the generalized output vector is measurable and where the high-frequency gain is an unknown positive-definite symmetric matrix. The proposed approach corresponds to a Lyapunov-based adaptive control strategy, which consists of three fundamental components: a measurable auxiliary filter that imitates the original system, an adaptive slave system that follows the filter system, and an adaptive controller that uses the states of the slave system. We provide a numerical experiment to highlight the effectiveness of the proposed adaptive control strategy in accomplishing the regulation problem.

Keywords: Adaptive control design, Learning theory and algorithms

Keywords: Adaptive control design, Learning theory and algorithms, Robotics. Mechatronics
Abstract: Time-invariant controllers can lead to suboptimal reference tracking for robotic systems faced with variable loads and dynamics. To mitigate this, adaptive control techniques such as model reference adaptive control (MRAC) and adaptive incremental nonlinear dynamic inversion (INDI) adapt to varying dynamics. However, these adaptive control methods require partial knowledge of the system dynamics to obtain the sensitivity derivatives/Jacobian--the direction in which to adapt. The prior knowledge required by these methods can increase the development/design time and limit controller flexibility when faced with highly variable dynamics. Here, inspired by insect control and cerebellum learning systems, we propose an `implicit' adaptive feedforward (IAFF) control architecture. In contrast to other methods, it does not require any prior knowledge of the system dynamics, and can be deployed in a plug-and-play manner. IAFF uses one filter to learn the sensitivity derivatives and another filter to generate the feedforward actuation command. We demonstrate this adaptive controller in simulations and on a Parrot Mambo mini-drone. We observe that from random initial filter weights, the controller improved the drone’s altitude reference tracking capabilities within 30 seconds, and position (pitch and roll) control within minutes. The low computational cost allowed the real-time learning algorithm to be executed on the onboard Arm A9 microprocessor.

Keywords: Adaptive control design, Linear system identification, Identification methods design and analysis
Abstract: The paper addresses the problem of parametric convergence enhancement up to the finite time convergence (FTC) in direct adaptation schemes. Two FTC mechanisms are proposed to preserve the alertness of the FTC property. In contrast to the most of parameter identification schemes with FTC, the main novelties of the proposed FTC mechanisms consist in the following tree properties: 1) the proposed FTC mechanisms can be combined with a wide class of standard algorithms of adaptation (like gradient, algorithms with dynamic or memory regressor extension); 2) zeroing of the control error is always guaranteed for any continuous regressor without any additional conditions (like not square integrability of some signals) that is important property in direct adaptive control; 3) FTC can be provided for robust modification of standard algorithms of adaptation.

Keywords: Adaptive observers and estimators, Nonlinear systems, Robotics. Mechatronics
Abstract: In this paper, we address the problem of online estimation of spherical features in the field of camera for robotics. Specifically, we consider a mobile robot equipped with inertial measurement units (IMUs) – providing linear acceleration and rotational velocity measurements in the body-fixed frame – and a pinhole camera that projects 3D points onto the image plane. To tackle this problem, we adopt the parameter estimation-based observer (PEBO) approach on manifolds to design a feature observer. Under a sufficient excitation condition, our design guarantees a globally exponentially convergent estimate of both the radius and the center coordinates of spherical targets. Simulation results validate the theoretical analysis and demonstrate the performance of the proposed feature observer.

Keywords: Algorithms and guarantees, Learning theory and algorithms, Intelligent learning in control systems
Abstract: The point of interest recommendation system is the core function of location-based social networks platforms. How to provide users with accurate and effective point of interest recommendations has become an urgent problem to be solved. The existing point of interest recommendation algorithms mainly have the following problems: ignoring the spatio-temporal connection between non-adjacent access points of interest; the problem of user interest drift has not been taken seriously; the cold start problem limits the performance of most algorithm models. In response to the above problems, this paper proposes a point of interest recommendation algorithm model based on spatio-temporal attention, which uses graph neural networks to connect the weight information between points of interest to explore users' short-term intentions, thereby coping with the problem of user interest drift; Attention mechanism is used to learn and capture the spatio-temporal correlationship between non-adjacent points of interest in ; a time preference matching mechanism is proposed to solve the cold start problem. This paper conducts comparative experiments with baseline models on multiple datasets, and the results show that the model has significant improvements on these datasets.

Keywords: Identification methods design and analysis, Linear control design, Knowledge-based Systems
Abstract: As physical systems are often central to many typical cyber-physical human systems, effectively modeling physical systems is crucial to integrating and controlling embedded systems. A set of general explicit expressions is derived to identify a simple transfer function model based on the describing function approach. Using these expressions, parameters of transfer function models are obtained. However, this approach leads to significant errors in determining transfer function model parameters due to various dynamic errors. On the basis of these expressions, a large set of data is generated to train machine learning models for various transfer function models. This procedure is proposed to increase the accuracy of the auto-tune identification method. Simulation examples show promising results by the proposed auto-tune identification method

Keywords: Optimal control design, Intelligent learning in control systems
Abstract: This paper addresses the challenge of designing resilient control systems under Denial of Service (DoS) attacks for discrete-time systems. A learning-based framework is proposed to reconstruct lost measurements and compute optimal controllers using input-output data, eliminating the need for a complete system model. By leveraging state reconstruction techniques, the framework estimates missing information during DoS periods, ensuring robust control performance. Two algorithms, based on policy iteration (PI) and value iteration (VI), are developed to learn the optimal feedback control policy. The effectiveness of the proposed methodology is illustrated via a numerical example.