In this paper, we synthesize controllers for linear time-invariant (LTI) systems using collected offline data. We first synthesize control-invariant sets from offline collected data using backward reachable set computations and then propose a data-driven predictive controller equipped with terminal constraints. After formulating the optimal control problem and designing the terminal constraints and costs, we ensure the recursive feasibility of the optimization problem, asymptotic stability of the closed-loop system, and satisfaction of input and state constraints. We develop an overall online algorithm for our approach that does not require the initial state to be included in the control invariant set, guarantees optimal behavior of the system’s operation, and does not have Lyapunov constraints that restrict the feasibility region. Furthermore, we extend our developed data-driven control algorithm to stabilize interconnected systems in a decentralized manner, where the satisfaction of a small gain condition is additionally required. We illustrate the effectiveness of our approach through a detailed example.

A novel deep neural network (DNN) based controller for omnidirectional robots is proposed. The controller decomposes the prescribed reference path, corresponding to a fixed prediction horizon, into multiple paths of shorter horizons. This implicitly enforces a Hankel structure in the input and consequently also on the output. Taking advantage of this, a convolutional vision transformer model is used to realize the controller which is then trained to predict state and controls over multiple prediction horizons. Model training is performed in a self-supervised manner using a synthetic dataset. The proposed controller is shown to be more efficient than a model designed for a single prediction horizon. In comparison to a model predictive controller, the proposed approach exhibits competitive performance in path following tasks and is three times faster on average for the same prediction length.

We consider the problem of designing controllers based on measurements affected by noise, for linear systems with unknown dynamics, that ensures one or more performance specifications. In particular, we consider (i) the$\mathcal{D}-\mathbf{stabilization}$problem, where performance specifications are given in terms of placing the eigenvalues of the closed-loop system in a predefined region$\mathcal{D}$of the complex plane, (ii)$\mathcal{H}_{\infty}$performance, (iii)$\mathcal{H}_{2}$performance, and a combination of some of the above. For$\mathcal{D}- \mathbf{stabilization}$problem, a general convex region$\mathcal{D}$defined by a quadratic matrix inequality (QMI) is considered. For this general region$\mathcal{D}$, we provide sufficient conditions, given in terms of data-based linear matrix inequalities, for controller design. For some regions of practical interest, these conditions are necessary and sufficient. We further consider the problem of designing data-driven controllers such that multiple performance specifications, not necessarily given in terms of stability regions, are guaranteed.

There is a great interest in developing robust and efficient controllers for autonomous cars. A special issue is related with the high energy consumed by the embedded GPUs for tasks like sensory data processing, feature extraction and decision making. In this paper, the benefits of using Semantic-Segmentation and Variational-Autoencoder methods integrated with a camera-based end-to-end controller are shown. With Semantic Segmentation an agent trained using Deep Deterministic Policy Gradient algorithm exhibits a faster convergence, a better performing policy and increased robustness when compared with the alternative raw RGB input. The Variational-Autoencoder, used for dimensionality reduction, represented a significant decrease in the GPU usage and in the neural network processing time.

A hierarchical algorithm involving two-layer optimization-based control policies with varying degrees of abstraction is proposed, including upper layer task scheduling and lower layer local path planning. A scenario with two robot arms performing cooperative pick-and-place tasks for moving objects is specifically addressed. The main focus of the paper lies on the bottom layer of the hierarchical control scheme, more precisely on the online generation of the synchronous robot trajectories using distributed minimum-time model predictive control (DMPC) algorithms. To this end, we introduce a decelerating coupling term in the cost functions of the individual distributed optimization algorithms to synchronize the overall robot motion. The performance of the algorithm is illustrated by extensive simulations with high-fidelity robot dynamic models.

We suggest a procedure for dynamic parameter estimation of serial robot manipulators. Its basic idea relies on the synthesis of an optimal manipulation trajectory, which is based on properly introduced parameter aggregates to ensure a collection of numerically well-conditioned data-sets, yielding an accurate computation of parameter estimates. The optimal trajectory itself is computed by using a memetic algorithm, which represents a metaheuristic combination of genetic and gradient based algorithms. The algorithm is experimentally verified by estimating the parameters of the UR5 robot by Universal Robots.

The paper presents control architecture for steer-by-wire (SBW) vehicles based on conventional ones. Estimation techniques linked to state space design enter in the control loop. The active observer (AOB) is extended to under-actuated multiple-input-multiple-output (MIMO) systems. Stochastic strategies are presented. Simulation results for the model-matching problem are presented. In the presence of road disturbances (e.g. lateral wind), SBW vehicles have superior performance.

A force-feedback actuation loop for a steer-by-wire vehicle is developed. It is shown that the performance of this loop can be essentially improved by the introduction of a torque sensor. Model reference based control algorithms based on disturbance observer (DOB) and active observers (AOB) are applied to enhance the robustness vs. non-modelled dynamics and uncertain driver impedance.