This paper presents a reinforcement learning (RL)-based altitude controller for a coaxial octorotor UAV performing landing on a moving platform under sudden, high-magnitude disturbance. The Deep Deterministic Policy Gradient (DDPG) algorithm is employed to learn vertical thrust commands directly from interaction with the environment, while proportional-integral-derivative (PID) loops handle horizontal positioning. A custom reward penalizes both residual altitude error—including steady-state error—and excessive thrust changes. In MATLAB/SIMULINK simulations, the RL controller maintains precise descent trajectories, drives steady-state error to near zero, and adapts to both severe and mild gusts without retuning, outperforming a conventional PID in all tested scenarios. These results demonstrate the promise of purely RL-based altitude control for robust UAV operations in challenging wind conditions.

The advancements of sensor technologies, improvements in computational power, and massive amounts of data from the increasing number of sensors deployed on structures have augmented the need for data-driven techniques in structural health monitoring (SHM), particularly for fatigue crack detection. Data-driven methods and tools are increasingly used in analyzing structures to ensure the absence of any fatigue crack in various engineering fields including civil, mechanical, aerospace, and maritime engineering. This overview aims to present the most recent research that utilizes data-driven methods for fatigue crack detection in engineering structures that fall within the realms of signal processing (SP) and machine learning (ML). It focuses on studies utilizing Deep Learning (DL) for crack detection, highlighting the importance of Transfer Learning (TL).

The objective of this paper is to compare the performance of a recent Model-Free Control (MFC) technique to the performance of well-known Sliding Mode Control (SMC) technique. The application of these two techniques is performed using MATLAB/Simulink on a popular case-study system, namely the coupled Three-Tank system. This comparison is not to say that one technique is better than the other, but rather to see the capacities of each technique in terms of simplicity of implementation, tuning parameters, performance, effects of disturbances and noise. Simulation results show that each technique has advantages with respect to some criteria and disadvantages with respect to others.