Optimal Control of Drones for a Train-Drone Railway Diagnostic System

The inspection of railway systems with traditional wayside detectors allows mainly the detection of wheels and axle bearings defects and can be time-demanding, unsafe, and heavily dependent on humans. To overcome these issues and then optimize and automate the rail and track diagnosis process, drones can be an excellent solution thanks to their onboard state-of-the-art cameras and sensors. Thus, with the aim of rapidly collecting highly accurate data, an innovative hybrid movable railway diagnostic architecture, consisting of a diagnostic train and a fleet of drones, is defined in this work. From the control point of view, the main interest is in optimally managing the crucial phase of drones returning to and landing on the moving train when the railway inspection mission is completed. To control the fleet of drones, a combination of consensus algorithm in the leader-following mode and linear quadratic regulator (LQR) is implemented for the flight formation phase and the landing phase on the moving base platform (i.e., the diagnostic train), respectively. The landing phase is performed both as vertical or oblique descent through a go to goal and as oblique descent along a predefined path. The obtained results of the railway diagnostic architecture simulations are presented and discussed in detail. In particular, they show that the vertical and oblique descent performed as go to goal are certainly faster than the oblique descent along a predefined path.