A Neural-network-based Control System for a Dynamic Model of Tractor With Multiple Trailers System

Tractors with multiple trailers are widely applied means of transport in manufacturing systems. There exist numerous designs of trailers and tractors, making the estimation of the system trajectory and the required transportation corridor a complex task. It is also difficult to achieve the same trajectory for a manually operated tractor for multiple runs. The problem is complicated if there are multiple towed trailers or a dynamic drive on slippery ground. One approach is to replace the driver with an automated steering system. This paper presents a dynamic model of a tractor with multiple trailer system, based on the Lagrange formalism, which is controlled by artificial neural networks. To account for the slip phenomenon, a sigmoidal tire model was used. The algorithm of the artificial neural network provides the most appropriate input parameters for tractor steering for a given transportation area. The input parameters are the torques applied to the tractor wheels and are determined by the algorithm based on the data collected by the LiDAR scanner during the train run. These data include distances for each unit from the obstacle (e.g., wall), information about the occurrence of a collision, and the distance traveled by the tractor. The simulation results of the integration of the dynamic model and the neural network modeled are presented in a graphic form. The proposed algorithm ensures a collision-free ride of the system.