A Preview Neuro-Fuzzy Controller Based on Deep Reinforcement Learning for Backing Up a Truck-Trailer Vehicle

Eduardo Bejar, Antonio Moran Cardenas

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

8 Scopus citations

Abstract

Backing up truck-trailer vehicles is a required task in several industry sectors. Controllers that have been proposed to automate this process usually struggle when the vehicle is required to follow a nonlinear trajectory. This paper presents a neuro-fuzzy controller based on preview control and deep reinforcement learning for reverse parking truck-trailer vehicles. The controller consists of a deep neural network trained with reinforcement learning. A preview control signal is coupled into the trained controller to improve the control performance when tracking complex trajectories. Moreover, a fuzzy logic approach is used to avoid the jackknife state. Simulation results are presented to show that the controller is able to track circular and sinusoidal trajectories.
Original languageSpanish
Title of host publication2019 IEEE Canadian Conference of Electrical and Computer Engineering, CCECE 2019
StatePublished - 1 May 2019
Externally publishedYes

Cite this