Simultaneous Path Following and Obstacle Avoidance of Field Tracked Vehicles via Model Predictive Control with Deep Deterministic Policy Gradient

Yu Cheng Sung, Chun Ting Sung, Wen Chuan Tseng, Shean Jen Chen*

*Corresponding author for this work

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

Abstract

Unmanned ground vehicles (UGVs) will be widely adopted in agricultural applications. To accomplish autonomous cruising in farm, path following is an essential skill. However, in the process of field cruising, some obstacles such as wild animals or motorcycles are present. In this study, tracked vehicles are utilized with deep deterministic policy gradient (DDPG) compensating for model uncertainties and achieving collision avoidance simultaneously. Among all, the most important issue is to keep the UGV following the predetermined path in specific agricultural field environment and coping with the uncertainty of the surroundings. Path following and obstacle avoidance of field tracked vehicles are conducted by using model predictive control (MPC) with a controller (agent) trained by DDPG. Therefore, we proposed control algorithm fusion with MPC and model-free DDPG.

Original languageEnglish
Title of host publicationOptics and Photonics for Advanced Dimensional Metrology II
EditorsPeter J. de Groot, Richard K. Leach, Pascal Picart
PublisherSPIE
ISBN (Electronic)9781510651500
DOIs
StatePublished - 2022
EventOptics and Photonics for Advanced Dimensional Metrology II 2022 - Virtual, Online
Duration: 9 May 202220 May 2022

Publication series

NameProceedings of SPIE - The International Society for Optical Engineering
Volume12137
ISSN (Print)0277-786X
ISSN (Electronic)1996-756X

Conference

ConferenceOptics and Photonics for Advanced Dimensional Metrology II 2022
CityVirtual, Online
Period9/05/2220/05/22

Keywords

  • Path following
  • deep deterministic policy gradient
  • deep reinforcement learning
  • field tracked vehicle
  • model predictive control
  • obstacle avoidance

Fingerprint

Dive into the research topics of 'Simultaneous Path Following and Obstacle Avoidance of Field Tracked Vehicles via Model Predictive Control with Deep Deterministic Policy Gradient'. Together they form a unique fingerprint.

Cite this