Planning of Opposite Q Learning Based on Virtual Sub-Target in Unknown Environment

DOI

10.2991/ncce-18.2018.139How to use a DOI?

Keywords

Mobile robot; Virtual sub target; Opposite Q-learing; Unknown environment.

Abstract

Aiming at the problem of Q value update is slow and easy to produce dimension disaster for Q learning algorithm in complex unknown environment, A path planning algorithm for opposite Q learning robot based on virtual sub targets in unknown environment is proposed .The algorithm is according to the state trajectory of the mobile robot, two state linker are established to record the current sate-action pair and current state-reverse action pairs, from the value of the tail of a single chain, the current state, is traced back to the Q value at the end of a single linker head until the target is reached. Meanwhile, searching for optimal virtual sub target in local detection domain to solve the problem of Q-learning prone to dimension in a large-scale environment. The experiments show that the algorithm can effectively speed up the convergence of learning algorithm and improve the learning efficiency in complex unknown environment and achieve the robot navigation task with the better path.

Copyright

© 2018, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Shengmin Wang
AU  - Wei Lin
PY  - 2018/05
DA  - 2018/05
TI  - Planning of Opposite Q Learning Based on Virtual Sub-Target in Unknown Environment
BT  - Proceedings of the 2018 International Conference on Network, Communication, Computer Engineering (NCCE 2018)
PB  - Atlantis Press
SP  - 839
EP  - 844
SN  - 1951-6851
UR  - https://doi.org/10.2991/ncce-18.2018.139
DO  - 10.2991/ncce-18.2018.139
ID  - Wang2018/05
ER  -