First Author:Haibo Sun

Correspondence Author:Feng Zhu

Co author:YangyangLi,PengfeiZhao,Yanzi Kong,Jianyu Wang,Yingcai Wan,Shuangfei Fu

Title of Paper:Viewpoint planning with transition management for active object recognition

Journal:Frontiers in Neurorobotics

Volume:17

DOI Number:10.3389/fnbot.2023.1093132

Institution:Faculty of Robot Science and Engineering, Northeastern University

Teaching and Research Group:物理化学

Place of Publication:SWITZERLAND

Summary:Active object recognition (AOR) provides a paradigm where an agent can capture
 additional evidence by purposefully changing its viewpoint to improve the quality
 of recognition. OneofthemostconcernedproblemsinAORisviewpointplanning
 (VP) which refers to developing a policy to determine the next viewpoints of the
 agent. A research trend is to solve the VP problem with reinforcement learning,
 namely to use the viewpoint transitions explored by the agent to train the VP
 policy. However, most research discards the trained transitions, which may lead to
 an ine cient use of the explored transitions. To solve this challenge, we present a
 novel VP method with transition management based on reinforcement learning,
 which can reuse the explored viewpoint transitions. To be specific, a learning
 framework of the VP policy is first established via the deterministic policy gradient
 theory, which provides an opportunity to reuse the explored transitions. Then, we
 design a scheme of viewpoint transition management that can store the explored
 transitions and decide which transitions are used for the policy learning. Finally,
 within the framework, we develop an algorithm based on twin delayed deep
 deterministic policy gradient and the designed scheme to train the VP policy.
 Experiments on the public and challenging dataset GERMS show thee ectiveness
 of our method in comparison with several competing approaches.

Key Words:active object recognition, viewpoint planning, deterministic policy gradient, twin delayed deep deterministic policy gradient, viewpoint transition management, reinforcement learning

Document Code:WOS:000950219200001

Discipline:Natural Science

First-Level Discipline:Chemistry

Page Number:1093132

ISSN:1662-5218

Translation or Not:No