文章基本信息

标题：AVALIAÇÃO DE POLÍTICAS ABSTRATAS NA TRANSFERÊNCIA DE CONHECIMENTO EM NAVEGAÇÃO ROBÓTICA
本地全文：下载
作者：Beirig, Rafael Lemes ; Pereira, Fernando Andrade ; Koga, Marcelo Li 等
期刊名称：Revista de Sistemas e Computação - RSC
出版年度：2012
卷号：2
期号：2
语种：Portuguese
出版社：Revistade Sistemas y Computación
摘要：This paper presents a new approach to the problem of solving a new task by the use of previous knowledge acquired during the process of solving a similar task in the same domain, robot navigation. A new algorithm, Qab-Learning, is proposed to obtain the abstract policy that will guide the agent in the task of reaching a goal location from any other location in the environment, and this policy is compared to the policy derived from another algorithm, ND-TILDE. The policies are applied in a number of different tasks in two environments. The results show that the policies, even after the process of abstraction, present a positive impact on the performance of the agent. 800x600 COMPARISON AND EVALUATION OF ABSTRACT POLICIES FOR TRANSFER LEARNING IN ROBOT NAVIGATION TASKS Normal 0 21 false false false PT-BR X-NONE X-NONE MicrosoftInternetExplorer4
其他摘要：This paper presents a new approach to the problem of solving a new task by the use of previous knowledge acquired during the process of solving a similar task in the same domain, robot navigation. A new algorithm, Qab-Learning, is proposed to obtain the policy that will guide the agent in the task of reaching a goal location from any other location in the environment, and this policy is compared to the policy derived from another algorithm, ND-TILDE. The policies are applied in a number of different tasks in two environments. The results show that the policies, even after the process of abstraction, present a positive impact on the performance of the agent. 800x600 COMPARISON AND EVALUATION OF POLICIES FOR TRANSFER LEARNING IN ROBOT NAVIGATION TASKS Normal 0 21 false false false PT-BR X-NONE X-NONE MicrosoftInternetExplorer4
关键词：Relational representation;Abstract policy;Reinforcement learning;Inductive logic programming;Robotics;Representação Relacional;Políticas Abstratas;Aprendizado por Reforço;Programação Lógica Indutiva;Robótica
其他关键词：Relational representation; Abstract policy; Reinforcement learning; Inductive logic programming; Robotics