Learning and control in virtual reality for machine intelligence
Learning and control in virtual reality for machine intelligence
复制标题
DOI：
10.1109/icicip.2012.6391478
复制
发表时间：
2012-07
期刊：
影响因子：
--
通讯作者：
Xiaonan Fang;Haibo He;Zhen Ni;Yufei Tang
中科院分区：
文献类型：
--
作者：
Xiaonan Fang;Haibo He;Zhen Ni;Yufei Tang
提出了一种基于自适应动态规划（ADP）的虚拟现实（VR）机器智能学习与控制交互平台。近年来的研究结果有力地证明，ADP是类脑智能系统设计的关键技术，而VR是一种功能强大的人机界面，可以提供一个三维（3D）活跃的虚拟环境。结合这两个主题，我们设计了一个交互式系统来促进和演示ADP在VR环境中的学习和控制。本文的主要目标有两个。首先，我们证明了VR可以是一个有用的平台，通过模拟的3D环境来展示和可视化机器智能研究。其次，将虚拟现实技术融入机器智能研究中，可以为智能系统与非结构化环境之间的实时交互提供强大的仿真、验证和促进平台。讨论了虚拟现实平台的详细设计策略，并演示了基于三连杆倒立摆基准的交互系统性能。
This paper presents a Virtual Reality (VR) interactive platform for learning and control for machine intelligence based on Adaptive Dynamic Programming (ADP). Recent research results have provided strong evidences that ADP could be a key technique for brain-like intelligent systems design, and VR is a powerful human-computer interface which can provide a three-dimensional (3D) active virtual environment. Converge these two subjects, we design an interactive system to facilitate and demonstrate the learning and control in VR environment with ADP. Our main goal in this paper is two-fold. First, we demonstrate that VR could be a useful platform to demonstrate and visualize machine intelligence research through the simulated 3D environment. Second, the integration of VR into machine intelligence research can provide a powerful platform to simulate, validate and facilitate the real-time interaction between the intelligent system and an unstructured environment. We discuss the detailed design strategy of the VR platform, and also demonstrated the interactive system performance based on the triple link inverted pendulum benchmark.