室内场景拟人交互研究进展  

Research progress in human-like indoor scene interaction

作　　者：杜韬 胡瑞珍 刘利斌 弋力 赵昊 Du Tao;Hu Ruizhen;Liu Libin;Yi Li;Zhao Hao(Institute for Interdisciplinary Information Sciences,Tsinghua University,Beijing 100084,China;Shanghai Artificial Intelligence Laboratory,Shanghai 200232,China;Shanghai Qi Zhi Institute,Shanghai 200232,China;College of Computer Science and Software Engineering,Shenzhen University,Shenzhen 518061,China;School of Intelligence Science and Technology,Peking University,Beijing 100871,China;Institute for AI Industry Research,Tsinghua University,Beijing 100084,China)

机构地区：[1]清华大学交叉信息研究院,北京100084 [2]上海人工智能实验室,上海200232 [3]上海期智研究院,上海200232 [4]深圳大学计算机与软件学院,深圳518061 [5]北京大学智能学院,北京100871 [6]清华大学智能产业研究院,北京100084

出　　处：《中国图象图形学报》2024年第6期1575-1606,共32页Journal of Image and Graphics

基　　金：国家自然科学基金项目(62322207)。

摘　　要：人类智能是在与环境交互中进化的,因而如何实现智能体与环境的自主交互是推进智能演化的关键。环境自主交互是一项涉及计算机图形学、计算机视觉和机器人等多个学科领域的研究课题,引起广泛的关注和探究,学术界已围绕这一热点研究问题从不同视角和技术维度开展了一系列研究工作。本文着眼于室内场景拟人交互,全面梳理数字人与机器人在室内环境下学习完成特定交互任务过程中需要涉及的仿真交互平台、场景交互数据和交互生成算法3方面基本要素的研究进展。在仿真交互环境搭建方面,本文梳理了仿真环境涉及的仿真技术和研究进展,并对代表性的拟人交互仿真平台进行了介绍;在场景交互数据构建方面,本文从场景交互感知数据集、场景交互运动数据集以及交互数据规模的高效扩充3方面对国内外研究现状进行了详细介绍;在拟人交互感知与生成方面,本文介绍了以交互为导向的场景可供性分析的相关工作,并以交互生成为线索,分别梳理了数字人—场景交互生成、机器人—场景交互生成的相关工作。基于对国内外相关工作的梳理和讨论,最后从交互仿真、交互数据、交互感知和交互生成4个方面,总结了该领域目前仍面临的挑战,并对未来的发展趋势进行了展望。Human intelligence evolves through interactions with the environment,which makes autonomous interactionbetween intelligent agents and the environment a key factor in advancing intelligence.Autonomous interaction with theenvironment is a research topic that involves multiple disciplines,such as computer graphics,computer vision,and robot⁃ics,and has attracted significant attention and exploration in recent years.In this study,we focus on human-like interac⁃tion in indoor environment and comprehensively review the research progress in the fundamental components including simulation interaction platforms,scene interaction data,and interaction generation algorithms for digital humans androbots.Regarding simulation interaction platforms,we comprehensively review representative simulation methods for vir⁃tual humans,objects,and human-object interaction.Specifically,we cover critical algorithms for articulated rigid-bodysimulation,deformable-body and cloth simulation,fluid simulation,contact and collision,and multi-body multi-physicscoupling.In addition,we introduce several popular simulation platforms that are readily available for practitioners in thegraphics,robotics,and machine learning communities.We classify these popular simulation platforms into two main cat⁃egories:simulators focusing on single-physics systems and those supporting multi-physics systems.We review typical simu⁃lation platforms in both categories and discuss their advantages in human-like indoor-scene interaction.Finally,we brieflydiscuss several emerging trends in the physical simulation community that inspire promising future directions:developing afull-featured simulator for multi-physics multi-body physical systems,equipping modern simulation platforms with differen⁃tiability,and combining physics principles with insights from learning techniques.Regarding scene interaction data,weprovide an in-depth review of the latest developments and trends in datasets that support the understanding and generationof human-scene interactions.We

关 键 词：环境交互 交互仿真 交互数据 交互感知 交互生成 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]