欢迎访问《图学学报》 分享到：
图学学报

图学学报 ›› 2021, Vol. 42 ›› Issue (3): 462-469.DOI: 10.11996/JG.j.2095-302X.2021030462

• 计算机图形学与虚拟现实 • 上一篇    下一篇

基于深度强化学习的虚拟手自适应抓取研究

  

  1. 1. 南京理工大学自动化学院，江苏 南京 210094；  2. 南京理工大学计算机科学与工程学院，江苏 南京 210094
  • 出版日期:2021-06-30 发布日期:2021-06-29
  • 基金资助:
    “十三五”装备预研项目(61409230104，1017，315100104)；中央高校基本科研业务费专项(30918012203)；上海航天科技创新基金 (SAST2019009) 

Research on adaptive grasping of virtual hands based on deep reinforcement learning 

  1. 1. School of Automation, Nanjing University of Science and Technology, Nanjing Jiangsu 210094, China;  2. School of Computer Science and Engineering, Nanjing University of Science and Technology, Nanjing Jiangsu 210094, China
  • Online:2021-06-30 Published:2021-06-29
  • Supported by:
    Thirteenth Five-Year Plan Equipment Pre-research Project (61409230104, 1017, 315100104); Fundamental Research Fund for Central Universities (30918012203); Shanghai Aerospace Science and Technology Innovation Fund (SAST2019009)

PDF (PC)

202

摘要： 在计算机角色动画的抓取研究中，生成动作序列的自然性、稳定性及自适应能力三者难以同时 得到保证，即自然又稳定的抓取控制器往往泛化能力有限，无法适用于其他类型、尺寸物体的抓取任务。通过 引入和抓取类型相对应的手部示教数据、设计回报函数，构建了一种基于深度强化学习的虚拟手自适应抓取控 制器。实验结果表明，该控制器能够生成兼具自然性和稳定性的抓取运动序列，同时对素材库中不同尺寸、不 同类型的基元物体也具备较好的自适应能力。

关键词: 深度强化学习, 示教学习, 运动生成, 虚拟手, 动作捕捉数据

Abstract: For the grasping of computer character animation, it is difficult to guarantee the naturalness, stability and adaptability of the generated action sequence at the same time. In other words, the natural and stable grasping controller are often limited in generalization and cannot be applied to other types of grabbing tasks. A virtual hand adaptive grasping controller was constructed based on deep reinforcement learning by introducing hand teaching data corresponding to the grasping types and by designing the reward function. Experimental results show that the designed controller can generate a grasping motion sequence with both naturalness and stability, and are also highly adaptive for different sizes and types of primitive objects in the material library. 

Key words: deep reinforcement learning, demonstration learning, motion generation, virtual hands, mocap data 

中图分类号: