Synthesis of humanoid whole-body motion with smooth transition

Synthesis of humanoid whole-body motion with smooth transition
复制标题

DOI:
10.1080/01691864.2015.1024284
复制
发表时间:
2015-05
期刊:
影响因子:
2
通讯作者:
C. Sung;T. Kagawa;Y. Uno
C. Sung;T. Kagawa;Y. Uno
中科院分区:
计算机科学4区
文献类型:
--
作者:
C. Sung;T. Kagawa;Y. Uno

文献摘要

相似文献

在规划特定的类人连续运动时,希望从一个运动转移到另一个运动,而在运动之间不停止。通过连接运动来实现连续性能的问题在机器人领域被称为运动合成。为了解决这个问题,重要的是分配一个适当的过渡运动,连接一个先前的运动到一个后运动。在本研究中,我们提出了一种在保持平衡和避免碰撞等各种约束条件下产生平滑过渡运动的新方法。我们的算法包括两个基于优化技术的顺序过程。首先,我们用极大极小问题的形式来分配运动的过渡时刻。其次,我们平滑地连接转换瞬间,以满足整个转换过程中的各种约束。应用该方法合成了仿人机器人HOAP-3的行走和踢腿动作。我们通过连续的动作连接来确认机器人完成给定的任务。此外,通过对过渡运动的规划,提高了运动性能。特别是,机器人在更稳定的条件下快速踢球,以保持平衡。结果表明,该方法通过解决运动综合问题,实现了有效的运动规划。图形抽象
In planning specific sequential motions of a humanoid, it is desired to transfer from one motion to another motion without stopping between motions. The problem of achieving continuous performance by connecting motions is referred to as motion synthesis in the robotic fields. To solve this problem, it is important to assign an appropriate transition motion that connects a prior motion to a posterior motion. In this research, we propose a new method of generating smooth transition motions under various constraints such as maintaining balance and collision avoidance. Our algorithm comprises two sequential processes based on optimization techniques. First, we assign transition instants of the motions with the formulation of a minimax problem. Second, we connect the transition instants smoothly to satisfy various constraints throughout the transition. Our method is applied to synthesize the walking and kicking motions of the humanoid robot HOAP-3. We confirm that the robot achieves a given task by connecting motions continuously. Furthermore, the motion performance is improved with the planning of transition motion using our method. In particular, the robot kicks a ball quickly under a more stable condition for maintaining balance. These results show that the proposed method achieves effective motion planning by solving the motion synthesis problem. Graphical Abstract