【作者】 程彬；

【导师】 陈小平；

【作者基本信息】 中国科学技术大学 ， 计算机应用技术， 2015， 硕士

【摘要】 机器人技术在过去的几十年取得了巨大进展,已经广泛应用于工业,医疗,教育以及家庭服务等方面。然而,在某些特殊条件下,用硬质材料制成关节并依靠电机驱动的刚体机器人的应用仍受到一些限制。如在室外凹凸不平的地面移动,穿越弯曲狭小的管道,抓取光滑的球体等,其无法良好的适应复杂的环境。由于具有顺从性并且能和周围环境柔和的交互,软机器人已经成为新的研究热点。然而目前大多数软机器人都是依靠研究人员的直觉和经验进行设计加工,缺乏系统性的理论分析,如受力分析,形变分析等,需要进行大量实验不断对假设进行验证,研发成本较高。在缺乏合适材料的情况下,无法对软机器人进行加工制作,相关的控制算法也无法得到验证。对于软机器人的研究人员来说,一个功能完善的仿真平台是非常必要的。本论文构建了一个蜂巢软体气动手指模型,它是由内嵌的蜂巢气动网络单元构成,克服了矩形气动单元易破裂,对材料韧性要求高的缺点。为了分析气动单元内气压变化对手指形变产生的影响,文中建立了相应的理论模型和物理仿真模型。由于Bullet物理引擎自身不能有效的支持气动机器人的仿真,所以我们对其软体仿真部分进行扩充和改进,基于Bullet构建了一个可以仿真蜂巢气动手指的物理仿真平台。在理论模型中,根据气动手指随气压变化而产生的形变特征我们对气囊建立了气压-形变模型。首先分析了单个气动单元是如何随气压变化而产生形变,在此基础上又分析由多个气动单元构成的气动手指如何发生形变产生弯曲效果。在物理仿真模型中,我们以质点-连接的形式表示蜂巢气动手指,将其宏观形变映射到微观下质点间约束的变化,建立了气压-形变-质点间约束的映射关系,使得仿真平台可以支持气动手指在三维环境下进行动态形变仿真,文章同时还实现了软体手指与刚体物体的动态碰撞检测方法。在本论文工作的基础上,无需考虑材料的选择和加工问题,即可完成虚拟气动手指的气压-形变分析。我们利用这个不含任何刚体关节和电机的气动手指构建了一个虚拟手爪,并对其进行三维环境下的物理仿真。通过控制气动手指内气动单元的气压变化,手指可以实现弯曲的效果,并且可用该虚拟手爪顺从地抓取球形物体,能完成传统硬关节机械手较难完成的任务。如果将此工作应用到传统机器人仿真平台中,可以使其仿真带有软体气动部件的机器人,为软机器人的研究工作提供有效的分析工具。文章的最后给出蜂巢气动手指的理论模型和物理仿真条件下的误差分析,并展望基于仿真平台下软机器人研究前景。更多还原

【Abstract】 The robot technology has made great progress in the past few years, and it’s widely used in industry, education, medicine and home service. However the application of the hard robots which are made of hard material and depend on the motor drive still meets many restrictions. Rigid robots cannot adapt to the environment well such as moving on the outside uneven ground, passing through the pipe and grabbing the smooth sphere. Soft robots have become a new research hotspot because of its compliance with the surroundings and friendly interactive with the environment. However, most of the soft robots are designed and fabricated by the researchers’intuition and experience, which lack of systemic theoretical analysis, such as the analysis of stress and deformation. It requires a large number of experiments to validate the assumptions. The cost of time and material of the research is very high. When in the absence of suitable materials to fabricate soft robots, the control algorithms can hardly be verified. So it’s very necessary for researchers to have a functional simulation platform.In this paper we construct a soft honeycomb pneumatic finger, which is composed of embedded honeycomb pneumatic network. It aims at grasping rigid ball through controlling the variation of its inner pressure. It overcomes the shortcoming of easily rupture and highly required of the toughness of the material. In the meantime, in order to analyze the effect of the aerodynamic changing to the shape of the finger, we also establish a theoretical model and a physical simulation model, which is as the extension part of Bullet soft dynamic simulation. In the theoretical model, we establish a pressure-shape model according to the characteristic of the deformation along with the changing pressure. In the physical simulation model, we use the form of mass-link to represent honeycomb pneumatic finger, mapping the macroscopic deformation to the microscopic changing of the link constraint, thus building the relationship among the pressure, deformation and the link constraint. We can use this simulation platform to simulate the dynamic deformation of this pneumatic finger in3D environment. We also implement the dynamical collision detection method between soft finger and rigid body.Based on this work, without the need to consider the selection and processing of materials, we can complete the analysis of the pressure-deformation of the virtual pneumatic finger. Subsequently we use the pneumatic finger without any rigid joints and motors to construct a virtual hand which is simulated with the use of open source physics engine Bullet. Through the control of inner air pressure of the pneumatic unit the finger can achieve bending and thus the virtual hand can grasp a spherical object obediently. If we apply our work in the impeccable robot simulation platform, making it supporting the simulation of soft components of robots, it will provide very useful validation and analysis tool. Finally we give the error analysis between the theoretical model and physical simulation model and the prospects of soft robot simulation platform based on our work.更多还原