柔带网格试衣机器人设计与仿人形研究

How to let the customer perceive the effect of selecting garment on his/her body is an important problem remaining to be solved in garment e-commerce. The emerging technology of shape-changeable robotic mannequin(SCRM) that can imitate human body shapes provides a feasible solution to the above problem. Customer can perceive the selecting garment effect by checking the image or graphic data of the garment on a SCRM with its imitated shape close to his/her body. According to the key problem of how to improve the flexibility and controllability of SCRM’s shape imitation, in this project, innovative research on designing mechanical structure and shape imitation method of SCRM will be carried out. (1) Improving shape imitation flexibility by designing flexible mechanical structure. A new mechanical structure, flexible-belt-net controlled SCRM will be designed. The flexible-belt-net is construed by length adjustable flexible-belts. The flexible-belts are positioned by elastic curved-bars and envelopes the elastic curved-bars. Each pair of transverse elastic curved-bar and lengthways elastic curved-bar are articulated in cross shape. Thus the flexible-belts are woven into a net. Positions of the elastic curved-bars are controlled by linear actuators. Postures and the bending deformations of the elastic curved-bars are controlled by the tension of flexible-belts. Thus the shape of the flexible-belt-net can be controlled, and stretchable cloth hugging the flexible-belt-net are deformed to imitate different human body shapes. Shapes and mechanical characters of the elastic curved-bars will be designed based on shape space of human bodies. (2) Improving shape imitation controllability by making the shape imitation computable. Bidirectional mapping will be established between flexible-belt’s tension, the positions of relative elastic curved-bars and the shape of flexible-belt, which makes the forward and backward shape imitation of the flexible-belt-net solvable; SCRM’s geometric surface will be computed by designing a flexible-belt-net-based data-driven model. Research on this project will provide new theories and methods for design and shape imitation of SCRM.

如何让顾客感知着装效果是服装电子商务领域一个亟待解决的关键问题。近年来出现的试衣机器人技术通过机器仿形真人并试穿服装，为解决该问题提供了可能。本项目围绕试衣机器人研究中的核心问题，即如何提高仿形灵活性和可控性，从结构设计和仿形控制两方面开展创新研究：（1）由结构的柔性提高仿形灵活性。提出柔带网格试衣机器人，由长度可调的柔带连接并包络位置可控的十字铰接弹性曲杆，构建柔带网格；利用柔带张力控制弹性曲杆的姿态与弯曲变形，从而调节柔带网格形状，驱动穿在柔带网格上的弹性紧身衣仿形不同人体；提出基于人体形状空间的弹性曲杆形状与力学性能设计方法。（2）由仿形可计算化提高仿形可控性。建立柔带形状与柔带张力、弹性曲杆位置之间的双向映射，使柔带网格正向、反向仿形可解；建立基于柔带网格的试衣机器人表面形状数据驱动模型。本项目研究将为试衣机器人设计和仿形控制提供新的理论与方法。

我国服装网购规模已达万亿，但其退货率高达25%。高退货率的一个重要原因在于顾客在网购时很难确切感知服装在其身上的着装效果，导致真实服装与顾客期望之间存在不能接受的差异。目前主流的基于图形图像形变技术的虚拟试衣，其模拟效果的真实感程度仍很难满足商业需求。本项目研究设计能仿形不同人体的可变形试衣机器人，使服装在形变的试衣机器人上呈现其在不同人体上的着装效果，从而解决服装电子商务中的试衣真实感难题。这对促进服装电子商务的健康发展，降低服装网购高退货率具有重要的经济和社会意义。.项目围绕如何提高试衣机器人的仿形灵活性、可控性和实用性等关键科学问题，从结构设计、仿形控制和试衣应用三个方面展开研究。在机构上，提出了柔带网格结构、分块结构、柔带约束分块结构三种方式的试衣机器人。设计了以柔带包络多弹性曲杆，通过柔带张紧力使柔带-弹性曲杆耦合形变的机构，并在人体形状空间计算设计弹性曲杆；将几何上的人体特征控制曲线和人体参数曲面实物化为形状可调可控的柔带及其控制的柔性分片曲面，从而建立人体数字几何模型与试衣机器人物理表面之间的形状映射理论模型和实物样机。在仿形控制上，提出了柔带与弹性曲杆的形变解耦求解方法，使柔带如同参数曲线一样形变可控；提出了试衣机器人表面形变的物理计算方法，使其正向和反向仿形可解。在试衣应用方面，为试衣机器人仿形提出了人体感知的多种方法，包括基于机电传感网络的人体尺寸感知，基于色彩及色彩-深度图像的人体形状感知等；提出了将服装迁移到目标人体呈现着装效果的计算方法。.项目开展过程中，搭建了3类试衣机器人样机，发表了15篇相关学术论文，其中国际SCI检索期刊论文12篇；申请了12项国家发明专利，其中5项已获得授权。项目建立了从试衣机器人结构设计、仿形控制到虚拟试衣应用的较为完备的理论方法体系，构建了较为完整的基于试衣机器人的虚拟试衣技术链路，为试衣机器人服务于产业应用奠定了理论和技术基础。