【作者】 徐孟；

【导师】 孙守迁；

【作者基本信息】 浙江大学 ， 计算机科学与技术， 2006， 博士

【摘要】 人机工程仿真分析能够在设计早期对产品的人机因素进行分析和评价，利用计算机建立人体和机器的计算模型，融入人体生理特征，模拟人操作机器的各种动作，进而将人、机相互作用的动态过程可视化，通过结合人机工程学的各种评价标准和算法，对产品(机)的人机因素进行量化分析和评价。人机工程仿真分析一方面可以大大降低产品开发的成本，为产品创新提供强有力的支持；另一方面，可以大大降低一些危险性产品的测试风险。 虚拟人体模型是人机工程仿真分析的核心，它作为真实人体的替身，是检验和分析产品设计方案中人机工程性能的关键。现有人体模型在人体数据支持，模型精度以及功能应用等方面都存在很多缺陷，使其无法满足越来越高级、复杂的人机工程分析、评价算法的需要。论文以人机工程仿真设计、分析、评价为目的，建立了由几何模型、运动模型、内力模型和外力模型四部分组成的虚拟人体生物力学模型，并围绕上述需求展开相关理论、技术和方法的研究工作。 论文首先讨论了面向人机工程仿真的人体几何模型的建模方法，从人体解剖学出发，通过对人体的动、静态测量数据进行参数化，以及对关节运动约束机制的改进，使模型具有更详尽的人体数据咨询能力；为了更好地支持肌肉的生物力学分析，探讨了骨骼肌的几何建模方法。 针对人机交互过程中上述几何模型的运动建模问题，论文给出了两种互补的方法。一种方法是对传统的基于逆向运动学的改进，以快速、有效地调整人体姿势为目的，给出了基于人体生理约束的IK算法及基于时空约束的IK改进算法；另一种方法以运动捕获技术为基础，针对简单骨骼到复杂骨骼的运动重定向情况，提出了新概念及新方法来处理重定向过程中产生的关节冗余问题。从而达到精确地再现人体动作，保证人机分析的准确性的目的。 为了对骨及关节的受力情况进行人机分析，论文建立了针对人体各环节的骨及关节的外力和扭矩的求解模型，亦称为外力模型。以多刚体简化模型和人体生物力学参数为建模基础，以逆向动力学及多刚体系统动力学为理论依据，针对动、静态两种情况，分别给出了基于生物力学的外力求解模型。 以实现肌肉施力分析为目的，论文讨论了肌肉在动、静性收缩两种情况下，更多还原

【Abstract】 Ergonomics simulation technology, which gives the industrial designer the opportunity of including ergonomic factors at an early stage in product development, has become more and more important recently. As a main research content of computer aided conceptual design, researchers all over the world are absorbed in it.Instead of a real man, virtual human model can be used to improve the ergonomic quality in advance. Many complicated ergonomics evaluation methods are based on the position and status of human bones and muscles and so on. However, the current models can not satisfy the requirements of them. A more real and accurate virtual human model is needed.Current animation methods which focus more on visual effects than physiological reasonability are more appropriate for entertainment other than ergonomics. Furthermore, it is deficient in sustaining the biomechanical analysis with bone, joint and muscle. Accordingly, this dissertation proposed a virtual human biomechanical model which is composed of geometric model, motion model, internal muscle force model and external force dynamic model.In the dissertation, we first discussed the virtual human geometric modelling methods. To meet the requirement of ergonomic simulation, and for parameterized the muscle force, we presented a musculoskeletal model with the ability of representing both static and dynamic anthropometric parameters in order to make the model be any man. Based on anatomy knowledge, we also discussed the algorithms of joint motion constrains.In order to get a more accurate human machine interactive animation, we took an in-depth study both on inverse kinematics and motion capture techniques. We proposed a method which used constraint-based inverse kinematics method to retarget motion between two virtual human models with different architectures, keeping space-time constraints and human physiological constraints at the same time. Then different motion capture data can be used to get a real ergonomics emulation.更多还原