Smart Generation of Human Motions for Designing Working Spaces

本计画发展一套精确的人体动作与行为模拟与动画机制，以从人体工程学的观点评估与检视设施环境的效率与舒适度。我们已经开发了运动合成机制，通过插值的运动样本，使用一种精确的预测方法称为克里格，这是在地质统计学中提出的，并已证明其优越性，在预测精度的径向基函数的基础上的方法与客观的分析。我们还证明了该方法对于预测人类足迹避开障碍物的有效性。智能生成运动数据需要从大量运动数据集中充分选择运动片段，因此我们通过将运动数据集中包含的代表性姿势图标分布到二维显示器上来开发用于运动数据检索的图形用户界面。作为人类行为预测机制的一部分，基于城市工程或认知心理学中发现的知识或信息，我们已经开发了步行行为的模拟。该仿真在宏观上能够很好地模拟不同空间密度下的人体行走过程，在微观上能够生成反映人体心理特征的合理的避碰行为。基于上述技术，本项目开发了基于Web3D可视化工具的运动序列仿真中间件，从估算工作时间的程序手册和X3D提供的设施数据。

This project developed an accurate simulation and animation mechanisms of human motions and behaviors for estimating and inspecting the efficiency and comfortableness of facilities environment from the viewpoint of human engineering.Simulations of human motions often require the predictive generation of motions for the different conditions of motion capture session. We have developed the motion synthesis mechanism through the interpolations of motion samples by using an accurate prediction method called kriging, which was propose in geostatistics, and have demonstrated its superiority in prediction accuracy over the method based on the radial basis functions with an objective analysis. We also demonstrated the effectiveness of this method for predicting human foot prints avoiding obstacles.Smart generation of motion data requires to adequately select motion clips from huge amount of motion data sets, and we therefore have developed the graphical user interface for motion data retrieval by distributing representative posture icons included in motion data sets onto a two dimensional display.As a part of prediction mechanism of human behaviors, we have developed the simulation of walking behaviors on the basis of the knowledge or the information discovered in urban engineering or cognitive psychology. This simulation can well-approximate the flow of human walking for various spatial densities of walkers in a macro level, and can generate plausible collision avoiding behaviors reflecting psychological features in a micro level.Based on the above mentioned technologies, this project has developed the middleware for simulating a sequence of motions with visualization tools of Web3D, from the procedural manual for estimating working time and facilities data given with X3D.