基于融合多运动模态多视点几何的多机器人系统队形控制研究

In this research, we apply multiple view geometry theory into the tasks of coordination motion and formation control in multi-robot system to solve some present complicated problems. We first promote the theory of multiple view geometry and propose the multiple view geometry unifying multi-motion modalities, which is a uniform multiple view geometry to represent the cameras in static, linear motion and curvilinear motion, and at the same time the cameras are tracking the non-rigid motion in 3D space to compute the multifocl tensor. This multiple view geometry can competent in dynamic environment to accomplish some tasks such as 3D reconstruction, view transfer, view orientation, etc. We also propose the optimal method on multi-motion modalities multiple view geometry with which it is much easier to derive multiple view geometry and the system is not so sensitive to noise. We make use of optimal dynamic view geometry to realize the formation control of multi-robot system in complicated landform environment including formation maintenance and formation change. Without 3D reconstruction, we can figure out the position of the robots accurately in the next time instance to realize the visual navigation and coordination motion. We plan to complete this research in three years. The results will be presented by high quality papers and a real robot system.

我们首次将多视点几何理论引入多机器人系统的协调运动与队形控制的任务中，拟解决现有的一些技术难题。.我们将首先深度开发及提升现有的多视点几何理论，提出基于融合多运动模态的多视点几何理论，使得摄像机在静止、直线运动、曲线运动的状态下都可以用一种综合的多视点几何表示，同时它可以捕捉空间中的非刚体运动，进行多视点几何的计算，从而实现从普通摄像机画面中提取三维信息、视点转换、视角定位等功能。我们还将对所提出多模态多视点几何理论进行优化，以降低多视点几何的计算难度、削弱系统对噪声的敏感度、提高系统的鲁棒性。再利用优化的多模态多视点几何理论实现多机器人系统在复杂地形环境里的视觉导航，通过对机器人下一时刻的准确定位，实现精确的多机器人系统的队形控制和协调运动。.本项目拟3年内完成，以高质量的学术论文和实装应用等形式呈现。

在进行三维重构以及描述多个摄像机之间的关系时，多视点几何所起的作用尤为重要，它使得很多难题迎刃而解。我们首次将多视点几何理论引入多机器人系统的协调运动与队形控制的任务中，拟解决现有的一些技术难题。.我们首先深度开发及提升现有的多视点几何理论，提出基于融合多运动模态的多视点几何理论，使得摄像机在静止、直线运动、曲线运动的状态下都可以用一种综合的多视点几何表示，同时它可以捕捉空间中的非刚体运动，进行多视点几何的计算，从而实现从普通摄像机画面中提取三维信息、视点转换、视角定位等功能。我们还对所提出多模态多视点几何理论进行优化，以降低多视点几何的计算难度、削弱系统对噪声的敏感度、提高系统的鲁棒性。再利用优化的多模态多视点几何理论实现多机器人系统在复杂地形环境里的视觉导航，通过对机器人下一时刻的准确定位，实现精确的多机器人系统的队形控制和协调运动。