网络机器人系统协同定位、标定与建图问题解耦及算法实现

项目名称： 网络机器人系统协同定位、标定与建图问题解耦及算法实现

项目编号： No.61305113

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 吴培良

作者单位： 燕山大学

项目金额： 25万元

中文摘要： 网络机器人系统是智能空间传感器网络与服务机器人技术的交叉领域。机器人定位、传感器网络标定与环境建图是网络机器人系统自身状态估计和周围环境感知的三个基本问题，是其提供高效智能服务的基础。本申请项目提出网络机器人系统协同定位、标定与建图的概念，分析该问题的耦合关系并加以分解，展开由简到繁的递进式研究。首先，由协同定位与标定双耦合问题入手，分析其联合条件概率表示，探讨基于贝叶斯公式和马尔科夫特性的解耦策略，及各解耦项在Rao-Blackwellized粒子滤波框架下的算法实现；其次，分析协同定位、标定与建图三耦合问题并设计其解耦策略；最后，提出多源信息融合意义下嵌套新型联邦滤波算法的Rao-Blackwellized粒子滤波框架，给出该框架下网络机器人系统协同定位、标定与建图的完整算法。作为网络机器人系统提供高效服务的基础性工作，本项目研究成果对网络机器人系统理论成熟及应用推广具有一定促进作用。

中文关键词： 网络机器人系统；协同定位;标定与建图；SLAM；多源信息融合；三维场景重建与物体标注

英文摘要： Network robot system is a cross research field of intelligent space sensor network and service robot. As the bases of efficient and intelligent service, robot localization, sensor network calibration and environment mapping are three basic issues for network robot system to estimate it statement and to recognize its surroundings. The concept of simultaneous localization, calibration and mapping is raised in this project, and their coupled relations are analyzed and decomposed. The researches will be carried out in the following simple-to-complex way. Firstly, the double coupling of simultaneous localization and calibration is chosen, its joint conditional probability distribution will be analyzed and decoupled according to Bayesian and Markov properties, and the two decoupled items will be implemented under the framework of Rao-Blackwellized particle filter. Secondly, the tripe coupling of simultaneous localization, calibration and mapping will be analyzed and then the synergetic decoupling strategy will be designed. Finally, in order to fuse multi-information, two novel algorithms of federated particle filter and federated EKF filter will be designed and embedded into the Rao-Blackwellized particle filter framework. Under the framework the whole algorithm will be given for simultaneous localization, calibration

英文关键词： network robot system；simultaneous localization; calibration and mapping；simultaneous localization and mapping；multisource information fusion；3D scene reconstruction and object labelling