一种普适机器人系统同时定位、标定与建图方法

机器人定位、传感器网络标定与环境建图是普适机器人系统中三个相互耦合的基本问题, 其有效解决是普适机器人系统提供高效智能服务的前提. 本文提出了普适机器人系统同时机器人定位、传感器网络标定与环境建图的概念, 通过分析三者之间的耦合关系, 给出同时定位、标定与建图问题的联合条件概率表示, 基于贝叶斯公式和马尔科夫特性将其分解为若干可解项, 并借鉴Rao-Blackwellized粒子滤波的思想分别求解. 首先, 联合传感器网络对机器人的观测、机器人对已定位环境特征的观测以及机器人自身控制量,设计了位姿粒子的采样提议分布和权值更新公式; 其次, 联合传感器网络对机器人运动轨迹及已定位环境特征的观测,设计了传感器网络标定的递推公式; 然后, 联合传感器网络和机器人对(已定位或新发现)环境特征的观测,设计了环境建图的递推公式. 给出了完整的同时定位、标定与建图算法, 并通过仿真实验验证了该算法的有效性.

A Solution to Simultaneous Localization, Calibration and Mapping of Ubiquitous Robot System

In ubiquitous robot system, robot localization, sensor network calibration, and environment mapping are three basic issues which are coupled with each other, and solutions to these three issues are prerequisites for efficient and intelligent service of ubiquitous robot system. In this paper, the concept of simultaneous localization, calibration and mapping of ubiquitous robot system is proposed. The joint conditional probability distribution is designed to describe the coupled question, and then is decomposed into several analytic terms according to Bayesian and Markov properties. The Rao-Blackwellized particle filtering is used to solve the analytic terms. Firstly, sensor network observation of robot, robot observations of mapped environments, and robot controls are combined to deduce the proposal distribution of robot pose and formula for updating particle weight. Secondly, sensor network observations of both robot path and mapped environment are combined to deduce the recursive formula of sensor network calibration. Thirdly, both sensor network observations and robot observations of environment (localized or newly found) are combined to deduce the recursive formula of environment mapping. The whole algorithm of simultaneous localization, calibration and mapping is designed according to Rao-Blackwellized particle filtering, and its efficiency is verified by simulated experiments.