协作网络因子图中粒子形式的信息传递算法

Message Passing Algorithm through Particles on Factor Graph in Cooperative Positioning Network

和积算法结合因子图可以用分布式方式实现协作定位。和积算法是一种信息传递算法,然而在非线性、非高斯环境下用参数法实现信息传递误差较大,不能满足定位需要,提出一种算法用粒子形式来实现信息传递。因子图中的信息计算包括求和与求积两个过程。粒子形式的信息传递算法利用重要性采样得到求和信息,利用吉布斯采样得到求积信息。提出的算法能简化复杂的网络节点的联合后验概率。与基于参数的信息传递算法相比,粒子形式表示方法提高了在非线性、非高斯环境下的定位精度。

Combining the sum-product algorithm with the factor graph can achieve cooperative positioning by a distributed manner. The sum-product algorithm is a message passing algorithm. However, the parameter method cannot meet the positioning needs because of the large error in the non-linear non-Gaussian environment. Therefore, this paper presents an algorithm to achieve message passing in the form of particle. The calculation of the message in the factor consists of summation and quadrature processes. The proposed method obtains the sum of messages by importance sampling, obtains the product of messages using Gibbs sampling. Particle-based information passing algorithm can simplify the complex joint posterior probability distribution. Compared with message passing algorithm based on parameters, themethod based on particles improves the positioning accuracy in non-linear, non-Gaussian environments.