Distributed Robust Data Fusion Based on Dynamic Voting
Data association mistakes, estimation and mea surement errors are some of the factors that can contribute to incorrect observations in robotic sensor networks. In order to act reliably, a robotic network must be able to fuse and correct its perception of the world by discarding any outlier information. This is a difficult task if the network is to be deployed remotely and the robots do not have access to ground truth sites or manual calibration. In this paper, we present a novel, distributed scheme for robust data fusion in autonomous robotic networks. The proposed method adapts the RANSAC algorithm to exploit measurement redundancy, and enables robots determine an inlier observation with local communica tions. Different hypotheses are generated and voted for using a dynamic consensus algorithm. As the hypotheses are computed, the robots can change their opinion making the voting process dynamic. Assuming that at least one hypothesis is initialized with only inliers, we show that the method converges to the maximum likelihood of all the inlier observations in a general instance. Several simulations exhibit the good performance of the algorithm, which also gives acceptable results in situations where the conditions to guarantee convergence do not hold.
Mobile Robotic Systems Robust Data Fusion Distributed Algorithms Dynamic Voting Sensor Network Calibration
Eduardo Montijano Sonia Mart(i)nez Carlos Sagues
DIIS - I3A,University of Zaragoza department of Mechanical and Aerospace Engineering,Univ. of California,San Diego
国际会议
2011 IEEE International Conference on Robotics and Automation(2011年IEEE世界机器人与自动化大会 ICRA 2011)
上海
英文
5893-5898
2011-05-09（万方平台首次上网日期，不代表论文的发表时间）