A particle filter and SVM integration framework for fault-proneness prediction in robot dead reckoning system

  • Authors:

  • Lingli Yu;Min Wu;Zixing Cai;Yu Cao

  • Affiliations:

  • School of Information Science and Engineering, Central South University, Changsha, Hunan, China;School of Information Science and Engineering, Central South University, Changsha, Hunan, China;School of Information Science and Engineering, Central South University, Changsha, Hunan, China;School of Information Science and Engineering, Central South University, Changsha, Hunan, China

  • Venue:
  • WSEAS TRANSACTIONS on SYSTEMS
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper proposes an integrated framework for fault prediction in the robot dead reckoning system. The integrated framework is built by particle filter and support vector machine (SVM). On the basis, the weighted fault probability parameters can be extracted to train the prediction model. Different from the traditional particle filter fault prediction model, the proposed framework can overcome difficulties of the empirical threshold setting for decision-making. On the other hand, particle filter can not only estimate the system state values, but also obtain the residual errors that are yielded by comparing with the actual measured values. The average relative error is calculated to reduce its computing complexity in fault prediction process. Furthermore, an improved particle filter combined with support vector machine (PF-SVM) integration framework for fault-proneness prediction was devised in robot dead reckoning system. This framework estimates the particle filter process according to system state values and observed parameters to train SVM model. Finally the simulation experiments demonstrate that the PF-SVM integration framework can increase computing efficiency for fault-proneness prediction, and keep relatively high prediction accuracy at the same time. Besides that, the corresponding time step of malfunction can be precisely predicted.