Comparative Study between Neural Network and MHSE in State Estimation for Skid-steer Mobile Robots

Authors

  • Anna R. S. Ferreira Mechanical Engineering Department, Pontifical Catholic University of Rio de Janeiro, RJ
  • Elias D. R. Lopes Mechanical Engineering Section, Instituto Militar de Engenharia, RJ
  • Vivian S. Medeiros Mechanical Engineering Department, University of São Paulo, EESC, São Carlos
  • Helon V. H. Ayala Mechanical Engineering Department, Pontifical Catholic University of Rio de Janeiro, RJ
  • Marco A. Meggiolaro Mechanical Engineering Department, Pontifical Catholic University of Rio de Janeiro, RJ

Keywords:

State estimation, Skid-steer robot, MHSE, Neural network, EKF

Abstract

The present work addresses the state estimation of a skid-steer mobile robot using a Moving Horizon State Estimation (MHSE) approach and the approximation of this estimator with a neural network. Initially, the results of the MHSE are compared with different prediction horizons, along with an extended Kalman filter, yielding results that demonstrate the robustness of the MHSE. Subsequently, the outputs of this estimator are used to train a neural network, which will be employed to replace the estimator. The results showcase the potential of this neural network, as it would eliminate the computationally expensive optimization process in the state estimation.

Downloads

Published

2024-10-18

Issue

Vol. 4 No. 1 (2024): CBA2024

Section

Articles