Hybrid Position Forecasting Method for Mobile Robot Transportation in Smart Indoor Environment

Abstract

Indoor mobile robot position forecasting can improve safety and robustness of the robot navigation systems in indoor environment. In this paper, a novel hybrid computing framework for indoor mobile robot position forecasting is proposed, namely EWT-MOFEPSO-MRMRMI-ORELM. The proposed model consists of three parts, decomposition, feature selection and forecasting. The EWT (Empirical Wavelet Transform) algorithm is utilized to decompose raw series into several more predictable sublayers. For each sublayer, the MRMRMI (Maximum Relevancy Minimum Redundancy Maximum Interaction) model optimized by MOFEPSO (Multi-objective Feasibility Enhanced Particle Swarm Optimization) is applied to generate Pareto set of candidate feature set. The best feature set is selected as the one with minimum forecasting error in validation data. The selected best feature set is used as input of the ORELM (Outlier Robust Extreme Learning Machine) model to generate forecasting value. The ORELM can prevent adverse effect of outlier. Two experiments are carried out to verify the effectiveness of the proposed computing framework. The results indicate that (a) the proposed hybrid computing framework has excellent forecasting performance, (b) The EWT, MOFEPSO-MRMRMI can enhance performance of the ORELM significantly; and (c) the proposed forecasting method can guarantee the safety transportation of mobile robots in smart indoor environment.