Multimodal Stacked Bidirectional LSTM 구조를 활용한 거리 센서 기반의 3D 위치 인식

Abstract

In this paper, we propose the multimodal stacked bidirectional LSTM that takes range-only measurements as input and outputs robot's position by end-to-end mapping. Out proposed neural networks receive multimodal range measurements. and the experiments show that the performance of our proposed method is better than that of Monte Carlo Localization(MCL)-based localization and stacked bidirectional LSTM architecture with fewer parameters.