Multi-Sensor Fusion Method Based on Artificial Neural Network for Mobile Robot Self-Localization

• Carlos Magrin UFPR
• Eduardo Todt UFPR

Resumo

This paper presents a hierarchical sensor fusion (HSF) method with an artificial neural network (ANN) to solve the problem of mobile robot self-localization with sonars octagon, digital compass, and wireless network signal strength measure to determine the location of an autonomous mobile robot. The multilayer perceptron (MLP) is used with supervised learning, backpropagation technique, to train the network in hierarchical fusion step and determine the robot localization in a map. In order to validate this work, a comparison between the HSF methods, artificial intelligence, and the matching algorithm, using the same training and testing UFPR-RSFM Dataset. Finally, the HSF method with artificial intelligence technique can determine the robot localization in a different indoor environment, using low-cost sensors, and support the relevance of hierarchical sensor fusion in mobile robot localization.

Palavras-chave: Robot sensing systems, Neural networks, Sensor fusion, Sonar measurements, Mobile robots