Classification of Service Robot Environments Using Multimodal Sequence Data

Abstract

The usage of autonomous robots is getting increased day by day. Most of the applications are moving toward automation with the help of robots. This paper mainly focuses on service robots and understanding their working environments. A few robotic scenarios are created using Webots tool, and the data from there are collected as a sequence of images and lidar sensor values. The lidar values are collected with both single layer and multilayer. The environments are analyzed with the help of the collected data. The collected multimodal data are preprocessed in order to reduce the number of features. After that, the collected data are sorted out to suitably characterize each environment, and the machine learning techniques are applied to classify the environments. Different machine learning algorithms like Naive Bayes classifier, support vector machine, decision-tree-like random forest tree, and simple logistic regression are used for the classification, and results are compared with each other.