STABILITY PREDICTION OF QUADRUPED ROBOT MOVEMENT USING CLASSIFICATION METHODS AND PRINCIPAL COMPONENT ANALYSIS

Abstract

. This paper presents a new stability prediction technique of quadruped robot locomotion based on central pattern generator (CPG). In this method, proposed stability prediction is executed using classification methods and principal component analysis (PCA). The purpose of this study is to anticipate the stability of robot movement using modified controlling parameters called features and the target is stability or instability condition. In MATLABTM/SIMULINK®, 82 observations of robot locomotion are simulated using different parameters in which 62 and 20 samples are considered as train and test dataset respectively. In order to determine whether the corresponding locomotion is stable or not, machine learning (ML) techniques are applied using classification methods and PCA. To classify stability condition, six classification methods are deployed including K-nearest neighbors (KNN), support vector classifier (SVC), Gaussian Naïve Bayes (GaussianNB), logistic regression (LR), decision tree (DT), random forest (RF) by Scikit-learn, an open-source ML library in Python. The results of six classifiers are evaluated based on four metrics, such as precision, recall, accuracy and F1-score. The results demonstrate KNN and SVC have high metric values in comparison to other classifiers for stability prediction.