Classifying Movement Articulation for Robotic Arms via Machine Learning

Articulation via target-oriented approaches have been commonly used in robotics. Movement of a robotic arm can involve targeting via a forward or inverse kinematics approach to reach the target. We attempted to transform the task of controlling the motor articulation to a machine learning approach. Towards this goal, we built an online robotic arm to extract articulation datasets and have used SVM and Naïve Bayes techniques to predict multi-joint articulation. For controlling the preciseness and efficiency, we developed pick and place tasks based on pre-marked positions and extracted training datasets which were then used for learning. We have used classification as a scheme to replace prediction-correction approach as usually attempted in traditional robotics. This study reports significant classification accuracy and efficiency on real and synthetic datasets generated by the device. The study also suggests SVM and Naïve Bayes algorithms as alternatives for computational intensive prediction-correction learning schemes for articulator movement in laboratory environments.