Neural Network Based Configuring of Linear Delta Robots with Reduced Energy Consumption

In this paper, a neural network based methodology of joints configuring is proposed for linear Delta robots, optimal in terms of minimizing the consumed energy. Since the techniques developed for serial robots are severely limited in parallel robotics, where many parameters affecting robot behaviour are either unstable or absent, a new purely data-driven approach has been developed and studied in the conditions of incomplete prior knowledge of robot geometric and mass properties. In the current research, the forward and inverse kinematics solvers are designed, the workspace of the robot is estimated, forward and inverse statics problems are resolved, and the most economical joint configurations are found with the help of the neural network robot model validated experimentally on the linear Delta robot EXPT-45-E1. The method offered in this study can be used in many other robotics applications, especially for the pick-and-place operations, where energy saving comes out on top. Key words. Energy consumption, linear Delta robots neural network based model, pick-and-place robotics.