Generating Constant Screw Axis Trajectories With Quintic Time Scaling For End-Effector Using Artificial Neural Network And Machine Learning

In this work, the application of Neural Networks (NNs) and Machine Learning (ML) algorithms within the trajectory generation framework is presented. The main objective is to demonstrate that the trajectory of an end-effector of a multilink robotic system can be obtained by generating a path via the Bézier curve, imposing a finite jerk constraint, and using NN and ML algorithms to obtain the Constant Screw Axis (CSA) trajectory closest to the chosen path. These constraints are of importance because the former forces the jerk to be finite throughout the motion of an end-effector, which reduces the vibrations within the robotic arm or any other manipulator, and the latter smooths the trajectory in Special Euclidean (SE(3)) space. First, a Bézier curve is chosen to define the path that an end-effector should follow. Subsequently, the finite jerk constraint is imposed. Lastly, NN and ML algorithms are used to obtain the closest CSA trajectory to the desired path. Since there are multiple CSA trajectories satisfying the initial configuration and final position, NN and ML algorithms are employed to minimize the Euclidean distance between the desired path (Bézier curve) and the actual obtained CSA trajectory.