Improving Smoothness and Stability in Robotic Arm Trajectories via Optimizer-Assisted Gradient Descent

The focus of this study is to optimise the running path of a six-degree-of-freedom (6-DOF) robotic arm between two fixed points in three-dimensional space. The gradient based method combined with an optimiser including Adagrad, Adadelta and Momentum method is utilized for optimization of robot manipulator actuation efficiency with minimum energy consumption when initial and final poses of the robot manipulator are known. A gradient descent-based execution optimisation framework is presented in the paper to iteratively optimise the joint motion trajectories of the robotic arm based on different optimisers. The general aim is to further reduce execution time while enhancing the smoothness and stability of the robotic arm motion to achieve overall efficiency. The experimental results indicate that the stability of robot arm motion combination of Adam optimiser can greatly improve the efficiency of gradient descent compared with other optimizers, the balanced performance, and the path execution efficiency between two fixed positions. This paper can provide a reliable reference for path execution optimization of robotic arms in industrial automation and accuracy control.