The Robotic Arm Velocity Planning Based on Reinforcement Learning

In order to improve the performance of the robotic arm effectively, this study established a robotic arm velocity planning model developed by artificial intelligence in the simulation system. The model not only considered the dynamic factors of the robotic arm but was also able to set different customized conditions such as machining accuracy and rotation angle. The study could be divided into three parts. First, the simulation environment was constructed with the ABB IRB140 six axes multipurpose industrial robot. To be consistent with real-world situations, a Vortex physics engine was applied to the simulation supplying varying locomotion parameters. In this research, friction, kinematics, and inertia were considered. Second, artificial intelligence was imported into the robotic arm through the establishment of connecting V-rep and Python. The proposed model was developed in the Python environment by deep deterministic policy gradients. Eventually, a design of the appropriate reward function governing the ultimate results was presented. Compared with traditional velocity planning, the proposed method can decline moving error by 0.05 degrees under the considerations involving dynamic factors in a robotic arm. Besides, the proposed velocity planning strategy could be obtained after taking the training time of one hour which can meet the demand for the time cost of the industry.