Trajectory Planning Optimization of Robotic Arm Based on Mixture of Gaussian Processes

In recent years, robotic arm trajectory planning plays an important role in enabling smooth, precise and efficient motion in robotic applications and industrial automation. Methods like Particle Swarm Optimization (PSO) gained attention and employed widely to perform optimization of trajectory planning in robotic arm. However, this method faced challenges due to limited trajectory smoothness, limited global exploration and premature convergence which leads to energy inefficiencies and joint impact forces. To overcome these challenges, a novel method, Mixture of Gaussian Processes for Trajectory Optimization (MGPTO) is proposed for efficient and smooth trajectory planning of robotics' arm. This method designs the robotic arm trajectories as a Mixture of Gaussian Processes (MGPs) to ensure smooth and energy efficient motion planning with continuous time trajectory representation. Furthermore, pose planning and velocity planning is also integrated to determine initial trajectory. Finally, the MGPTO utilizes probabilistic inference and stochastic gradient-based updates to optimize the trajectories. The proposed MGPTO achieved better results with a reduced execution time of 2.187s after optimization, outperforming the existing Robot Trajectory Planning Particle Swarm Optimization (RTPPSO).