Model Predictive Path Integral with Integrated Human Pose Prediction Network for Robotic Arm Motion Planning

With the development of robotics technology, robots began to replace humans to carry out various types of repetitive operations. At the same time, robotic arms and humans in the same workspace to carry out collaborative work has become an important trend for future development. In the fixed program state, the traditional robotic arm executes a fixed trajectory, when the human body appears in the workspace of the robotic arm, it is often difficult to adjust the trajectory in time to avoid collision with the human body. Due to the high randomness of the human body movement, the conventional path planning method is difficult to adapt to the highly dynamic and strongly random human body collision avoidance task in the human-machine collaboration scenario. For this reason, this paper designs a human-machine collaborative collision avoidance method based on human pose prediction, by predicting the human body pose in the future moment, so that the robotic arm avoids the collision planning in advance for the predicted human body pose.