Development of a Control System Based on a Deep Generative Model for a 6-DoF Robotic Manipulator

Abstract

The functioning of robotic intelligent complexes in a dynamic environment for the effective solution of object manipulation tasks is a challenging scientific problem. Intelligent robotics involves not only the performance of tasks by a robot manipulator, but also safe interaction with humans, which is why collaborative robots are increasingly in demand. The efficiency of robot manipulators in performing tasks depends on the hardware architecture of the robotic device, but to a greater extent on the control system. The use of dynamic models based on physical parameters of the environment allows for the implementation of a high-quality robot control system capable of generalization, but this approach requires complete information about the environment. The use of machine learning models in the development of a robot manipulator control system allows the robot's actions to be generated based on the parameters of the environment perceived by sensors. This approach allows complex behavior algorithms to be implemented in real time in a changing robot operating environment. This paper proposes an architecture for a robot manipulator control subsystem based on a deep generative model.