Flexible manufacturing and human-machine cooperation have become an important development trend of robots because of the great application potential in the field of intelligent manufacturing. During the deployment of the robot, it is of great challenging to make the robot to adapt to rapidly changing working scenes, tasks and dynamic obstacles. Compound environment constraints refer to the complex workspace constraints defined by working environment and task. Constrained by the nonlinear characteristics of robot body structure, the robot control under compound environmental constraints usually adopts the method of pre modeling and off-line programming, which has some shortcomings, such as low planning efficiency and poor environmental adaptability.
The GIIM Robotics team, Guangdong Academy of Sciences, proposed a flexible modeling method and real-time kinematic control method manipulators in compound constraint environment. This method proposes a feasible space modeling method based on level set function. By describing the robot workspace with "soft limit", and abstracting the composite working environment with a group of level set functions, the modelling method could obtain the robot workspace model under analytical constraints. Then, this method takes the neural network with strong parallel computing ability as the basic architecture of real-time kinematic control, and adjusts the neural network structure according to the robot model information, which can greatly reduce the neural network structure and improve the computational efficiency and convergence of the neural network.
This method has been preliminarily applied in the field of logistics and storage. The application verification of special-shaped box rapid loading in SF express station, Dongguan shows that this method has the characteristics of rapid adaptation to changing environment, high motion control accuracy and good real-time performance.
Fig.1 Real time kinematic control scheme of robot in compound constraint environment
This research has been published in the international authoritative journal IEEE Transactions on industrial electronics, with Dr. Zhihao Xu as the first author and GIIM, Guangdong Academy of Sciences as the first affiliation.
Paper Information：Z. Xu, X. Zhou, H. Wu, X. Li and S. Li, (2021). Motion Planning of Manipulators for Simultaneous Obstacle Avoidance and Target Tracking: An RNN Approach with Guaranteed Performance. IEEE Transactions on Industrial Electronics. doi: 10.1109/TIE.2021.3073305.