Safety of the cooperative screw assembly must be ensured through robot awareness with AI-based volumetric representation method of the workspace. Prediction of potential human movements ensures efficient and safe human-robot collaboration (HRC).
To develop an efficient HRC application with dynamic responsiveness and safe collision avoidance.
Human poses are tracked and predicted using different neural network architectures such as CNN, DNN, RNN and Autoencoder. The AI-based system learns motion predictions via a long short-term memory (LSTM) architecture, which memorizes workflows from the motion databases. Four RGB-D cameras are used for 3D workspace monitoring and GPU-Voxels for virtualization.
Real setup can be used as TestBed for use-case specific algorithms and can be transferred to various industrial applications.