A Tensor‐based Regression Approach for Human Motion Prediction

Collaborative robotic systems will be a key enabling technology for current and future industrial applications. The main aspect of such applications is to guarantee safety humans. To detect hazardous situations, commercially available rely on direct physical contact the co-working person. further advance this technology, there are multiple efforts develop predictive capabilities systems. Using motion tracking sensors pose estimation combined with adequate models, potential episodes collisions between humans robots can predicted. Based provided information, system avoid by adjusting speed or position. A approach perform human prediction machine learning methods like artificial neural networks (NNs). In our approach, patterns past seconds used predict ones applying linear Tensor-on-Tensor Regression model, selected according similarity measure sequences obtained dynamic time warping (DTW). For test validation proposed pseudo assembly tasks were recorded capture system, providing unique traceable Cartesian coordinates ( x , y z ) $(x, y, z)$ each joint. repetitive motions associated tasks, whose data vary significantly in length have highly correlated variables, has been achieved real time.