The paper looks at using neural networks to observe and learn a small number of human reaching motions that have been scaled to a robot body. The neural network is then able to generalise these examples for new requirements, so the robot can generate new human-like trajectories on the fly.
Below is a word cloud of the paper, which was generated by Wordle.
This was the most recent piece of work I completed for my PhD on human motion synthesis for manipulator robots.
