Planning with learned motions

Oftentimes, robotic motion planning algorithms generate very bizarre paths that don’t make sense to people. For my Cognitive Robotics (6.834J) final project, I developed an algorithm that seeks to generate motions that would be intuitive/predictable, by piecing together motions that have been demonstrated by a human. It uses motions learned by demonstration as the primitives in RRT* instead of straight line segments.

More details coming soon!