Modeling human skills to control dynamic systems

摘要

There are many successful industrial applications of control methods based on imparting human skills. However, all applications to date are task-lever controllers, i.e., when switching from one task to another, they need to be retrained. We do not focus simply on task-level human learning control, but on how to produce a general control law in which we can input control targets, i.e., to enable us to change control targets for which controllers have not been trained. Using a dynamically stabilized but statically unstable single-wheel mobile robot - Gyrover - we use this new method to control and thus confirm the validity and efficiency of the method.