Learning elementary movements jointly with a higher level task

摘要: Many motor skills consist of many lower level elementary movements that need to be sequenced in order achieve a task. In learn such task, both the primitive as well higher-level strategy acquired at same time. contrast, most learning approaches focus either on combine fixed set options or just single options. this paper, we discuss new approach allows improving performance actions while pursuing higher The presented is applicable wider range skills, but employ it for games where player wants improve his individual game still performing game. We propose using Cost-regularized Kernel Regression and form Policy Iteration. two are coupled by their transition probabilities. evaluate side-stall-style throwing simulation with real BioRob.