利用學習機制進行機器人運動指令之簡化與調變

Abstract

經由觀察我們可以得知人類肢體運動並不是非常精確，這個現象引發一種 假說：人類神經肌肉控制系統可能以運動精確度的代價來換取運動指令的 簡化。由這個假說，我們提出了一個機器人學習機制可用以進行機器人運 動的精確度與運動指令簡化的交換。一組可以執行人類運動軌跡精確追蹤 的原始複雜機器人運動指令在經過簡化之後，可以利用結構比較簡單、需 要記憶體空間比較少的學習機制加以儲存與執行之。另外，這個學習機制 也可以在不需要系統動態重新計算之下執行運動指令的調變，完成相似但 具有不同運動距離與速度的機器人運動軌跡。以機器人執行人類運動的實 驗驗證了我們所提出的學習機制在實現運動精確度與指令簡化之間的交換 是可行的。 It has been observed that human limb motions are not very accurate, leading to the hypothesis that the human motor control system may have simplified motion commands at the expense of motion accuracy. Inspired by this hypothesis, we propose a learning scheme that trades motion accuracy for motion command simplification. When the original complex motion commands capable of accurate motion tracking are simplified into those in simple forms, the simplified motion commands can that be stored and manipulated by using learning mechanisms with simple structures and scanty memory resources, and they can be fast and smoothly executed. In addition, this learning scheme can also perform motion command scaling, so that simplified motion commands be provided for a number of similar motions of different movement distances and velocities without system dynamics re-calculation. Experiments based on robot motions are reported that demonstrate the effectiveness of the proposed learning scheme in implementing this accuracy-simplification tradeoff.