Learned Manipulation at Unconstrained Contacts Does Not Transfer across Hands
نویسندگان
چکیده
Recent studies about sensorimotor control of the human hand have focused on how dexterous manipulation is learned and generalized. Here we address this question by testing the extent to which learned manipulation can be transferred when the contralateral hand is used and/or object orientation is reversed. We asked subjects to use a precision grip to lift a grip device with an asymmetrical mass distribution while minimizing object roll during lifting by generating a compensatory torque. Subjects were allowed to grasp anywhere on the object's vertical surfaces, and were therefore able to modulate both digit positions and forces. After every block of eight trials performed in one manipulation context (i.e., using the right hand and at a given object orientation), subjects had to lift the same object in the second context for one trial (transfer trial). Context changes were made by asking subjects to switch the hand used to lift the object and/or rotate the object 180° about a vertical axis. Therefore, three transfer conditions, hand switch (HS), object rotation (OR), and both hand switch and object rotation (HS+OR), were tested and compared with hand matched control groups who did not experience context changes. We found that subjects in all transfer conditions adapted digit positions across multiple transfer trials similar to the learning of control groups, regardless of different changes of contexts. Moreover, subjects in both HS and HS+OR group also adapted digit forces similar to the control group, suggesting independent learning of the left hand. In contrast, the OR group showed significant negative transfer of the compensatory torque due to an inability to adapt digit forces. Our results indicate that internal representations of dexterous manipulation tasks may be primarily built through the hand used for learning and cannot be transferred across hands.
منابع مشابه
Learning Dextrous Manipulation Skills for Multiingered Robot Hands
We present a method for autonomous learning of dextrous manipulation skills with robot hands. We use heuristics derived from observations made on human hands to reduce the degrees of freedom of the task and make learning tractable. Our approach consists of learning and storing a few manipulation primitives for a few prototypical objects and then using an associative memory to obtain the require...
متن کاملLearning Dextrous Manipulation Skills for Multi ngered Robot Hands
We present a method for autonomous learning of dextrous manipulation skills with mul-tiingered robot hands. We use heuristics derived from observations made on human hands to reduce the degrees of freedom of the task and make learning tractable. Our approach consists of learning and storing a few basic manipulation primitives for a few prototypi-cal objects and then using an associative memory ...
متن کاملLearning Dextrous Manipulation Skills Using Multisensory Information
In this paper we present a method for autonomous learning of dextrous manipulation skills with multiin-gered robot hands. We use heuristics derived from observations made on human hands to reduce the degrees of freedom of the task and make learning tractable. Our approach consists of learning and storing a few basic manipulation primitives for a few prototypical objects and then using an associ...
متن کاملLearning dextrous manipulation skills using the evolution strategy
This paper presents an approach based on the evolution strategy for autonomous learning of dextrous manipulation primitives with a dextrous robot hand. We use heuristics derived from observations made on human hands to reduce the degrees of freedom of the task and make learning possible. Our system does not rely on simulation; all the experimentation is performed the 16-degree-of-freedom Utah/M...
متن کاملA Machine Learning Approach to Dextrous Manipulation
We present an approach for autonomous learning of dextrous manipulation skills. We use heuristics derived from observations made on human hands to reduce the degrees of freedom of the task and make learning tractable. Our approach consists of learning and storing a few basis manipulation primitives for a few prototypical objects and then using a nearest-neighbor method to compute the required p...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره 9 شماره
صفحات -
تاریخ انتشار 2014