Dynamic sensorimotor model for open-ended acquisition of tool-use
| Autor: | Raphael Braud, Philippe Gaussier, Alexandre Pitti |
|---|---|
| Rok vydání: | 2016 |
| Předmět: |
0209 industrial biotechnology
Sequence Social robot Computer science business.industry Control (management) Sensory system 02 engineering and technology Task (project management) 020901 industrial engineering & automation Reachability 0202 electrical engineering electronic engineering information engineering Robot 020201 artificial intelligence & image processing Artificial intelligence business Robotic arm |
| Zdroj: | ICDL-EPIROB |
| DOI: | 10.1109/devlrn.2016.7846834 |
| Popis: | The learning of sensorimotor primitives in an open-ended manner is important to achieve all the possible tasks a robot can do, even those never experienced before. In this short paper, we propose a neural architecture called Dynamic Sensorimotor Model (DSM) (1) that learn co-variation rules between sensors and motors for sensorimotor prediction, (2) use these predictions for action planning. This archtecture can achieve off-the-shelf a tool-use task with a tool never used before. By simulating possible motor activities and alternative sensory pattern, our DSM is able to control the robot and to create sub-goals on-the-fly. Our experiment consists on a 6-degree of freedom robotic arm and a camera. At first, the robot learns to predict sensory variations through its motor activities and a given sensory pattern, which can serve for reaching. After experiencing the tool in its grip and the new visuo-motor relationship, the robot is capable to use this rule and to generate a sequence online when a target is farther than its reachability area in order to reach it. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|