Abstract | ||
---|---|---|
An Interactive Machine Learning (IML) approach for training a dribbling engine for humanoid biped robots in RoboCup competitions (Standard Platform League) is presented. The proposed dribbling approach solves two decision problems: the determination of the dribbling direction and the calculation of the walking velocities required for pushing the ball toward the desired direction. Moreover, the prediction of the position of moving balls is used for improving the dribbling performance, when it is needed to intercept a moving ball. A combination of batch and incremental learning is used for shaping the policies of the dribbling controller. Results obtained from previous RoboCup competitions, and also from specific experiments, validate the proposed methods. |
Year | DOI | Venue |
---|---|---|
2017 | 10.1007/978-3-030-00308-1_30 | ROBOCUP 2017: ROBOT WORLD CUP XXI |
Keywords | Field | DocType |
Robot soccer, Learning from demonstration, Robot behavior, Human feedback | Control theory,Decision problem,Simulation,Computer science,Ball (bearing),Incremental learning,Learning from demonstration,Artificial intelligence,Behavior-based robotics,Robot,Machine learning | Conference |
Volume | ISSN | Citations |
11175 | 0302-9743 | 0 |
PageRank | References | Authors |
0.34 | 8 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Carlos Celemin | 1 | 18 | 4.49 |
Rodrigo Perez | 2 | 0 | 0.34 |
Javier Ruiz-del-Solar | 3 | 865 | 105.92 |
Manuela Veloso | 4 | 8563 | 882.50 |