Environment and Human Behavior Learning for Robot Motion Control
Date Issued
2008
Date
2008
Author(s)
Yu, Yueh-Chi
Abstract
THE Nearness Diagram (ND) method provides a reactive algorithm for robot motionontrol. It uses a decision-tree to classify the environment into several situations. mapping function is used to generate the control commands from the situations.owever, the decision-tree and the mapping function are pre-defined andany parameters need to be manually tuned. Besides, the generated path is not humanlike.he imitation learning method is an approach that aims to make robot behave as a human.t is based on the Markov decision process (MDP) which is a framework for modelinghe environment. In the imitation learning method, it tries to extract the reward function inDP under given human’s control behavior. Then, the reward function is used to generatehe control command which imitates human’s behavior. Unfortunately, the true rewardunctions in their mind are hard to describe for general users. Thus, we have difficulty onomparing the learned reward function and the ground truth.n this thesis, we combined the ND method and the imitation learning method. Weo not use a pre-defined decision tree to classify the environment in the ND method. Also,e do not solve the reward function in the imitation learning method. Instead, we try toind a mapping from the environment information to the human’s control behavior.ur system is simply described below. Several users are asked to control the robot atirst. Then, the environment information and users’ control data are gathered as trainingata. The incremental K-means method is used to classify the training data into differentituations, such as straight or turns. We use the concept of scale-invariant feature transformSIFT) in computer vision. A SIFT-like temporal feature is proposed to mark the differentituations and try to eliminate noise. The Adaptive Boosting (AdaBoost) algorithms applied to train one classifier for each situation. Finally, a nearest neighbor controller isroposed to generate the control command.
Subjects
environment
human behavior
learning
robot
control
Type
thesis
File(s)![Thumbnail Image]()
Loading...
Name
ntu-97-R95922030-1.pdf
Size
23.32 KB
Format
Adobe PDF
Checksum
(MD5):16b390b7f57949e5f6f7598bf99ed343