Robotic intelligence with behavior selection network for Bayesian network ensemble

Scene understanding is an important and difficult problem in intelligent robotics and computer vision. Since visual information is uncertain due to several reasons, we need a novel method that has robustness to the uncertainty. Bayesian probabilistic approach is robust to manage the uncertainty and powerful to model high-level contexts. Moreover, Bayesian network can be adapted to environment efficiently by learning. In this paper, we propose a Bayesian network ensemble technique based on behavior selection network. The method includes how to handle uncertainty based on probabilistic approach, and how to combine multiple Bayesian networks. An experiment with a mobile robot simulation presents how the proposed ensemble method works and can be used effectively.