Improved and Scalable Online Learning of Spatial Concepts and Language Models with Mapping

  • 2019-01-04 07:36:32
  • Akira Taniguchi, Yoshinobu Hagiwara, Tadahiro Taniguchi, Tetsunari Inamura
  • 0

Abstract

We propose a novel online learning algorithm, called SpCoSLAM 2.0, forspatial concepts and lexical acquisition with high accuracy and scalability.Previously, we proposed SpCoSLAM as an online learning algorithm based onunsupervised Bayesian probabilistic model that integrates multimodal placecategorization, lexical acquisition, and SLAM. However, our previous algorithmhad limited estimation accuracy owing to the influence of the early stages oflearning, and increased computational complexity with added training data.Therefore, we introduce techniques such as fixed-lag rejuvenation to reduce thecalculation time while maintaining an accuracy higher than that of the previousalgorithm. The results show that, in terms of estimation accuracy, the proposedalgorithm exceeds the previous algorithm and is comparable to batch learning.In addition, the calculation time of the proposed algorithm does not depend onthe amount of training data and becomes constant for each step of the scalablealgorithm. Our approach will contribute to the realization of long-term spatiallanguage interactions between humans and robots.

 

Introduction (beta)

None

 

Conclusion (beta)

None