Developmental Neurorobotics Models of Counting and Numerosity Judgment

  • Leszek Pecyna

Student thesis: PhD

Abstract

This thesis presents novel developmental robotics models of number cognition which take into account the embodied basis of this process. The ability to perceive quantity is undoubtedly crucial for both humans and animals. Nowadays, automation is becoming increasingly important in society; it seems that the way, similar to humans, to perceive and process numbers is one of the pressing issues for the future of robotics. This is important in the context of proper interaction between robots and humans which requires machines - at least on some level - to understand complex/abstract words such as number names. This thesis takes into account the development of human number cognition and investigates how a similar process can be implemented on machines. This not only can help to build a number knowledge in robots but also to understand the mechanisms responsible for these abilities in humans. This dissertation focuses on two numerical skills: numerosity judgment and counting. The embodied effects are linked with both of these abilities. Finger number representation and usage of pointing gestures while counting will be investigated in three sets of experiments. Both of these embodied effects seem to have an important role in human development of numerical system. The first set of experiments studies the influence of finger counting on the performance of numerosity judgment. The second and the third examine the impact of pointing gestures on the counting process. Especially in this third and last experiment, the model is trained and tested in a complex procedure resembling the process of children learning to count. This model is also the most complex in its architecture and can process real images from the robot cameras. The PhD research contributes to knowledge through introduction of novel neurorobotics models of these embodied phenomena and by conducting a series of simulation experiments. In the context of numerosity judgment, the studies provide evidence on the impact of finger counting and of the importance of unsupervised preliminary training (where the model is trained without any required output, and having only visual input provided that it gets habituated with). The experiments that concern counting show that pointing gestures can significantly boost the performance of the model and that a proper training routine - composed of learning to point or to recite numbers - has a significant impact on the development of counting skills. Finally, and based on the conducted experiments, in the context of simulating of number cognition, the thesis provides detailed guidelines for future studies and model design on robotic platforms.
Date of Award2021
Original languageEnglish
Awarding Institution
  • University of Plymouth
SupervisorAngelo Cangelosi (Other Supervisor)

Keywords

  • Counting
  • Developmental Robotics
  • Numerosity Judgment
  • Pointing
  • Neural Network
  • Deep Learning
  • Number Cognition
  • iCub
  • Machine Learning

Cite this

'