• ai
• biosensors
• computer vision
• convolutional neural network
• cricket
• deep learning

Animal-robot interaction is an interesting topic that have emerged very recently and it is still under development, thus offering many areas of research that has not yet been investigated. The main contribution of this thesis is to introduce artificial intelligence methods in the loop in animal-robot interaction. Specifically, this thesis deals with the classification of the interaction that occurred between an insect (Orthoptera: Ensifera) and two chemical agents (sugar and ammonia) through the readings of the animal’s antennae.
This was carried out in two steps: first, a deep neural network was developed and trained, using state-of-the art computer vision techniques, to locate the position of specific parts of the cricket in order to generate reliable numeric sequences; subsequently, these obtained sequences were used to identify the type of interaction that happened, using one-dimensional convolutional deep learning techniques.