Designing Learning Mechanisms in Robotics

: Investigating Adaptive Phototaxis through Reinforcement Learning in Partially Observable Markov Decision Processes

Abstract

This research ventures into the crossroads of adaptive behaviour, phototaxis and reinforcement learning (RL), attempting to bridge the gap between biological phenomena and learning mechanisms. The study aims to validate the applicability of RL in the context of phototaxis under Partially Observable Markov Decision Processes (POMDPs). However, the complexity of the problem space and time constraints prevented the realisation of a learning behaviour stage. Despite these challenges, the study provides interesting insights into the interaction between adaptive phototaxis and RL. The agent's performance in a variety of scenarios highlights the potential of RL in complex behaviours and the obstacles it presents. Despite its preliminary nature, this work sets a promising course for future exploration in this wide area of research.

Date of Award	19 Jun 2023
Original language	English
Awarding Institution	University of Namur
Supervisor	Elio Tuci (Supervisor)