データ駆動型による空間依存的な昆虫の匂い源探索行動のモデル化

抄録

In this study, We propose a new concept to approach Chemical Plume Tracing (CPT) problems in complex environments expected in emergencies such as disasters and terrorism as data-driven models. CPT, which uses odor information to plan the robot’s movement in preparation for obstacles and noise in emergencies, has been actively studied recently as it has huge engineering value. Since Considering all the factors in a complex environment in which robots equipped with a data-driven model are expected to be deployed takes high engineering costs, it aims to reduce the cost by dividing the complex environment into element environments with different spatial characteristics, conducting CPT experiments in each element environment, building an independent model with that data, and switching between the models autonomously to determine the next step action.