Surpassing Natural Experts in Odor Source Localization Through Inverse Reinforcement Learning

I am using deep inverse reinforcement learning to learn the odor search strategy of the silkworm moth in nature and export a reward function from it to apply to the autonomous mobile robot for safety and rescue tasks.

Problem statements:

  • Finding gas source in obstacle region.
  • Learning from natural expert which is silkworm moth.

Images and Results:

Robot

VRmoth system for obtaining silkworrm moth behavior

Additional Information:

Location: Kurabayashi Laboratory, Institute of Science Tokyo, Japan.
Date: From 2022 October to 2025 September.
Context: This is my PhD Dissertation at the Institute of Science Tokyo.
Publications: Learning a Generic Olfactory Search Strategy From Silk Moths by Deep Inverse Reinforcement Learning (2022).