Real-time Remote Tracking and Autonomous Planning for Whale Rendezvous using Robots

Abstract

We introduce a system for real-time rendezvous with sperm whales using an autonomous uncrewed aerial vehicle. Our system extends a model-based reinforcement learning approach that leverages in situ sensor data alongside an empirical whale diving model to inform navigation decisions. Critical challenges include real-time acoustic sensing in the presence of multiple whales, distributed communication and decision making for real-time UAV deployment, and enforcing realistic battery constraints in the decision making process. We evaluate our system with proof-of-concept field experiments at sea with sperm whales, hardware experiments on land, and simulations at scale using whale trajectories.