| Definition | Simulating and optimizing odor-based target tracking strategies in partially observable settings |
| No of environments | 1 |
OTTO is a Python package designed to learn, evaluate, and visualize strategies for odor-based target tracking in dynamic, partially observable environments. It simulates the source-tracking POMDP (Partially Observable Markov Decision Process) problem, where agents, such as robots or animals, must find a hidden odor source in a turbulent flow, receiving only partial observations to guide their search.
The “olfactory-search” module is a simplified implementation of this problem, providing a Gymnasium environment for studying olfactory search strategies. It serves as a foundation for testing various policies, such as infotaxis, by offering a flexible environment where agents must search for an odor source under partial observability, making it ideal for reinforcement learning applications focused on real-world search problems.