A Python rewrite of FEBOL.jl by Louis Dressel. It aims to also be a general framework for simulating UAV-based localization of a single target. The user can choose from a variety of sensors, filters, and policies (or make his own). The package was created to better interface with reinforcement learning packages created in Python.
Currently provides:
- particle filter with stratified resampling
- discrete (histogram) filter
- bearing only sensor
- FOV sensor
- various cost models, incorporating entropy, covariance, distance, etc.
- a policy class that allows for creation of seeker and target policies
- a search domain for the seeker and target to live in
Also check out deep-drone-localization for an implementation of DQN that works with multiple inputs, and a gym environment that uses all the stuff from PyFEBOL.
pip install PyFEBOL