Active SLAM Based on Information Theory

Information Theoretic Exploration with Gaussian Process (GP) and Bayesian Optimization (BO)

Overview

Here is a simulation of infomation-theoretic exploration with python. ROS-based method is released by Robust Field Autonomy Lab. To make the principle and mechanism more clear for active slamer, I reproduce it with python and revise some functions.

Related Work

Some basic knowledge is suggested for understanding this work.

• Gaussian Process: https://krasserm.github.io/2018/03/19/gaussian-processes/
• Bayesian Optimization: http://krasserm.github.io/2018/03/21/bayesian-optimization/
• Information-Theoretic Exploration with Bayesian Optimization: https://personal.stevens.edu/~benglot/Bai_Wang_Chen_Englot_IROS2016_AcceptedVersion.pdf

Source Code

• Jupyter Notebook
• 

Results

3 pictures from each step. The first picture illustrates the candidates. The second picture illustrates the top 5. The third picture illustrates the chosen next location and its receptive field.

Authors

Junyi Ma, Beijing Institute of Technology

Acknowledgement

• Source Code from Robust Field Autonomy Lab
• Information-Theoretic Exploration with Bayesian Optimization