/cit_prob_robotics_19

The python notebook of the homework.

Primary LanguageJupyter Notebook

CIT_PROB_ROBOTICS_19

This repository is created by Shuhei Kozasa, kozasa0820jps@gmail.com

This repository is for the final report for the Probabilistic Robotics class 2019. The task is to

  • Create a grid world
  • Move the agent to the goal by using a reinforcement learning method

Results

The image Result_01.png is the result of the Sarsa version. Here, alpha = 0.5, gamma = 1.0 The agent did reach the goal but it looks like the path isn't that efficient.

result_01

The image Result_02.png is another result of the Sarsa version. Here, alpha = 0.8 and gamma = 1.0 This time, I set an obstacle and the agent managed to go around it.

result_01

Future works

Doing this task was interesting. It is my first time trying out something related to machine learning.

I would love to get back to this code once I finish my thesis paper.

Reference

Source code from the class in 2017 https://github.com/ryuichiueda/probrobo2017/blob/master/09.ipynb