omtodkari / uas-obstacle-avoidance

This is the modification of the existing repository under some author

UAS Obstacle Avoidance Integrating Safety Bound with Reinforcement Learning

Author: Jueming Hu, Arizona State University

Email: [email protected]

This module demonstrates UAS obstacle avoidance with a risk-based safety bound using reinforcement learning.

The detailed information can be found here.

  @inproceedings{hu2020uas,
    title     = {UAS Conflict Resolution Integrating a Risk-Based Operational Safety Bound as Airspace Reservation with Reinforcement Learning},
    author    = {Hu, Jueming and Liu, Yongming},
    booktitle = {AIAA Scitech 2020 Forum},
    pages     = {1372},
    year      = {2020}
  }

Files

• main_RL_train_result.ipynb: main file for training and visualization.

• SafetyBound.py: obtain the size of safety bound.

• geometryCheck.py: check potential collision among different shapes.

• ObstacleAvoidanceENV.py: define RL environment, including transition model and reward function.

• plotting.py: visualization of RL training process

• draw.py: visualization of learned trajectories

• Q_learning.py: Q-learning algorithm

Required python packages

• gym
• itertools
• matplotlib
• numpy
• pandas
• sys
• matplotlib
• collections
• shapely
• math
• random
• sys