This repository contains the code for the paper Enhancing Navigational Safety in Crowded Environments using Semantic-Deep-Reinforcement-Learning-based Navigation.
Navsafe-Arena is based on arena-rosnav, which is a flexible, high-performance 2D simulator for testing robotic navigation, and pedsim_ros, which is a pedestrian simulator implementing the social force model. The agent is trained to learn an object-specific navigation behavior, keeping different safety distances towards different types of humans such as adults, children, and elders. An efficient DRL approach called CPU/GPU asynchronous A3C using curriculum learning is utilized. Within the simulation, it is assumed that the agent knows the accurate position and type of humans in the vicinity for understanding the interaction between the robot and humans. In real-world experiments, this information can be acquired using computer vision approaches.
 |
 |
 |
|---|---|---|
 |
 |
 |
| raw | static zone | dynamic zone |
We recommend starting with the start guide which contains all information you need to know to start off with this project including installation on Linux and Windows as well as tutorials to start with.
- For Mac, please refer to our Docker.
Please refer to Installation.md for detailed explanations about the installation process.
We provide a Docker file to run our code on other operating systems. Please refer to Docker.md for more information.
Please refer to DRL-Training.md for detailed explanations about agent, policy and training setups.
DRL agents are located in the agents folder.
- In one terminnal, start simulation
roslaunch arena_bringup start_training.launch num_envs:=1 #switch useDangerZone to be false if normal zone needed- In another terminal, load the pretrained agent
workon the_name_of_your_virtual_env
roscd arena_local_planner_drl && cd scripts && cd deployment
python run_agent load MLP_HUMAN_DANGER_ZONE -s scenario2 #scenario is not used but should be denoted - In the third terminal, visualize the simulation
roslaunch arena_bringup visualization_training.launch rviz_file:=human_nav