Arena-Rosnav

A modular ROS 2 (Humble) platform for researching and benchmarking autonomous robot navigation in 2D and 3D simulated environments. It supports classical planners (Nav2), deep-RL planners (rosnav_rl), and a variety of simulators (Gazebo, Isaac Sim).

---

Installation

curl https://raw.githubusercontent.com/voshch/Arena/humble/install.sh > install.sh
bash install.sh

cd ~/arena5_ws # replace with your actual workspace path
source arena
arena update
arena build
arena feature gazebo install # optional
arena feature isaac install # optional
arena feature training install # optional

Usage

cd ~/arena5_ws # replace with your actual workspace path
source arena
arena launch sim:=gazebo                         # default — Gazebo simulator
arena launch sim:=isaac                          # Isaac Sim
arena launch local_planner:=rosnav_rl agent_name:=<your_agent>  # DRL planner
arena launch sim:=gazebo local_planner:=rosnav_rl env_n:=2 train_config:=<path to config.yaml> # DRL training 

DRL quick-start: place your trained agent folder inside Arena/arena_training/agents/<agent_name>/ (must contain training_config.yaml and best_model.zip), then launch with local_planner:=rosnav_rl agent_name:=<agent_name>. Refer to the arena_training README for training instructions.