Arena is a platform for developing and benchmarking navigation algorithms in human-centric social environments. We offer a wide variety of different social force models, robots, planners, and world generation algorithms, and many more to use. All functions are abstracted and can be run across three widely used simulators: Flatland 2D, Gazebo, Unity 3D, and Isaac Sim (added 2025). Arena Rosnav also offers a complete evaluation pipeline for benchmarking the performance of robots and planners based on standard metrics, and a trainings pipeline for navigational models based on DRL and PPO. With this pipeline our own DRL planner ROSNavRL was created.

Follow the documentation for details how to use the platform:

Documentation

• Automatic installation script for the arena rosnav environment

• 4 different simulators including Isaac, Unity, Gazebo and Flatland

• We offer prebuilt, realistic simulation environments, including offices, hospitals, canteens, warehouses, and much more

• Dynamic Map Generation including dynamic mazes

• Variety of Task Modes for robots and pedestrians

  Task Mode	Short Description	Robots	Obstacles
  scenario	load scenario file	✓	✓
  random	generate random positions	✓	✓
  parametrized	more fine-tuned random		✓
  guided	waypoint sequence	✓
  explore	explore map	✓

• Variety of Robots including the go1 quadruped robot

• Variety of Planners including our own DRL planner ROSNavRL

• Variety of social force models for pedestrians

• Pipeline for evaluating approaches and analysing them based on standard metrics with our Arena Evaluation package.

• Pipeline to train planner agents based on reinforcement learning approaches from stable baselines3

• Modular and flexible structure for extension of new functionalities and approaches

• Fully integrated Move Base Flex in our Arena-Rosnav ecosystem

• ROSNavRL: Our own planner based on neural networks.
• Dragon: from the BARN challenge
• Trail: from the BARN challenge, TRAIL lab
• Applr: a hybrid approach by Xuesu et al.
• RLCA-ROS: a DRL-based colision avoidance approach from Long et al.
• CADRL: a DRL-based colision avoidance approach from Everett et al.
• SARL-Star
• Crowdnav-ROS: a DRL-based colision avoidance approach from Chen et al.
• TEB: a classic approach by Rösmann et al.
• DWA: the standard ROS local planning approach by Marder-Eppstein et al.
• MPC: a classic approach by Rösmann et al.
• and many more (added with Arena 3.0 (2024), Arena 5.0 (2025))

turtlebot3-burger	jackal	ridgeback	agv-ota	tiago

Robotino(rto)	youbot	turtlebot3_waffle_pi	Car-O-Bot4 (cob4)	dingo

Hospital	Canteen	Campus	Factory	Warehouse

Hospital	Restaurant	School	Japanese Garden	Warehouse

• Arena 4.0 (ICRA2025): Improved World Generation and Realsim
• Arena 3.0 (RSS2024): Social Navigation Platform
• Arena-Web (RSS2023): Web-based Development and Benchmarking Platform for Autonomous Navigation Approaches
• Arena-Rosnav 2.0 (IROS2023): A Development and Benchmarking Platform for Robot Navigation in Highly Dynamic Environments
• Arena-Bench (RA-L+ IROS22): A Benchmarking Suite for Obstacle Avoidance Approaches in Highly Dynamic Environments
• Arena-Rosnav (IROS21): Towards Deployment of Deep-Reinforcement-Learning-Based Obstacle Avoidance into Conventional Autonomous Navigation Systems
• All-in-One (ICRA22): A DRL-based Control Switch Combining State-of-the-art Navigation Planners