ICRA-RoboMaster AI Challenge Simulator

Background

This project is aimed to build a simulator for ICRA RoboMaster AI Challenge. Through the simulator, a lot of Deep Reinforcement Learning algorithms can be tested and train a good policy for the competition. It may also require transfer learning to overcome the gap between sim to real.

Features

• Multi-agent navigation based on multi-jackle ROS package.
• Omni-directional chassis combine with RoboRTS.
• Rotatable turret with a camera on the robot model.
• Connect gazebo with OpenAi gym framework by openai_ros ROS package.
• Speed up the gazebo simulation.
• Referee system based on the competition’s rule.
• Custom Debuff layer was added to avoid the debuff zone in the arena.

Project Video

Project Code

https://privilger.github.io/RoboMaster-Simulator