Multi control of mini-robots

Robotics Platform for Traffic Simulation

  • Design and build a reliable open-source physical robotic platform for multi-agent learning and control research at an acceptable cost.

  • Designed and built (with 3D printer) two kinds of mini-robots based on Arduino-like MCU, and NVIDIA Jetson.

  • Implemented a very robust and precise localization and speed detection algorithm to perceive the environment information for the traffic planning. with a central camera, infrared cameras, and onboard tracking tags.

  • Built the communication network among the controlling server, the central camera, and 20 mini-robots via ROS

  • Transferred learning from simulations to this controlled robotic platform to validate the effectiveness of the Deep Reinforcement Learning based control design of Flow framework

Robotics Control Machine Learning Reinforcement Learning
Xiyu Zhai
Software Engineer

My research interests include robotics, multi-agent control and graphics.