In Shandong Province Key Research Plan, to address the safety concerns associated with manned operations on the launch and recovery platforms, our team have initiated the development of an intelligent assistant robot capable of conducting patrol inspections in the satellite launch and recovery platform environment. This robot system is designed to adapt to the specific requirements of satellite launch and recovery platforms and will be integrated into the development of maritime satellite launch command and control systems.

Summary

The maritime rocket launch system consists of a rocket launch command ship and a rocket launch support ship. During rocket launch, there are safety issues with manned operations on the launch platform and recovery platform of the launch support ship, requiring assistance from intelligent robots to conduct patrol inspections. Two local area networks are set up to remotely control the patrol of quadruped robots on the rocket launch support ship from the rocket launch command ship. The video captured by the cameras mounted on the quadruped robots is transmitted back to the rocket launch command ship, achieving real-time monitoring of the rocket launch support ship’s condition. We fine-tuned the classic path planning algorithm $D^{*}$ to improve the obstacle avoidance success rate of quadruped robots patrolling on typical rocket launch support ship decks by adjusting the forward direction of path planning. Finally, we constructed simulation training platforms for quadruped robots based on two simulation physics engines, mujoco and pybullet, respectively. We further use Reinforcement Learning algorithm to train a control policy for quadruped robots based on simulation platforms.

Keywords and Definitions

$D^{*}$ Algorithm, Obstacle Avoidance, Reinforcement Learning

Final Draft are Here.