Reinforcement learning enables quick robot training for rough terrain travel
Towards Reinforcement Learning for Space Robotics: Fast Training and Reliable Real-World Transfer
RoboticsArtificial Intelligence
Summary
Robots that explore planets need to move across uneven ground safely and quickly. The authors show a way to teach a four-legged robot to walk on tricky surfaces using a computer simulation that trains faster than before. After less than 20 minutes of practice on a regular computer, the robot uses what it learned in simulation to walk reliably on real rough surfaces, including places like the moon. This work brings us closer to robots that can handle difficult terrain in space missions.
What this means in practice
- •For space robotics engineers: Train walking robots quickly in simulation to perform well on rough planetary surfaces without expensive hardware setups.
- •For robotics development teams: Develop agile quadruped robots that adapt to diverse real-world environments using fast reinforcement learning training.
Authors
Abhishek Naik, Michael Wu, Michael O'Sullivan, Colin Bellinger, Yunli Wang
Abstract
This work demonstrates the use of a fast simulation-based training framework to learn control policies using reinforcement learning. The hardware requirements are minimal and the simulation-learned policies transfer reliably to the real world. With this setup, we show that a quadruped robot can learn to traverse diverse real-world terrains reliably---including in a lunar-analog facility---after less than 20 minutes of training on a single consumer-grade GPU. These results provide a glimpse of a future where reinforcement learning is a crucial tool in developing more capable and robust space robotic systems. Video demonstrations are available at https://youtu.be/vUFJiZdJaQ0.