Train a balance bot with reinforcement learning

T1Covered by 1 source · first reported by Raspberry Pi / Arduino / Adafruit at 28 Oct, 16:13 UTC

Shawn Hymel wrote a guide in Make: Volume 99 on training a two-wheeled balance bot with reinforcement learning instead of a classical control loop. The agent is trained with proximal policy optimization, or PPO, the algorithm OpenAI published in 2017 that is still used for robot control and for…

ai

Sources

• T1Train a balance bot with reinforcement learningRaspberry Pi / Arduino / Adafruit28 Oct, 16:13