Start from the trot: behaviour cloning and residual policies
Training from scratch found small, fast steps that hobby servos may not follow. Two ways to start from the gait that already walks instead: copy it into the policy network and let PPO improve it, or keep the trot running and learn only a small correction on top. Both trained, measured on random robots and compared.
This lesson is part of the paid course. You can read the first lesson of every course for free.
Order by email€24We reply within a day with the payment details (bank transfer or PayPal) and email you the files right after payment.