[RL] Quadruped Blind Trot Walking with Reinforcement Learning

Описание к видео [RL] Quadruped Blind Trot Walking with Reinforcement Learning

Zero-shot sim2real Experiment
As input to the neural network, Encoders and IMU information are used.
(External observation sensors such as cameras, lidar, and depth cameras, etc. were not used.
all computations are performed on the robot's internal CPU (without GPU)

00:00 Intro
00:03 Balancing Test
00:30 Walking Test
01:25 Walking & Disturbance Test
03:00 Walking on Small Rock Terrain
03:20 Walking & Disturbance Test
04:00 Walking on Rigid Sand Terrain & Disturbance Test
05:50 Step Down Test
06:40 Walking & Disturbance Test

Комментарии

Информация по комментариям в разработке