Page 113 - Read Online

Zhang et al. Intell Robot 2022;2(3):275­97 I http://dx.doi.org/10.20517/ir.2022.20 Page 297

https://github.com/Alescontrela/AMP_for_hardware

https://github.com/Mehooz/vision4leg https://github.com/PaddlePaddle/PaddleRobotics/tree/main/QuadrupedalRobots/ETGRL

/ / /
Per- Terrin, and Ini- Force, Joint loss, Friction, Friction, Sensor
Velocity (Rough Training, Mass, External Mass, Friction Adaptation,and Foot Noisy

Mass, (Robot of Centre and Lateral and

Base Randomisation Curriculum Termination. Randomisation Orientation, Mass, Damping Limit). Inertia, Domain Latency, Kd), Kp,

(Friction, Multiplier). Force), Trajectory Velocity, FrictionCoefficient, CurriculumLearning, andRandomized (Body Joint Torque and KD, (KP, Mass,MotorFrictionandStrength,andSensorLatency),and Setting, (Control Motor and

Randomization Gain Motor Domain Model, External Domain Model, and Position Sampling. Randomization Reference, and Gain, P Friction, Randomization Input. Depth Learning Randomization Mass, Leg

Domain turbation, Actuator Disturbances, Joint-Velocity-Based Actuator Joint tial Height Domain Position Geom Domain Random Teacher-Student Domain Mass, Base Input.

Motion RL another controller. walking Genera-

Shepherd from Data MPC an Generator. Trajectory dog of Dataset behaviors [111] . Trajectory

German Dataset [119] . Motion or policy Foot MoCap turning and / Evolutionary tor [120] .

Make Multi- in Priors loco- robots Learn- Movement Animal Locomotion with Gener- for

Priors through Motion Learning [56] perceptive and Vision-Guided Cross-Modal Learning Trajectory Approach Locomotion [62]

Motion GoodSubsitutesforComplexRe- Functions [91] Skills robust quadrupedal Repurpose: Robot Human with General

Adversarial ward Advanced Adversarial ple Reinforcement Learning for motion wild [8] the in and Imitate Reusable ing From Skills Behaviors [12] Learning Quadrupedal End-to-End Transformers [13] Reinforcement Evolutionary A ator: Quadrupedal