Locomotion

Use PlotJuggler to inspect SRBD, MPC, feet references, and wbid_solve_time in the G1 MuJoCo MPC/WBID stack.

Compare posture PD and contact-aware WBID for G1 MuJoCo: gains, q_init, qddot, contact forces, tasks, and constraints.

Design a low-cost egocentric video protocol for WholeBodyVLA-style locomotion LAM pretraining.

Compare GRAIL manipulation reconstruction with static-object mode for stairs, slopes, curbs, and sitting.

Transferring locomotion policy from simulation to real robot — domain randomization, actuator networks, and best practices.

Challenges of bipedal robot control — balance, fall recovery, terrain adaptation from Cassie to Digit and humanoid.

Analyzing Extreme Parkour and SoloParkour — teaching robots to jump, climb stairs and overcome complex terrain with RL.

Analyzing Walk-These-Ways paper — train one single policy for multiple movement types with multiplicity of behavior.

Hands-on train locomotion policy with legged_gym and deploy to 4-legged Unitree Go2 robot — from sim to real.

How RL is applied to locomotion — from reward design, observation space to training pipeline with Isaac Lab.

History and theory of robot movement control — ZMP, CPG, inverse kinematics before RL dominated.