Atlas rough-terrain paper demonstrates traversability-aware humanoid path planning

Key points

• The paper presents a humanoid navigation planner that uses traversability assessment over 2.5D height maps for rough terrain.
• The framework combines graph search, trajectory optimization, footstep planning, and balance control, with demonstration on a DRC Boston Dynamics Atlas robot.
• The work gives Atlas a benchmark for humanoid path planning over rough terrain beyond flat-floor locomotion demos.