LP-NavOA: Integrated Local Navigation and Obstacle Avoidance for Humanoid Robots under Limited Perception

In each trial, the coordinate frame is recalibrated using the robot's initial pose before execution. The target is then specified at a fixed distance directly in front of the robot in this local frame, and the run is terminated once the robot reaches the target x-position rather than requiring convergence to the target y-position. This protocol is intended to evaluate whether the policy can produce timely forward progress while accurately avoiding obstacles and recovering toward the task goal after bypassing. Indoor experiment videos are not uploaded at this stage because laboratory management rules require some internal content to be visually masked, while large-scale masking of the laboratory environment is impractical; the corresponding indoor results are instead fully illustrated by the figures in the paper.