Simulation

Simulation#

The simulated side of the experiments: the Habitat-sim renderer, the stand-in controllers that execute trajectories without physics, and the HSSD scenes. Node parameters and topics are in the open_apero_habitat and open_apero_controllers READMEs; Installation sets up Habitat-sim and the dataset.

The Habitat Renderer#

habitat_renderer_node runs in Habitat-sim’s own Python environment. For every camera pose the controller publishes on /camera_transform it renders, in arrival order, a depth image (32FC1, planar z, in meters), a color image and the camera info, stamped with the pose’s stamp. Once up, it publishes /open_apero/simulation_ready, which mission_coordinator_node waits for before it starts the planner.

  • Depth, in order: NaN and infinite readings become 0, readings above max_depth become far_depth_sentinel (so voxblox clears free space up to the observation boundary), readings below min_depth become 0.

  • Intrinsics: fx = fy = width / (2 tan(hfov / 2)), cx = (width - 1) / 2, cy = (height - 1) / 2.

The Stand-In Controllers#

dummy_controller_node publishes the scene’s initial pose until the first trajectory arrives, then steps through its waypoints at time_from_start / speedup_factor, publishing each as TF, robot_state and, with publish_camera_transform: all, /camera_transform, which makes Habitat render a frame. After the last waypoint it publishes command/result with the trajectory’s command_sequence and final state. keyboard_teleop_node drives by hand: each key moves, turns, raises or lowers the pose and sends it to dummy_controller_node as a one-waypoint trajectory.

Scenes and the HSSD Dataset#

The scenes come from the HSSD dataset (HSSD-hab on Hugging Face, CC BY-NC 4.0, non-commercial use only), found through HSSD_DATA_ROOT (set in datasets.env). open_apero_ros/config/scenes/HSSD/ holds one file per scene, the paper’s eight and a ninth, 103997562, each setting the stage and the start position; scene_id (default 102816036) selects one. The README’s scene files section lists them, and Reproducing the Paper gives the paper’s budgets.