Humanoid elevation-frame integration

This repository integrates MID-360 lidar/inertial odometry with humanoid elevation mapping. It credits ANYbotics elevation mapping and FAST-LIO-related work; the contribution described here is engineering integration and additional frame scripts, not a new mapping algorithm or reproduced perception result. [1] [6] [7]

Its README clone command names smoggy-P/elevation_mapping_humanoid, while this entry preserves the inspected ArthurAllshire repository identity. Comparing the stated revisions found 236 identical blob paths, ten changed common paths and three candidate-only files. The added gravity-correction and corrected-frame scripts provide a concrete distinction beyond a simple mirror. That comparison does not establish a preferred canonical maintainer or justify silently substituting repositories. [9]

The documented setup uses ROS/catkin, grid_map, kindr, PCL, Eigen and a Livox driver. The README requires source changes in specified preprocessing/mapping files when choosing driver2 instead of driver1. The inspected laserMapping.cpp already includes livox_ros_driver2/CustomMsg.h. Inspect preprocess.h, preprocess.cpp and laserMapping.cpp at this revision before applying the README replacements; consistency across all three files is not established by the supplied excerpts. Its realsense_demo.launch name should not substitute for inspecting the actual MID-360 topics/configuration. No workspace was built or launch run. [1] [4]

Inspect fast_lio_mid360/script/publish_tf.py for mount transforms. The candidate-specific gravity_correction_publisher.py consumes /livox/imu, averages initial samples and corrects odom_torso toward odom_corrected, using a hard-coded axis convention described as trial-and-error. The publish_corrected_frame.py script publishes height_map_2 using torso_link translation and yaw-only orientation. The README's map-output check is /elevation_mapping/elevation_map; none of those frames or outputs was validated here. These are actual implementation assumptions to inspect against the sensor/torso frame tree, not measured extrinsics or universally correct gravity alignment. No map, TF tree or sensor data was validated. [6] [7]

The inherited directories carry separate BSD-style and GPL notices; no blanket grant for the new integration scripts was established. Public source is not permission over every component. Begin with the exact sensor frames, mounts and driver variant, then inspect the additional frame scripts and map-output topic. No native dependency build, lidar/IMU session, map accuracy, locomotion or real-time performance was tested. [1] [2] [5] [6]