How ashflamer/lidar-robot-ml-3d-suite works
This repository combines a browser-based autonomous rover simulation and sensor comparison with a real-LiDAR data collection and model-training workflow. The main user path is starting the warehouse mission, where sensing and learned motion cues inform planning, path smoothing, and robot control; a separate workflow collects, cleans, trains, and exports real-data models for the browser. The README documents several simulation internals whose implementation is in index.html, but that source was not sampled, so those links are documented rather than source-verified.
Main entry points: train_real_lidar_ml.py Model Training · index.html LiDAR Perception · lidar_dataset.csv Clean LiDAR Dataset · export_browser_models.py Browser Model Export
Components 19
Rover Simulation
- Warehouse Mission
index.html - LiDAR Perception
index.html - ML Inference
index.html - Safety Bumper
index.html - Space-Time Planner
index.html - Path Smoothing
index.html - Pure Pursuit
index.html
Data And Models
- Track Cleaning
clean_lidar_dataset.py - Clean LiDAR Dataset
lidar_dataset.csv - Model Training
train_real_lidar_ml.py - Trained Artifacts
ml_artifacts - Browser Model Export
export_browser_models.py - Browser Model Pack
ml_models.js
Comparison And Evaluation
- Sensor Comparison
comparison.html - Navigation Benchmark
benchmark_navigators.py
LiDAR Collection
- ROS LiDAR Collector
collect_live_lidar.py - Raw LiDAR Dataset
lidar_dataset_raw.csv
Other components
- User
- YDLidar ROS Driver
Connections 22
- User → Warehouse Mission: starts mission
README.md - Warehouse Mission → LiDAR Perception: senses scene
README.md - LiDAR Perception → ML Inference: supplies tracks
README.md - ML Inference → Space-Time Planner: informs costs
README.md - LiDAR Perception → Safety Bumper: monitors corridor
README.md - Space-Time Planner → Path Smoothing: plans route
README.md - Path Smoothing → Pure Pursuit: provides path
README.md - Pure Pursuit → Warehouse Mission: drives rover
README.md - User → Sensor Comparison: opens comparison
README.md - User → Navigation Benchmark: runs benchmark
README.md - YDLidar ROS Driver → ROS LiDAR Collector: publishes scans
collect_live_lidar.py - ROS LiDAR Collector → Clean LiDAR Dataset: logs measurements
README.md - Raw LiDAR Dataset → Track Cleaning: provides records
README.md - Track Cleaning → Clean LiDAR Dataset: writes cleaned tracks
clean_lidar_dataset.py - Raw LiDAR Dataset → Model Training: provides training data
README.md - Clean LiDAR Dataset → Model Training: provides sequences
README.md - Model Training → Trained Artifacts: saves models
README.md - Model Training → Browser Model Export: invokes export
train_real_lidar_ml.py - Trained Artifacts → Browser Model Export: loads weights
export_browser_models.py - Browser Model Export → Browser Model Pack: writes model pack
export_browser_models.py - Browser Model Pack → ML Inference: supplies models
README.md - Navigation Benchmark → Sensor Comparison: documents results
README.md