Local Mapping System

Overview

The Local Mapping system provided critical real-time environment perception capabilities for the Navigation Stack at Miko.AI, enabling immediate awareness of the robot’s surroundings.

Technical Details

Cyclic Buffer 3D Voxel System

• Implemented cyclic buffer of 3D voxels for efficient real-time mapping
• Maintained three specialized buffer types for comprehensive spatial representation
• Designed system to avoid recalculation at each timestep through cyclic updates
• Created modified octomaps optimized for dynamic local mapping

Voxel Buffer Types

• Free Voxels: Tracked confirmed free space around the robot
• Occupied Voxels: Mapped occupied space and obstacles in the environment
• Distance Transform Voxels: Maintained closest obstacle distances for each 3D space point
• Integrated distance transform data with DWA planning for collision avoidance

Sensor Fusion

• Implemented sensor fusion algorithms for combining pointcloud data with robot pose estimate
• Developed multi-sensor calibration systems
• Designed real-time sensor data processing pipeline

System Integration

• Designed interfaces with local planning system
• Implemented real-time data sharing with global mapping
• Created performance monitoring tools