Nvblox is a library designed for real-time 3D reconstruction and mapping, specifically targeted at robotics applications. It builds a voxel-based map from depth images and 3D LiDAR scans, outputting a mesh of the world in real-time and a cost map for path planning. The library is optimized for discrete NVIDIA GPUs and Jetson devices.

Features

• Voxel-Based Mapping: Constructs a 3D voxel grid using depth images and 3D LiDAR scans.
• Real-Time Mesh Output: Provides real-time mesh generation of the environment.
• Path Planning: Generates a cost map for use in path planning.
• TSDF and ESDF: Utilizes Truncated Signed Distance Function (TSDF) and Euclidean Signed Distance Function (ESDF) for high-quality surface reconstructions.
• Robotics Applications: Suitable for navigating Autonomous Mobile Robots (AMRs) and motion generation for robotic arms.
• Optimized for NVIDIA Hardware: Specifically optimized for discrete NVIDIA GPUs and Jetson devices.

Benefits

• High-Quality Reconstructions: Provides higher quality surface reconstructions compared to traditional methods.
• Collision Detection: Facilitates collision detection for future robot positions.
• Efficient Path Planning: Enhances path planning with accurate distance fields.
• Real-Time Performance: Ensures real-time performance for dynamic environments.