Spatio-Temporal Voxel Layer

What it does

• Stores the environment as a configurable voxel grid in OpenVDB, sized by voxel_size, with optional approximate voxel or passthrough filtering of sensor data.
• Expires voxels over time with linear, exponential or persistent decay models, and accelerates decay for voxels inside current sensor frustums instead of clearing them by raytracing.
• Models depth cameras with cubical frustums and 3D lidars with hourglass-shaped FOVs, using parameters such as hFOV and vertical FOV offset.
• Supports a mapping mode that keeps the full voxel grid and saves maps through services, with maps viewable in vdb_viewer and convertible to point clouds.
• Sources can be toggled at runtime through a ROS service, and configuration is done through costmap YAML files with a plugin entry in the local and global costmap params.
• Ships with a jemalloc LD_PRELOAD workaround for a known OpenVDB issue on Ubuntu 20.04 binaries.

Written by Robotheca from the project's README (opens in a new tab) at commit 16cdf3b.

Getting started

Difficulty

Intermediate

Hardware it needs, as the project states it

It names no required minimum hardware, and it requires ROS Kinetic (or Foxy/Noetic per the note) with OpenVDB and TBB.

How to start

The project's README (opens in a new tab)

Maintenance

Status

Active: a commit in the last 90 days

Newest commit

17 Sept 2026 on ros2

Latest release

2.7.0 (5 Aug 2026)

Checked on GitHub on 16 Oct 2026.

Links

• Repository on GitHub: github.com/SteveMacenski/spatio_temporal_voxel_layer (opens in a new tab)