EvergreenSLAM: always-on 2D lidar SLAM as the long-term spatial memory of an agent
Year: 2026
Type: software
Abstract: Always-on, hands-off, agent-friendly 2D lidar SLAM. Maps are created, frozen and extended automatically across power cycles, with pose-graph trimming by marginalization and Chow-Liu sparsification to keep the graph bounded. Named places (anchors) bind to graph nodes rather than coordinates, and an agent reads and writes its spatial memory as plain files through a CLI and an MCP server. Source: github.com/ThomasZB/EvergreenSLAM
Author: Chen, Hang
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7225451129
