SLAM-MER: Mitsubishi Electric Research framework for visual SLAM

About

This documentation is generated for SLAM-MER, a Visual Simultaneous Localization and Mapping (Visual-SLAM) framework developed at Mitsubishi Electric Research Laboratories (MERL).

Many legacy SLAM codebases are difficult to maintain, debug, and extend for new research directions. With this in mind, we implemented SLAM-MER as a modern, low-latency, modular Visual-SLAM pipeline that explores recent advances in deep learning and optimization while keeping strong geometric foundations. In particular, we integrate vision and depth priors into the framework to improve the robustness, accuracy, and efficiency of the SLAM system. The pipeline is implemented in C++ and is designed to run in indoor and outdoor environments. The current code uses Eigen, CudaSift, GTSAM, PoseLib, FAISS, Rerun, libpng, libjpeg-turbo, optional LibTorch/ALIKED support, and optional ONNX Runtime-based modules.

The project is under active development, and we continue to add features and improve the documentation to make integration and extension easier.

Repositories

The source tree is split into the main framework and optional module repositories:

  • Main framework: slam-mer. This repository contains the core C++ Visual-SLAM pipeline, integration code, build configuration, and main documentation.

  • Depth module: thirdparty/depth_inference. This repository provides the standalone depth estimation module (ONNX Runtime) used as an optional depth prior in SLAM-MER.

  • Visual place localization module: thirdparty/visual_place_localization. This repository provides the standalone visual place recognition/localization module used by SLAM-MER for retrieval and relocalization workflows.

  • Examples and experiments repository: slam-mer_examples. This repository contains runnable examples and experiment-style scripts that demonstrate how to use and evaluate the framework.

To install and use the framework, see Installation and Examples. For ROS integration notes, see ROS2 Wrapper.

MERL logo

This project is supported by the Mitsubishi Electric Research Laboratories (MERL), 201 Broadway, Cambridge, Massachusetts.