cuNLS Documentation

cuNLS provides GPU-accelerated nonlinear least-squares optimization for batched geometric estimation problems.

User Guide

• Introduction
  • Purpose
  • Nonlinear least-squares problems
  • Factor Graphs
  • Core concepts
  • High-level solve flow
  • Supported optimization patterns
• Installation
  • Prerequisites
  • Build locally
  • Build with Docker
  • Direct CMake build (manual path)
  • Notes
• Tutorial
  • Overview
  • Common build commands
  • Sparse Bundle Adjustment
  • Pose Graph Optimization
  • Custom Factor
• Quick Start
  • Step 1: Install cuNLS
  • Step 2: Create a minimal source file
  • Step 3: Create CMakeLists.txt
  • Step 4: Build and run
• Build and Run Tests
  • Step-by-step
  • Alternative: Run the aggregated binary directly
  • Coverage mode
• Licensing
  • Runtime / core build dependencies
  • Optional dependency
  • Python bindings (pycunls) dependencies
  • Test-only dependency
  • License texts in repository

Python Guide (pycunls)

• pycunls Installation
  • Prerequisites
  • Install from source (pip)
  • Build a wheel
  • Dependencies
  • Verify the installation
• Python Quick Start
  • Step 1: Install pycunls
  • Step 2: Create a minimal script
  • Step 3: Run
• Python Tutorial
  • Overview
  • Sparse Bundle Adjustment
  • Pose Graph Optimization
  • Custom Warp Factor
  • Custom Warp State
• Python Tests
  • Prerequisites
  • Run all tests
  • Run a specific test module
  • Test modules

API Reference

• API Reference
  • Minimizer API
  • State API
  • Factor API
  • Robustifier API
  • Linear Solver API
  • Common API
  • Math API