icp

Introduction

This repository hosts CEV's implementation of Iterative Closest Points (ICP) as applied to scan matching.

Authors

• Ethan Uppal
• Jess Wang
• Utku Melemetci

Licensing

Please see LICENSE. Note that all code is licensed under the MIT license, with the exception of some additions to doxygen-style.css. Files with copyright headers denoting specific copyright holders belong to said holders. Files under book/asset are from doxygen_theme_flat_design under its MIT license (again with the exception of some additions to doxygen-style.css). Files without copyright headers are under the MIT license, Copyright (c) 2024-2025 Cornell Electric Vehicles.

Install

You can view installation instructions at INSTALL.md.

Usage

#include "icp/icp.h"
#include "icp/geo.h"
#include "icp/driver.h"
#include <iostream>

void align_clouds(const icp::PointCloud2& a, const icp::PointCloud2& b) {
    std::unique_ptr<icp::ICP2> icp = icp::ICP2::from_method("vanilla", icp::Config()).value();
    icp::ICPDriver driver(std::move(icp));

    driver.set_max_iterations(100);
    driver.set_transform_tolerance(0.1 * M_PI / 180, 0.1);
    auto result = driver.converge(a, b, icp::RBTransform2::Identity());

    std::cout << "Rotation: " << result.transform.rotation() << "\n";
    std::cout << "Translation: " << result.transform.translation() << "\n";
}

Documentation

Some documentation is available in the book folder. There are also Doxygen docs and other generated documentation hosted at cornellev.github.io/icp/, but we haven't update this system in a bit and it might be out of date.

Versions

Version information can be found in the releases page.