Skip to content

agents.md

Latest commit

 

History

History
112 lines (80 loc) · 3.4 KB

agents.md

File metadata and controls

112 lines (80 loc) · 3.4 KB

MOLA Academic Datasets — AI Agent Context Guide

Project Overview

This repository contains MOLA RawDataSource modules for well-known academic robotics datasets (KITTI, KITTI-360, EuRoC, MulRan, Paris-Luco) plus KITTI benchmark evaluation tools.

These packages were split from the main MOLAorg/mola repository to keep dataset-specific I/O code separate from the MOLA core.

Maintainer: Jose Luis Blanco-Claraco
License: GPL-3.0
Docs: https://docs.mola-slam.org/latest/

Architecture

Each package provides a RawDataSourceBase plugin that:

  1. Reads a dataset from disk in its native format
  2. Publishes mrpt::obs::* observations into the MOLA pipeline
  3. Supports offline playback (implements OfflineDatasetSource)

All dataset readers are configured via YAML (variable substitution via mola_yaml).

Directory Structure

mola_input_euroc_dataset/      # EuRoC MAV stereo+IMU dataset reader
mola_input_kitti_dataset/      # KITTI odometry/SLAM dataset reader
mola_input_kitti360_dataset/   # KITTI-360 panoramic dataset reader
mola_input_mulran_dataset/     # MulRan urban LiDAR dataset reader
mola_input_paris_luco_dataset/ # Paris-Luco outdoor LiDAR dataset reader
kitti_metrics_eval/            # KITTI benchmark evaluation CLI tools

Key Packages

mola_input_kitti_dataset

  • Reads KITTI odometry sequences (velodyne, stereo camera, ground-truth poses)
  • Key class: KittiOdometryDataset in include/mola_input_kitti_dataset/
  • Supports correction angle for LiDAR calibration

mola_input_kitti360_dataset

  • Reads KITTI-360 sequences (panoramic LiDAR, cameras, IMU, GPS)
  • Key class: Kitti360Dataset

mola_input_euroc_dataset

  • Reads EuRoC MAV sequences (stereo camera, IMU)
  • Key class: EurocDataset

mola_input_mulran_dataset

  • Reads MulRan sequences (Ouster LiDAR, GPS/IMU)
  • Key class: MulranDataset

mola_input_paris_luco_dataset

  • Reads Paris-Luco dataset (single outdoor LiDAR sequence)
  • Key class: ParisLucoDataset

kitti_metrics_eval

  • CLI tool for computing KITTI odometry metrics (translation/rotation errors)
  • Input: estimated trajectory + ground-truth poses

Build System

  • Build tool: CMake 3.22+ / Colcon (for ROS 2)
  • C++ standard: C++17
  • Config: colcon_defaults.yaml (symlink install, RelWithDebInfo)
  • Platforms: Linux Ubuntu 22.04/24.04, AMD64 and ARM64

Key External Dependencies

Dependency Purpose
mola_kernel RawDataSourceBase, OfflineDatasetSource interfaces
MRPT (≥2.1.0) Observation types, poses, serialization
mola_yaml YAML config parsing with variable expansion

Testing & CI

  • CI: .github/workflows/build-ros.yml — builds on ROS 2 Humble, Jazzy, Rolling
  • Style: .clang-format / .clang-tidy enforced in CI

Common Patterns

Adding a New Dataset Reader

  1. Inherit from mola_kernel::RawDataSourceBase (and OfflineDatasetSource if offline)
  2. Override initialize(), spinOnce(), datasetEntry(), datasetSize()
  3. Register the plugin via CMake macros from mola_common
  4. Write a YAML config referencing the dataset path and parameters

YAML Configuration

Dataset path and sequence are set via YAML variables, e.g.:

- type: mola_input_kitti_dataset::KittiOdometryDataset
  params:
    base_dir: ${KITTI_BASE_DIR}
    sequence: '00'