Introduction to curobo_ros

Welcome! This guide will introduce you to curobo_ros and the key technologies it uses. If you’re new to robotics software, ROS, or Docker, this is the perfect place to start.

What is curobo_ros?

curobo_ros is a ROS 2 package that brings GPU-accelerated motion planning to robotic arms. It allows you to:

  • Plan collision-free trajectories in milliseconds (not seconds!)

  • Solve inverse kinematics (IK) to find joint angles for desired end-effector positions

  • Avoid obstacles in real-time using depth cameras or manual object definitions

  • Visualize everything live in RViz

  • Control real robots or test in simulation

Think of it as a “smart motion planner” that uses your GPU to compute safe paths very quickly.

Key Technologies

1. What is ROS 2?

ROS (Robot Operating System) is not actually an operating system - it’s a framework for building robot software.

Why ROS?

  • Modular: Different parts of your robot (sensors, motors, planners) run as separate programs called nodes

  • Communication: Nodes talk to each other through topics (streaming data) and services (request/response)

  • Reusable: Use existing packages instead of writing everything from scratch

  • Visualization: Tools like RViz let you see what your robot is doing

Basic ROS Concepts

Concept

Description

Example

Node

An executable program that does one job

unified_planner (plans trajectories)

Topic

A named data stream (publish/subscribe)

/joint_states (robot joint positions)

Service

Request/response communication

/generate_trajectory (ask for a path)

Action

Long-running task with feedback

/unified_planner/execute_trajectory (execute a trajectory)

Message

Data structure sent between nodes

geometry_msgs/Pose (position + orientation)

Common ROS Commands

# List all running nodes
ros2 node list

# List all available topics
ros2 topic list

# See messages on a topic
ros2 topic echo /joint_states

# Call a service
ros2 service call /unified_planner/generate_trajectory curobo_msgs/srv/TrajectoryGeneration "{...}"

# Check node parameters
ros2 param list

2. What is Docker?

Docker is a tool that packages software with all its dependencies into a container - a lightweight, isolated environment.

Why Docker for this project?

Installing curobo_ros requires:

  • ROS 2 Humble

  • CUDA 12 (NVIDIA GPU libraries)

  • PyTorch with CUDA support

  • cuRobo library

  • Many Python dependencies

Setting this up manually can take hours and often breaks. Docker gives you a pre-built environment where everything already works!

Docker Key Concepts

Concept

Description

Analogy

Image

A template with all software pre-installed

Recipe for a cake

Container

A running instance of an image

The actual cake you baked

Volume

Share files between host and container

A shared folder

Build

Create an image from a Dockerfile

Following the recipe

For setup details, DEV vs PROD workflows, and daily usage, see the Installation Guide.

3. What is cuRobo?

cuRobo (CUDA Robot) is NVIDIA’s GPU-accelerated motion planning library.

Why GPU acceleration?

Traditional motion planning can take seconds. cuRobo uses your NVIDIA GPU to:

  • Solve inverse kinematics in milliseconds

  • Check thousands of collision scenarios in parallel

  • Optimize trajectories 100x faster than CPU-only methods

What cuRobo Does

  • Inverse Kinematics (IK): Given a target position/orientation, find joint angles

    • Example: “Move the gripper to (x=0.5, y=0.2, z=0.3)” → cuRobo calculates joint angles

  • Motion Generation: Plan a smooth, collision-free trajectory from A to B

    • Considers joint limits, velocity limits, acceleration limits

    • Avoids obstacles

    • Optimizes for smoothness and efficiency

  • Collision Checking: Uses voxel grids and sphere approximations

    • Robot represented as spheres

    • World represented as voxels

    • Ultra-fast GPU collision detection

Prerequisites

Before starting, make sure you have:

  • Ubuntu 20.04 or 22.04 (or Windows 11 with WSL2)

  • NVIDIA GPU with drivers installed (check with nvidia-smi)

  • ~30 GB free disk space (Docker image is large!)

  • Docker and NVIDIA Container Toolkit installed

  • Basic command line knowledge (cd, ls, etc.)

Next Steps

Ready to get started?

  1. Installation Guide - Set up Docker, build the image, and run your first trajectory

  2. Your First Trajectory - Generate your first motion plan

  3. Parameters Guide - Understand important configuration parameters

Getting Help

Let’s begin! → Installation Guide