1 of 19

Python, NumPy, ROS

Daniel Seita

CSCI 545: Introduction to Robotics (Fall 2025)

September 04, 2025

Based on slides from Varun Bhatt, Bryon Tjanaka, Chris Denniston, Stefanos Nikolaidis.

1

2 of 19

Python

Tutorial if you are new to python:�https://learnxinyminutes.com/docs/python/

​

Note: even though it is meant as a “quick” refresher, the best way to learn how to use a programing language is to use it (often).

2

3 of 19

Basic Python Usage

  • Hello world
  • Variables: int, float
  • Arithmetic operators: exponent; int vs float division
  • String: add, conversion, format, split
  • List: append, indexing, slicing
  • Dict: add, index
  • If else: elif, a < b < c conditions, and/or
  • For: lists, ranges, enumerate, dict keys
  • Functions: return, default values, keyword args
  • Classes: init, attribute, methods
  • Try except: raise

3

4 of 19

Python Pros

  • Great for prototyping - often cited as remarkably similar to pseudocode.
  • Huge user base, so easy to find solutions to simple issues.
  • Can talk to efficient libraries written in C/C++/Fortran (e.g., numpy, pytorch).
  • Many robotics researchers use it (labmates, anyone that uses learning, etc.).

4

Image from https://slides.btjanaka.net/545-python-intro/#/6 (actual source unknown)

5 of 19

Python Cons

  • Interpreted and not compiled, hence slow versus compiled C/C++ code.
    • This is why numpy/pytorch/etc call C/C++ code, though we “interface” using Python.
  • Poor concurrency support (Python GIL).
    • See this note (mid-July) about making GIL optional.
  • Type checking should be done manually.
    • Highly recommend to use linters (available as standalone, editor plugins, or in IDEs).
    • Also make use of assertions.
  • Weird at times…
    • See https://github.com/satwikkansal/wtfpython for some fun examples.
    • Though, arguably this is true for many programing languages…

5

6 of 19

Useful Tools: Python Management

  • Conda and virtual environments (e.g., uv).

​

    • Helps to isolate your Python development per project.
    • Avoid sudo pip install xyz !!
  • Formatting style.
  • Python debugger.
    • import pdb; pdb.set_trace()
  • Python profiling.

6

7 of 19

Useful Tools: Code Editors

My personal usage:

  • VS Code (always update to latest).
  • Extensions: Python, Python Debugger, Remote-SSH, GitHub Copilot, Pylint (for linting), Git Lens (for GitHub).
  • Google Python Style Guide.

Note: everyone’s needs and preferences will vary. Please do what works for you.

7

Good editors with plugins for autocompletion, linting, profiling, debugging, formatting, etc. (image is for VS Code)

Images from https://code.visualstudio.com/assets/docs/languages/python/jupyter.png

8 of 19

NumPy

Official tutorial:�https://numpy.org/devdocs/user/quickstart.html

​

This is one of the most widely used Python libraries.

Also closely related to PyTorch and other Deep Learning libraries.

8

9 of 19

Demo

  • Array creation: array(…), zeros, ones.
  • Array properties: shape (+ reshape), dtype.
  • Indexing (this can get complex).
  • Copying: a = b doesn't copy; explicit copy does.
  • Arithmetic: matrix multiplication (no operator in python 2, use dot), broadcasting, sum, mean (+ per axis).
  • Random: np.random for quick testing.
  • Vectorization.

9

10 of 19

Softmax with Numpy Arrays (Vectorized)

  • You can use np.exp() for ex.
  • x is a vector = [x1, x2, …, xn]
  • Watch out for numerical stability.
    • More near the end of this page.

10

11 of 19

ROS

Official tutorial:�http://wiki.ros.org/ROS/Tutorials

​

Widely used in robotics research code today (and historically).

https://spectrum.ieee.org/the-origin-story-of-ros-the-linux-of-robotics

​

Note: installing can be tricky.

Use Docker, VMware, follow instructions carefully.

Answers.ros.org is deprecated; use robotics.stackexchange.com, etc.

11

12 of 19

Typical Robot

Need some way for these modules to talk to each other

12

Image from https://img.lovepik.com/element/40145/4453.png_1200.png

Sensors (e.g., cameras)

Actuators (e.g., motors)

Algorithms (e.g., planning, localization)

13 of 19

ROS (Robot Operating System)

  • Not an operating system like the name suggests.
  • Provides tools to facilitate robot operation.
  • Designed for a heterogeneous computer cluster.
    • Abstracts low-level details about hardware through common messages.
  • Popular in robotics community and has well designed packages.
    • Need state estimation code for planning research? Just get a state estimation package.
  • Default for labs in this class:
    • ROS Noetic with Ubuntu 22.04/24.04

13

14 of 19

ROS2

14

15 of 19

Nodes, Topics, Publishers, and Subscribers

  • Nodes (green ovals): For running code, performs computation.
    • Formally: an executable that uses ROS to communicate with other nodes.
  • Topics (yellow boxes): Place to post/read messages (like message boards).
  • Publishers: Nodes that write to a topic.
  • Subscribers: Nodes that read messages from a topic.

15

Image from https://slides.btjanaka.net/545-python-intro/#/18 (actual source unknown) Code snippet from Daniel Seita here, just an example, not meant to be run.

16 of 19

Demo

  • roscore → Do this to get started (or roslaunch)
    • Sets up necessary infrastructure for ROS to operate, and for nodes to communicate.
  • Reading and writing messages:
    • rosnode list
    • rostopic list
    • rostopic pub
    • rostopic echo
  • Turtlesim:
    • rosrun turtlesim turtlesim_node
      • (the main node that controls the turtle)
    • rosrun turtlesim turtle_teleop_key
      • (node to read keyboard commands and send velocities)
  • Visualizations:
    • rqt_graph
    • rqt_plot, rviz

16

17 of 19

Lab 01

  • Catkin: Creating and installing of ROS packages
  • More fun with turtlesim
  • Rosbag: For recording data from topics and playing them back (maybe at a faster speed)
  • Rospy: Python wrapper for ROS
    • Python code for publishers and subscribers

17

18 of 19

Miscellaneous

  • ROS Services: alternative to topics for immediate response.
    • One node advertises a service.
    • Other nodes can send a request to those service and get a reply immediately.
    • Usually meant for one-time calls that end quickly, not for continuous data streams.
  • URDF: file format (in “xml”) to describe the geometry of a robot.
  • Robot simulators: for simulating robots in an environment.
    • E.g., Gazebo, Isaac Sim, MuJoCo, etc.

18

Gazebo

Isaac Sim

MuJoCo

19 of 19

Slide Credits

2023 Presentation by Varun Bhatt

2021 Presentation by Bryon Tjanaka

2020 Presentation by Chris Denniston

19