ICRA 2023 Tutorials on Non-Gaussian SLAM and Mapping

Welcome to the ICRA 2023 tutorials on Non-Gaussian SLAM and Mapping!

This hands-on tutorial series builds on the success and feedback from the ICRA 2022 workshop, where we discussed some of the canonical issues motivating Non-Gaussian SLAM.

In this tutorial series, we will take the problems further by:

    1. Defining ubiquitous mapping problems like contradictory information, partial map updates, and prior data fusion
    2. How formulating the problem as a factor-graph and solving with multiple solvers (non-Gaussian and Gaussian) addresses these issues
    3. Demonstrating these issues in relevant examples such as construction automation, warehouse automation,  and autonomous driving

Furthermore, the DIY examples in this tutorial will demonstrate how the same modeling philosophy is readily extensible to other non-Gaussian behavior, and how this new modeling freedom can simplify SLAM front-end processes while building enterprise-level maps for autonomous vehicles.

Learn more about the four drivers of non-Gaussian behavior in SLAM.

The tutorials are packaged for a zero-install, “bring-your-own-laptop” setup so that you can easily experiment with non-Gaussian SLAM code, and the results are published if you would prefer to simply peruse through the examples.

Want to redo our ICRA 2022 tutorials? Get started here!


Note: The tutorials will be provided both in-person as well as remotely via Gather.Town to allow for both physical and remote participation. Please feel free to join us from anywhere in the world via Gather.Town!

What you will learn in this workshop

Mapping and localization systems are a critical component in advanced autonomous systems, both for human decision support systems and for enabling humans and robot cooperation. Building maps from various data is perhaps easy to describe, but the technical difficulties of doing so robustly and at scale are enormous.

This tutorial is focused on:

    1. Driving debate on the state-of-the-art in non-Gaussian SLAM
    2. Demonstrating through do-it-yourself examples how to use non-Gaussian factor graphs to solve mapping and localization problems beyond parametric or multi-parametric methods
    3. Providing examples of ubiquitous problems in mapping like contradictory information, partial updates, and prior information fusion
    4. Building awareness around open-source multi-modal SLAM solvers such as Caesar.jl

Four interactive examples have been prepared and will be presented in 30-minute segments throughout the day using an open-discussion format.

The tutorials consist of both 2D and 3D real-world data sets, as well as illustrative examples to help attendees new to SLAM get started in factor graph based navigation. Each example demonstrates challenging problems in SLAM, such as using prior data when building maps and resolving contradictions during map updates.

Who Should Attend


Researchers in the fields of SLAM and robotics looking to go beyond the unimodal assumption

Industry Roboticists

Industry leaders looking to add robustness, trust, and safety to their applications

New Robotics Engineers

New roboticists learning about SLAM algorithms, data persistence, and SLAM in the cloud


The following DIY canonical examples (incorporating real-world use-cases) are provided to guide the reader on the four main drivers of non-Gaussian behavior in SLAM. 

You can work through any, or all, of the following examples during the day:

    1. Tutorial 1 – Creating and Solving Factor Graphs: An introduction to building and solving factor graphs, incorporating various sensor data, and producing a cohesive map
    2. Tutorial 2 – Mapping and Navigating Boston Harbor: A real-world example where we build a map of the Boston Harbor without GPS, solving ambiguity using a multimodal solver
    3. Tutorial 3 – The Power of Contradictory, Dynamic, and Prior Data: A more theoretical example where we incorporate CAD (and other) prior, potentially contradictory data
    4. Tutorial 4 – Building Dynamic 3D Maps using Multi-Modal SLAM: A full tutorial demonstrating how to build, update, and manage large maps for multiple robots in the same space


The tutorials are designed to be:

    • Zero-footprint setups that should take about 15 minutes to take you from problem definition to results and analysis
    • Run in a browser in a JupyterHub Notebook independently or pull the code to your local machine to review later
    • Partially guided in open an open forum setting – all tutorials will always be available and can be run at any time, however we will guide attendees through the problem, the results, and the relevance of the outcomes during each session
    • Interactive with time and resources set aside for discussions, so please feel free raise any questions at any point during the tutorials

More Information

For more information please feel free to reach out to us at any point for general information at info@navability.io.

For specific questions regarding the tutorials please feel free to reach out to us directly:

Dehann Fourie


Sam Claassens


Jim Hill

Technical Operations

Johan Terblanche

Remote Operations

Our Sponsors