Jeremy Leibs (He/Him), Chief Architect of Rerun, Sweden
Abstract: The Rerun visualizer and data platform need to work with multimodal data across a wide variety of formats. On one hand, it’s crucial to allow users to store their data in formats that mirror their native representations. On the other hand, generic processing tools like visualizers need that data converted to a more canonical structure. Multimodal data presents additional challenges as these conversions go beyond simple format-interpretation and require a greater degree of semantic context. Common examples include operations like temporal alignment of signals, coordinate-frame reprojections, or decoding of embeddings. This talk will highlight how several features of the rust language allow us to do these kinds of transformations with the level of safety and performance required by our users.
Dr. Angelo Corsaro (He/Him), CEO/CTO of ZettaScale Technology Inc., France
Abstract: Communication is at the heart of robotics —yet traditional middleware often limits what’s possible in terms of performance, scalability, and safety. That’s changing fast. Zenoh, a zero-overhead, pub/sub/query protocol designed for distributed and real-time systems, has been officially selected as an alternative RMW (ROS Middleware) for ROS 2, and is set to become Tier-1 with the upcoming Kilted Kayu release.
What makes Zenoh truly stand out is its native Rust implementation. By leveraging Rust’s memory safety guarantees, fearless concurrency, and zero-cost abstractions, Zenoh delivers ultra-low latency, predictable resource usage, and strong security—all critical for safety-critical and autonomous robotic applications. In this talk, we’ll introduce Zenoh and explore the rationale behind Zenoh’s integration into ROS 2.
Guillaume Binet (He/Him), Founder of Copper Robotics, United States
Abstract: As robotics systems become increasingly complex, integrating teams of algorithm developers, machine learning engineers, control and other specialists can present unique challenges—particularly when relying on traditional languages like C++. In this talk, I’ll take you through my journey of building Copper, a robot runtime designed to tackle these integration challenges.
While these specialists bring deep expertise in their respective fields, they often encounter difficulties in addressing the concurrency issues, undefined behavior, and subtle pitfalls that can arise in a C++-based codebase. Rust offers a solution by providing both systems-level performance and safety, while also being accessible to developers who may not have extensive experience managing these low-level concerns.
We’ll dive into specific examples of how Rust shines in both systems programming and algorithmic development, from safely managing concurrency to writing expressive, error-free algorithms. I’ll also share practical Rust code snippets, demystifying the language and showing how it can serve as a powerful tool in the robotics ecosystem, especially under the Copper framework.
By the end, you’ll see how Rust bridges the gap between low-level control and high-level algorithmic design, making robotics development more integrated, efficient, and less prone to costly errors.
Making AI robotics simple with rust
Haixuan Xavier Tao (He/Him), Founder of 1ms.ai, France
Abstract: Building AI robots requires a high level of understanding of both hardware and software. Making Python only approach often hard to scale up due to hardware and time constraints. This is where Rust shine since it’s able to deal with low level constraints as well as high level AI software. In this talk, we’ll showcase how dora-rs makes the AI robotic software stack fast and easy through a Live Demo.
sophus-opt: Rust library for generic and sparse non-linear least squares
Dr. Hauke Strasdat (He/Him), Co-founder of Farm-ng Inc. and Space-ng Inc., United States
Abstract: Sophus-opt is an early-stage Rust library for generic and sparse non-linear least squares optimization, inspired by widely used libraries like g2o and Ceres. These libraries are essential in robotics and computer vision for tasks such as bundle adjustment, camera calibration, and pose graph optimization. Developing such a library in Rust presents unique challenges, particularly due to the lack of advanced generic features such as const generics expressions. Despite these hurdles, sophus-opt aims to bring efficient non-linear optimization capabilities to the Rust ecosystem.
Generalizable Planning Algorithm in Rust
Abstract: TBD
Embedded optimisation using Rust
Dr. Pantelis Sopasakis(He/Him), Ruairi Moran(He/Him), Queen's University Belfast, United Kingdom.
Abstract: In many engineering applications such as control (e.g., model predictive control) and estimation (e.g., moving horizon estimation and maximum likelihood estimation) we need to solve optimisation problems in real time and on embedded devices, often having limited resources. To this end we need fast numerical optimisation algorithms with good convergence properties and implementations with appropriate considerations for memory management. The need for speed and safety makes Rust an excellent programming language for embedded numerical optimisation. In this talk we will present Optimization Engine (OpEn) — see http://doc.optimization-engine.xyz/ — a solver for embedded numerical nonconvex optimisation written in Rust accompanied by a code generation library in Python that facilitates the designer. We will present the numerical algorithm of OpEn and will demonstrate its use for solving nonconvex constrained optimisation problems using in application of control and estimation such as nonlinear model predictive control for obstacle avoidance.
The Applied Planning, Learning, and Optimization Toolbox: A Flexible Rust-Based Software Suite for Robotics Research
Dr. Daniel Rakita (He/Him), Yale University, United States.
Abstract: In this talk, I overview the Applied Planning, Learning, and Optimization (APOLLO) Toolbox, a flexible Rust-based software suite, specifically for robotics research. Central to this toolbox is a preprocessor, which extracts important kinematic and dynamics data from robot platforms and environments, storing this information in lightweight, easily parsable files. This approach allows robotics developers to rapidly achieve compatibility across numerous applications, as much of the necessary data and functionalities can be directly loaded rather than re-derived in a new language. Additionally, the toolbox’s broad flexibility improves compatibility with a wide range of graphics front-ends, including Bevy, a pure-Rust game engine. I will share the current state of development and highlight how Rust fosters a collaborative and efficient development environment within our lab and the broader robotics community.