From MuJoCo to Real Hardware: Running RL Policies Through ROS 2
Ekumen enabled reproducible ROS 2 deployment of MuJoCo-trained RL policies on real manipulators.
Innovation unleashed
Exploring cutting-edge technologies and real-world applications, our blog is your gateway to the future of engineeering. Join us as we delve into the latest advancements and unconver how they are shaping our world.
-
-
NuRec: Bridging the Visual Fidelity Gap in Simulation
We tested NVIDIA’s NuRec, revealing practical trade-offs for 3D robotics simulation. -
Humanoids Are Leaving the Lab. Now Comes the Hard Part
Humanoids and AMRs are scaling fast; success depends on the software stack. -
Nav2 + Open-RMF: How We Built a Scalable Fleet Coordination System with Andino
Bridging Nav2 and Open-RMF to transform individual robots into synchronized, managed fleets. -
Stop Archaeologizing: Bringing your company documentation to Life with VertexAI
Ekumen's Vertex AI agent, EKU, turns fragmented company documentation into instant, accurate chat answers. -
From Unit Tests to CI: A Practical Guide for Professional ROS 2 Projects
A practical guide to building robust, automated testing pipelines for ROS 2. -
Why Some ROS 2 Nodes Are Hard to Test (and How Design Fixes It)
Best practices for designing nodes that are easier to test, maintain, and evolve. -
Modeling LeKiwi’s omni-base in MuJoCo
Real-world experiences while modeling the 3-omniwheel LeKiwi’s base using MuJoCo. -
The High Cost of Cutting Corners: Why Testing in ROS 2 Robotics is Non-Negotiable
Why testing is essential for building reliable, scalable, and maintainable ROS 2 robotics systems. -
The Hidden Cost of Reactive Infrastructure
Ignoring infrastructure early drains money, time, and focus from growing tech teams. -
Deep Dive into OpenRMF
OpenRMF connects diverse robots, overcoming silos to improve coordination and system performance. -
AR4 Robot vs the Chessboard: A Simulation in Isaac Sim
Simulating the AR4 robot playing chess in Isaac Sim with ROS 2. -
Taking the AR4 Further: From Gazebo to Isaac Sim
We're excited to share the next step in our journey extending the AR4 simulation into Isaac Sim and MuJoCo. -
andino-rs: Driving educational robotics with Rust and AI
Introducing andino-rs, a project bringing the Rust language to the andino educational robot for AI applications. -
Exploring Raw Data of an RGB-D Dataset from an iPhone
A technical exploration of raw RGB-D data from an iPhone using Open3D for visualization, processing, and preparation for 3D reconstruction. -
9 out of 10 Leading Companies Outsource. Why Not Yours?
Outsourcing boosts innovation, reduces risk, and accelerates projects with specialized global talent. -
Can You Map a Building with a Smartphone?
Exploring smartphone-based 3D mapping with neural fields and open-source software. -
Bringing the AR4 Robotic Arm to Life in Simulation - Part 1
Simulating the AR4 arm in ROS2 and Gazebo with MoveIt control. -
Migrating from ROS 1 to ROS 2: what you need to know (before the robot apocalypse)
How to approach robotic software upgrades without losing control—or your mind. -
Use Case: Closing the Sim2Real Gap with automated photorealistic data generation
Discover how photorealistic simulation is transforming ML training, offering unparalleled efficiency, scalability, and precision. -
Robotics solutions: finding the right talent for your project
Finding the right team for robotics and simulation projects is crucial. -
Ekumen at FOSDEM 2025: Accelerating robotics development through simulation
A solution agnostic approach to accelerate robotics development with simulation. -
A Tale of Two Formulas: Benchmarking the Nav2 AMCL Likelihood Field Sensor Model
A story of unexpected math, retro-engineering code that just-works, and benchmarking. -
Big shoes to fill: Validating the performance of Beluga AMCL
Beluga AMCL is a modern reimplementation of the localization package in ROS using the Beluga library, but does it work? We demonstrate it is a viable alternative to Nav2 AMCL by testing it against a large number of real-robot public datasets. -
Dataset Generation with NVIDIA Omniverse™
Case study of a synthetic dataset generation pipeline using NVIDIA Omniverse™. -
Upgrade your robot's navigation stack with Beluga AMCL
Beluga AMCL is a drop-in replacement for Nav2 AMCL and a gateway to more advanced localization solutions for robotics. -
A Beluga in Monte Carlo
Beluga is an open source implementation of particle filters written in C++17 that works as a drop-in replacement for AMCL ROS. -
Andino: Empowering Education with an Open-Source Educational Robot
We proudly introduce Andino, an innovative open-source differential drive robot designed specifically for educational purposes