
Large-Scale Autonomous Gas Monitoring for Volcanic Environments: A Legged Robot on Mount Etna
Active volcanoes present some of the most hazardous field conditions on the planet, and reliable gas monitoring is essential for forecasting eruptions. In a new study, researchers deployed a quadruped legged robot equipped with a commercial quadrupole mass spectrometer on Italy’s Mount Etna to demonstrate autonomous volcanic gas sampling, addressing the long‑standing challenge of accessing unstable, toxic terrain without endangering personnel. The team integrated a modular autonomy stack that couples robust global localization with a terrain‑aware global‑to‑local navigation framework, allowing the robot to traverse loose soil, steep slopes, and rugged lava fields. Four field missions were conducted: three fully autonomous runs covering a crater rim, a crater descent at the Sylvestri crater, and a volcanic desert at the Legetto crater, each achieving autonomy rates above 90% and successfully locating multiple artificial gas releases. A fourth, tele‑operated mission inside the Legetto crater measured natural fumaroles, confirming the robot’s ability to collect scientifically valid data under extreme locomotion challenges. Notable results include the mass spectrometer’s detection of clear sulfur dioxide and carbon dioxide signatures that matched handheld reference measurements, validating the instrument’s performance in situ. The autonomous missions demonstrated the robot’s capacity to navigate to and sample from remote, hazardous locations, while the tele‑operated run highlighted its utility for detailed, on‑demand investigations of unstable terrain. The successful deployment signals a paradigm shift for volcanic monitoring: legged robots can now provide continuous, high‑resolution gas data while keeping scientists out of harm’s way. This capability promises earlier eruption warnings, richer datasets for modeling volcanic behavior, and a template for extending autonomous sensing to other extreme environments such as glaciers, disaster zones, and planetary surfaces.

Multi-Task Reinforcement Learning for Quadrotors
The video introduces a multi‑task reinforcement learning (RL) framework designed for quadrotor drones, showcasing a single “generalist” policy capable of handling disparate flight objectives such as high‑speed stabilization, agile racing, and velocity‑tracking commands. By partitioning sensor inputs into shared and...

Learning on the Fly: Rapid Policy Adaptation via Differentiable Simulation (RA-L 2026)
Learning on the Fly: Rapid Policy Adaptation via Differentiable Simulation (RA‑L 2026) introduces a novel online‑learning framework that lets quadrotor controllers adjust to previously unknown disturbances within seconds of real‑world deployment. The authors combine a differentiable low‑fidelity dynamics model with...

Build an App with Arduino App Lab - Arduino Uno Q Part 2
In this second installment of the Arduino Uno Q workshop, the presenter walks viewers through the Arduino App Lab—a nascent development environment that lets users build "apps" composed of Python code for the board’s Linux‑based micro‑computer and C++ for its...

Alpamayo Debuts at CES 2026
The video announces NVIDIA’s debut of Alpamayo at CES 2026, positioning the platform as the next frontier in autonomous‑driving technology. Alpamayo is presented as an open ecosystem that couples advanced AI models capable of reasoning with large‑scale real‑world driving data...

Atlas | Product Features | Boston Dynamics
This short video introduces some of the high level futures of the production version of our Atlas robot. Learn more about Atlas: https://bosdyn.co/4jrAuRG

The Beginning of Your Tomorrow I Boston Dynamics
The video, titled “The Beginning of Your Tomorrow,” serves as Boston Dynamics’ flagship announcement of Atlas, its most advanced humanoid robot. It frames the company’s three‑decade quest to translate the effortless movement of animals and humans into engineered form, positioning...

Introducing XGSynBot Humanoid Robot #ces2026
This video introduces the XGSynBot wheeled bi-arm mobile manipulator, announced this week at #CES2026.

Spatially-Enhanced Recurrent Memory for Long-Range Mapless Navigation
The video introduces a new research effort titled “Spatially-Enhanced Recurrent Memory for Long‑Range Mapless Navigation,” which tackles the challenge of enabling autonomous agents to traverse complex environments without pre‑built maps. The authors propose a neural architecture that couples a recurrent...

AI on the Edge: YOLO Object Detection on Raspberry Pi with IP Cameras
In this tutorial, Paul McQuarter walks viewers through extending his "AI on the Edge" series by deploying the YOLO object‑detection model on a Raspberry Pi 5 that ingests video from an IP (RTSP) camera rather than the native Pi camera. He begins...

Humanoid Robots at Dubai’s Biggest Tech Expo
The Dubai JITEC expo turned the city’s futuristic ambitions into a tangible showcase of humanoid robots and AI‑driven systems, underscoring the United Arab Emirates’ drive to shift from a consumer of artificial intelligence to a global producer. Visitors were...

Intro to Hesai Maxwell & Hertz
In 2025, Hesai claims to have become the world's first lidar company to surpass 2 million cumulative lidar deliveries, enabling large-scale deployments across automotive and robotics applications amid rapidly growing global demand for advanced autonomy. To date, Hesai has secured design...

AI on the Edge: Watch This Before Buying a Raspberry Pi HAILO Accelerator
In this tutorial, Paul McCarter of topteboy.com walks viewers through the process of deploying AI on the edge using a Raspberry Pi 5, focusing on running the YOLO‑v11 object‑detection model locally without any external GPU accelerator. He emphasizes the distinction between cloud‑based...

I Built a Real Star Wars AT-AT
The video showcases a maker’s ambitious project: a full‑scale, walkable replica of the iconic Star Wars AT‑AT. Using a combination of 3D‑printed plastic components, off‑the‑shelf actuators, and custom‑designed gear trains, the creator has assembled a four‑legged walking machine that can...

What If Evolution Was Metal Instead of Carbon?
The video "What If Evolution Was Metal Instead of Carbon?" offers a speculative narrative that mirrors biological evolution, but with robots as the subject of natural selection. It opens with a primordial single‑core robot and traces a millennia‑long journey from...

Machining Parts for Wall-E. Episode 03
The video documents the third episode of a DIY series in which the creator fabricates the tracks for a life‑size, functional Wall‑E robot. Starting with a salvaged 1‑inch aluminum plate, he outlines the plan to machine, weld, and assemble...

Multi Pick with PAL Ready
In this tutorial, David, a technical trainer at Rub Boutique, walks users through the quickest way to configure a multi‑pick application using a robotic smart infeed conveyor paired with a Power Pick multi‑gripper, both of which are PAL‑ready or optional on a...

Adaptive Grippers Introduction
The video serves as a product briefing from David, a technical trainer at Rubboutique, introducing the company’s line of adaptive robotic grippers. Designed for flexibility, reliability and seamless integration with leading collaborative‑robot (cobot) platforms, the grippers aim to handle everything...

Automatic Pick Position
The video introduces Roboutique’s new automatic pick‑position feature for palletizing robots, which eliminates the traditional manual teaching of waypoints by calculating the pick location from user‑entered box dimensions. This capability is positioned as a response to the imprecision and ergonomic...