EP.44 ANOTHER CAR MAKER STARTS MAKING ROBOTS
A sink-cleaning robot, Chinese EV-maker's humanoid, huge updates from Universal Robots & much more...
Chinese carmaker unveils a humanoid robot! 🚗
XPENG, a Chinese electric vehicle (EV) maker, recently revealed its new humanoid robot named "Iron" at the company’s 2024 AI Day event. Designed with over 60 joints and 200 degrees of freedom, Iron is capable of mimicking human movements with impressive precision.
The robot is already operational on XPENG's assembly lines, specifically aiding in the production of their EV cars. Powered by XPENG's Turing AI chip, which boasts a 40-core CPU, Iron is equipped to handle complex tasks in manufacturing and potentially other settings, like customer service. XPENG also announced plans for Iron’s future roles in retail, administrative, and even personal assistance applications.
Additionally, XPENG's innovations extend beyond robotics, as it unveiled the Kunpeng super-electric system—a high-efficiency engine and charging system for EVs, capable of ultra-fast charging, adding one kilometer of range per second. Looking ahead, XPENG aims to enter new industries and expand its presence to 60 countries by next year, setting ambitious goals in AI, robotics, and autonomous technologies.
From A-B with cobots 🧪
Up to 5,500 engines, including petrol, diesel, and electric vehicle components, are produced daily at the BMW Goup plant in Austria, one of the automotive companies largest. A key part of the production process involves moving engine connecting rods (conrods) between machines for further processing. The plant previously used a curved conveyor system for this task, but the rods would often get stuck or scratched.
To fix this issue, the plant switched to a two-belt linear conveyor system, which required a robot to pick the conrods from one belt and place them onto another. Given the limited space and the need for occasional human intervention, BMW chose the ABB GoFa™ cobot. Unlike traditional robots, cobots can work safely alongside human workers without the need for bulky safety barriers.
The introduction of the cobot solution has improved production efficiency, cut down on stoppages, and saved on floor space, all while ensuring higher quality by preventing damage to the connecting rods. Plus the ABB cobot can be simply reprogramed to handle different conrod designs, instead of switching the entire conveyor system around.
Read more here → Full success story
Robot learns to clean bathroom sink 🚰
Researchers at TU Wien in Austria have developed a robotic arm capable of cleaning a bathroom sink by learning through observation, rather than relying solely on detailed programming. The robot initially watched a person clean just the front edge of a sink, which had been covered with a dyed gel to imitate dirt.
Using this limited data, the robot then learned to clean the entire sink independently. A specialized sponge equipped with force and position sensors helped the robot understand how to adjust its angle, speed, and pressure based on the surface it was cleaning. The data collected was fed into a neural network, which transformed these observations into motion patterns the robot could follow.
The team sees this achievement as a promising step for robotic automation in various industries. This technique, known as "federated learning," allows robots to share and improve on each other's knowledge, potentially enabling them to perform complex tasks like painting, sanding, or welding in different environments.
Universal Robots is cooking together with NVIDIA! 👨🏻🍳
Universal Robots has introduced its AI Accelerator, a hardware and software toolkit designed to expand the capabilities of AI-powered collaborative robots (cobots). This toolkit, intended for both commercial and research applications, is tailored to help developers accelerate AI product development, reducing time to market.
Integrated with NVIDIA’s powerful Jetson AGX Orin and Isaac Manipulator tools, the AI Accelerator enhances UR's software platform, PolyScope X, bringing advanced AI features like object detection, pose estimation, and path planning to robotics solutions.
Equipped with a high-quality 3D camera, the toolkit offers a seamless, ready-to-use experience. The toolkit allows developers to freely choose their programming languages, making it highly adaptable for different AI projects. CEO Kim Povlsen sees this as a major step forward in bringing innovative, user-friendly AI solutions to market for Universal Robots’ partners and clients.
Read more here → AI Accelerator
Meme of the week 🤖
Jelly, jealous robots!
Build and Deploy SLAM Workflows with MATLAB! 🛻
Get a quick primer on Simultaneous Localization and Mapping (SLAM), a technology that enables robots and autonomous systems to build maps and track their positions in unknown environments.
In the latest MathWorks blog post, expert Jose Avendano Arbelaez delves into SLAM’s essential concepts and how to implement them in MATLAB.
From LiDAR and Visual SLAM to multi-sensor fusion, learn about the different algorithms and workflows you can apply to robotic and autonomous systems.
After reading this blog, you'll be able to:
Understand SLAM’s core concepts and applications in robotics and autonomous systems.
Build and test SLAM workflows in MATLAB using LiDAR, Visual, and Multi-Sensor SLAM.
Optimize SLAM for deployment, including error correction and ROS integration.
ROSCON in Odense was a blast! 🔥
I was invited by the City of Odense to participate in the Week of Robotics (including ROSCON - a place where every roboticist should go at least once in his life).
First ROSCON for me. Something amazing. The level of involvement, support, and robotics bashing - out of this world.
I met many inspiring people from the world of robotics, the founders of Universal Robots and a lot of passionate people.
The world of robotics has never been like this before. Robotics is ready to change the world!
Today I know that!
