EP.32 ROBO-TABLE THAT MOVES THINGS

Eyes for robots, a robot stitches up a corn kernel, a wheeled humanoid & much more ...

'Eyes’ for robots for transparent objects 👀

Understanding the intricacies of surface properties, including transparent and specular surfaces, is crucial for advancing 3D imaging technologies in handling real-world scenarios.

Zivid is a manufacturer of advanced 3D cameras that can detect virtually any object. They have become famous for their ability to also detect the most challenging transparent and translucent objects with very high accuracy. Their software outputs accurate data from the machine vision sensors, consisting of high-resolution 2D color images and 3D point clouds, essential for logistics automation. 

One of the significant challenges in piece-picking robot applications is the vast number of products that need to be identified and handled, ranging from clothes, food, and make-up to boxes and bottles. Some e-commerce businesses have millions of separate items and SKUs, and the machine vision system needs to capture it all. 

Another fundamental challenge is transparency, which is prevalent in e-commerce packaging. Transparent materials pose a significant challenge in 3D imaging, as they can create multiple reflective surfaces, making it difficult for machine vision systems to accurately detect and pick them. To overcome this,  cutting-edge imaging methods combining hardware, optics, and software are being employed by Zivid.

Learn more about transparent object detection!

---

A robot stitches up a corn kernel 🧵

Sony has unveiled a microsurgery robot capable of performing super-precise operations, demonstrated by stitching up a tiny slit in a corn kernel. This innovative robot is the first of its kind to auto-switch between different tools and has successfully been tested in animal surgery.

Designed for the field of super-microsurgery, where surgeons operate on tiny blood vessels and nerves, this robot aims to assist in procedures that require incredibly steady hands. The system features a low-latency remote surgery device with a stereoscopic 4K 3D camera system, allowing surgeons to perform high-precision work with ease.

The prototype has been tested at Achi Medical University, where it demonstrated its ability to enable non-specialized medical staff to perform complex microsurgery tasks. Sony plans to continue R&D work to improve the machine and verify its effectiveness, with the goal of contributing to the advancement of medicine through robotic technologies.

---

ABB Robotics AI Startup Challenge

💸 Winning prize: $30,000 cash prize and the opportunity to establish a long-term partnership with ABB Robotics, exploring joint go-to-market strategies and investment opportunities.

Join ABB's global network, unlock advanced robotics, and secure potential long-term partnerships:

🔹 Challenge Overview: Revolutionize robotics by integrating AI technology, focusing on natural language programming, efficient skill acquisition, and autonomous decision-making.

Participate in one or more of three challenge tracks:

1. Prompt-based Natural Language Programming

2. Skill Learning & Acquisition

3. Planning & Decision Making

🔹 Who Should Apply? Top-tier startups and scale-ups globally, developing AI technologies compatible with current robotic systems, targeting ease of use, flexibility, and efficiency within diverse industries. Focus on practical, industry-specific applications showcasing AI's impact on robotics.

🔹 Prizes: Win a 30,000 USD collaboration project with ABB Robotics, six months of free membership at ABB's startup accelerator SynerLeap, and additional growth and networking opportunities.

Join ABB Robotics in shaping the future of robotics and automation. Don't miss out on this fantastic opportunity – apply now and submit your proposal here!

---

Robotic table from the future 🧊

Researchers have developed ArrayBot, a distributed manipulation system featuring a 16x16 array of vertically sliding pillars integrated with tactile sensors. This innovative system can simultaneously support, perceive, and manipulate tabletop objects. To achieve generalizable distributed manipulation, the team employed reinforcement learning (RL) algorithms to automatically discover control policies.

To address the challenge of redundant actions, they reshaped the action space by focusing on spatially local action patches and low-frequency actions in the frequency domain. This approach enabled the training of RL agents that can relocate diverse objects using only tactile observations. Notably, the discovered policy demonstrated generalizability to unseen object shapes in simulations and transferred successfully to the physical robot without domain randomization.

The deployed policy has been used to accomplish a wide range of real-world manipulation tasks, showcasing the vast potential of RL on ArrayBot for distributed manipulation.

---

---

Meme of the week 🤖

Consider it a lesson learned!

---

---

Learn how to simulate mobile robotics virtually! 🦾

In this course, participants delve into the exciting realm of virtual robotics using MATLAB® and Simulink® Robotics Playground.

After completing this course, you will be able to:

• Setting Up and Programming: Establish a foundation by setting up MATLAB® and Simulink® Robotics Playground.

• Sensor Integration and Visualization: Program a variety of common robot sensors using Simulink®.

• Understanding Virtual Robotics: Acquire skills to operate within a virtual space.

---

Lovely humanoid with a wheeled base! 🛞

Reflex Robotics, a startup, showcased its humanoid robot at Modex in Atlanta, impressing attendees with its speed and accuracy. The robot, featuring an in-house designed "torso" mounted to a base, can dynamically move up and down to access shelves at various heights and navigate tight spaces.

While the system is primarily teleoperated, with a human operator controlling it remotely, the goal is to reduce the human-to-robot ratio as the robot becomes more autonomous. Currently, the ratio is 1:1, but the company aims to achieve 1:2 in warehouses and 3PLs by early next year.

The second-generation system is being piloted with potential customers, including some major companies, and the plan is to have 10 to 20 robots in operation soon, followed by hundreds next year. With a current price point below $50,000, Reflex Robotics believes its system can scale efficiently. The New York-based startup, founded by MIT alumni with experience from companies like Boston Dynamics and Tesla, has raised a seed round and plans to pursue a Series A later this year.

---