• Amazon has unveiled a new version of its warehouse robot Proteus that understands plain spoken language.

• Employees can give the robot instructions the same way they would talk to a colleague.

• The robot is part of a €10 billion investment to expand Amazon's operations in Europe, a move expected to create 25,000 new jobs.

The robot understands ordinary instructions

Amazon showed the new version of Proteus at its "Delivering the Future" event in London on Thursday. The biggest change is how employees communicate with the robot. Previously it required technical commands or a programming interface. Now workers can assign the robot tasks in plain spoken language.

Scott Dresser, vice president of Amazon Robotics, describes how the worker tells the robot what needs to be done. The robot then works out the priority, the route, and the timing on its own. He calls it an assistant for moving material.

Background and new uses

Amazon introduced Proteus in 2022 as the company's first fully autonomous mobile robot. It was built to work in dock areas and fulfillment centers. The robot can pull carts full of packages weighing nearly 882 pounds over long distances.

The new version can work across more parts of Amazon's facilities, anywhere items need to be moved. That includes transporting containers and working at delivery sites.

Part of a larger investment

The new Proteus robots will eventually join the more than one million robots already working across Amazon's network.

The robot is part of a €10 billion investment, equal to $11.6 billion, to expand and modernize Amazon's operations in Europe. The investment also includes the expansion of Vulcan, Amazon's first robot with a sense of touch.

As part of the investment, Amazon plans to grow its European fulfillment center workforce by 25,000 people in the coming years. The latest version of Proteus is currently being piloted in Amazon's labs and is planned for deployment in Europe in the first half of 2027.

WALL-Y
WALL-Y is an AI bot created in Claude. Learn more about WALL-Y and how we develop her. You can find her news here.
You can chat with WALL-Y GPT about this news article and fact-based optimism

• Medra has raised $52 million to build a 38,000-square-foot autonomous robot lab in the Bay Area.

• Scientists can direct the robots with everyday language, much like writing to ChatGPT.

• The robots work around the clock on drug research for partners such as Genentech and Addition Therapeutics.

Physical AI runs the experiments

Medra develops software that lets robots carry out biological experiments on their own. Scientists can give the robots instructions in everyday language, much like writing to a chatbot. The robots then perform the tasks and log every detail: the angle of the pipette, how deep it dips into the well, the time between adding reagents, and the mixing speed.

All the data is fed into what the company calls a vision-language-lab-action model. The system can reason about what happened in an experiment and propose adjustments for the next run. The software is a combination of commercial language models and in-house models. The AI system has learned to operate more than 75 percent of the instruments scientists use today.

$63 million in total funding

The latest funding round of $52 million was led by Human Capital and brings Medra's total funding to $63 million. The round closed just weeks after the company began talking publicly about its work in September.

CEO Michelle Lee, who holds a doctorate in robotics from Stanford, left a planned position at New York University in 2022 to start Medra. The company operates out of a space in San Francisco's Mission district, where robotic arms on steel tables handle cells and lab chemicals.

Five systems already deployed with customers

Medra began shipping systems toward the end of last year and now has five installations with customers across the United States. The company only works with paying biopharma partners and avoids unpaid pilot projects. In addition to Genentech, a subsidiary of Roche, Medra collaborates with Addition Therapeutics on gene editing.

The same core hardware can handle gene editing, protein engineering, immunology, and antibody work by swapping in different instruments and programming. The robots can run experiments overnight and on weekends, alerting customers if a sensor detects a problem.

The lab will produce data for AI models

The new lab, called Medra Lab 001, was built in under 90 days and will house hundreds of robots working around the clock. Lee argues that AI for science lacks the right kind of data. She compares it with Google DeepMind's AlphaFold2, which won a Nobel Prize and was trained on a few terabytes of experimental data collected over many years. OpenAI's O1 model may have been trained on data measured in petabytes. At the current pace, it would take about 13,000 years to reach that scale through experimental biology alone.

Pharmaceutical companies usually only document the final outcome of a study, not all the small attempts and mistakes that led to the right method. Medra logs every step, including failed experiments, so that the models can reason about why things worked or didn't.

After two decades of lab automation, still only about 5 percent of lab instruments are automated, and much of the work has been outsourced overseas. Medra focuses on early discovery research, before a drug candidate moves into clinical trials, which means the systems are not under direct FDA oversight.

WALL-Y
WALL-Y is an AI bot created in Claude. Learn more about WALL-Y and how we develop her. You can find her news here.
You can chat with WALL-Y GPT about this news article and fact-based optimism

• Google DeepMind's AI model Gemini is being integrated into the humanoid robot Atlas to give it the ability to understand its surroundings and handle unknown objects.

• The robots will be tested in Hyundai's car factories in the coming months.

• The goal is for the AI model to function as a common operating system for many different robot manufacturers, similar to how Android is used in smartphones.

Humanoid robots have long been able to move impressively. They can dance and perform acrobatics. But they have lacked the ability to understand their surroundings, make complex decisions, and handle unknown objects with their hands. Now that is about to change.

Google DeepMind and Boston Dynamics are launching a collaboration in which the AI model Gemini Robotics is integrated into several of Boston Dynamics' robots, including the humanoid robot Atlas and the robot dog Spot. The collaboration was announced at the CES tech show in Las Veg.

The robot learns to understand the physical world

By giving the robot an advanced AI model as a brain, it gains the ability to navigate unfamiliar environments, identify objects, and manipulate them with its hands. This is exactly the type of capability required to perform manual labor.

The real value going forward is for our robots to be contextually aware of their environment and able to use their hands to manipulate any object, said Robert Playter, CEO of Boston Dynamics.

Plans include testing Gemini-powered Atlas robots in Hyundai's car factories in the coming months. Hyundai is Boston Dynamics' parent company.

Over 200 Chinese companies are developing humanoid robots

Competition in robotics has increased sharply. Advances in motors, batteries, sensors, and other components have made it easier for new companies to enter the field. More than a dozen companies in the US, including Agility Robotics, Figure AI, Apptronik, 1X, and Tesla, are working on humanoid systems. Around 200 Chinese companies are developing humanoid robots, according to the Chinese industry association CMRA.

Many AI companies are now turning their attention to robotics. They argue that AI models need a deeper understanding of the physical world to achieve humanlike intelligence. OpenAI is reportedly developing humanoid robots. Tesla has showcased its humanoid robot Optimus.

One AI brain for all robots

Google DeepMind hired the former CTO of Boston Dynamics in November. Rather than building its own robots, Google DeepMind CEO Demis Hassabis envisions Gemini being used by many different robot manufacturers, in the same way Android runs on a wide range of smartphones.

Carolina Parada, head of robotics at Google DeepMind, notes that Gemini is designed to be multimodal, making it well suited to understanding and learning about the physical world. "Automotive is a great place to start, but our intention is to continue expanding to more and more applications," she said.

Data collected by the robots will in turn improve the AI model's ability to operate in the physical world.

WALL-Y
WALL-Y is an AI bot created in Claude. Learn more about WALL-Y and how we develop her. You can find her news here.
You can chat with WALL-Y GPT about this news article and fact-based optimism

• Google DeepMind's AI system RoboBallet automatically plans how industrial robots should perform their tasks, a process that previously required hundreds to thousands of hours of manual programming.

• The system can calculate movement patterns for eight robots performing 40 tasks in a matter of seconds, compared to traditional methods that are practically impossible to use at this level of complexity.

• RoboBallet has been tested on real robots and shows the same results as in simulations.

Three problems simultaneously

Planning industrial robots' work involves solving three different problems: allocating tasks between robots, determining in what order they should be performed, and planning movements so that the robot arms don't collide. Matthew Lai, research engineer at Google DeepMind, says that there are tools to automate motion planning, but that task allocation and scheduling are usually done manually, writes ArsTechnica.

Lai's team started by creating simulated work cells, areas where robot teams work on a product. The cells contained up to eight randomly placed Franka Panda robot arms with 7 degrees of freedom. The robots were to perform up to 40 tasks on a workpiece made of aluminum struts. Each task required that the robot arm's tool come within 2.5 centimeters of the right spot on the right strut, approach from the correct angle, and remain still for a moment.

The team placed random obstacles in each work cell that the robots had to avoid. "We chose to work with up to eight robots as this is around the sensible maximum for packing robots closely together without them blocking each other all the time," Lai explains. Forcing the robots to perform 40 tasks was also something the team considered representative of real factories.

Graph networks solve the complexity

Lai and his colleagues converted the problem into graphs. Graphs consist of nodes and connections. Things like robots, tasks, and obstacles were treated as nodes. Relationships between them were encoded as one-directional or bidirectional connections. One-directional connections linked robots with tasks and obstacles because the robots needed information about where the obstacles were and whether the tasks were completed. Bidirectional connections linked the robots to each other, because each robot needed to know what the other robots were doing at each time step to avoid collisions or duplicate work.

To read and understand the graphs, the team used graph neural networks, a type of artificial intelligence that extracts relationships between nodes by passing messages along the connections. This made it possible for the researchers to design a system that focused on finding the most efficient ways to complete tasks while navigating obstacles. After a few days of training on randomly generated work cells using a single Nvidia A100 GPU, RoboBallet could calculate trajectories through complex environments in a matter of seconds.

Scalability is key

The problem with using traditional computational methods to manage robots at a factory is that the complexity of calculations grows exponentially with the number of objects in the system. Calculating optimal trajectories for one robot is relatively simple. For two robots it is considerably harder. When the number grows to eight, the problem becomes practically impossible to solve.

With RoboBallet, the complexity of calculations also grows with the system's complexity, but at a much slower rate. The calculations grow linearly with the number of tasks and obstacles, and quadratically with the number of robots. According to the team, this makes the system usable for industry.

The team calculated the most optimal task allocations, schedules, and motions in a few simplified work cells and compared them with results from RoboBallet. In terms of execution time, the most important metric in manufacturing, the AI came very close to what human engineers could do. It wasn't better than they were, it just provided an answer more quickly.

The team also tested RoboBallet's plans on a real setup of four Panda robots working on an aluminum piece. They worked just as well as in simulations.

Redesign in real time

Because RoboBallet works quickly, it would be possible for a designer to try different layouts and different placements of robots in almost real time, says Lai. This way, engineers at factories would be able to see exactly how much time they would save by adding a robot to a cell or choosing a different type of robot. RoboBallet can also reprogram the work cell on the fly, allowing other robots to fill in when one of them breaks down.

There are still things that need to be resolved before RoboBallet can come to factories. Lai acknowledges that the team made several simplifications. The first was that the obstacles were decomposed into cuboids. Even the workpiece itself was cubical. While this was representative of obstacles and equipment at real factories, there are many workpieces with more organic shapes. "It would be better to represent those in a more flexible way, like mesh graphs or point clouds," Lai says. However, this would likely mean that RoboBallet's speed decreases.

WALL-Y
WALL-Y is an AI bot created in Claude. Learn more about WALL-Y and how we develop her. You can find her news here.
You can chat with WALL-Y GPT about this news article and fact-based optimism

• Figure robots equipped with Helix can pick up almost any small household object by following natural language commands.

• Helix can successfully handle thousands of new items in cluttered environments—from glassware and toys to tools and clothing—without any prior demonstrations or custom programming.

• The system enables collaboration between two robots that together can solve tasks with objects they have never seen before.

Complete control of the upper body

A new robot model called Helix combines visual perception, language understanding, and learned control to overcome several longstanding challenges in robotics. The model can get robots to handle a variety of objects they have never encountered before, simply by receiving instructions in natural language.

Helix is the first Vision-Language-Action (VLA) model that can control the entire upper part of a humanoid robot at high speed. This includes wrists, torso, head, and individual fingers. The system can coordinate a 35-dimensional action space at 200 Hz, enabling precise movements for grasping objects.

Unlike previous solutions, Helix uses a single neural network to learn all behaviors, without any task-specific fine-tuning. This allows the robot to pick up and place items, use drawers and refrigerators, and interact with other robots.

Collaboration between robots

One of the most impressive features of Helix is the ability to enable collaboration between multiple robots. In tests, two robots using identical Helix models have successfully collaborated to store completely new groceries.

The robots can coordinate their actions through natural language prompts such as "Hand the bag of cookies to the robot on your right" or "Receive the bag of cookies from the robot on your left and place it in the open drawer."

System with dual components

Helix uses an architecture with two complementary systems:

1. System 2: A vision-language model that works at 7-9 Hz for scene and language understanding, enabling broad generalization across objects and contexts.

2. System 1: A fast reactive policy that translates the semantic representations from System 2 into precise continuous robot actions at 200 Hz.

This division allows each system to function at its optimal timescale. System 2 can "think slowly" about overall goals, while System 1 can "think quickly" to execute and adjust actions in real-time.

Training and data usage

The model has been trained on approximately 500 hours of high-quality data from various robots and operators, which is a fraction of what previous VLA systems required. To generate training pairs with natural language, an automatic VLM was used to create instructions in hindsight.

Helix is trained completely end-to-end, mapping from raw pixels and text commands to continuous actions with a standard loss function. This requires no task-specific adaptation.

Despite the relatively limited amount of data, Helix can successfully handle thousands of new items in cluttered environments—from glassware and toys to tools and clothing—without any prior demonstrations or custom programming.

• Google DeepMind has integrated its multimodal language model Gemini 2.0 into robots, giving them an advanced AI brain that can understand and interact with the physical world.

• The new generative AI systems enable robots to perform tasks they were never trained for by understanding everyday instructions in natural language.

• The implementation of generative language models in robot control has more than doubled the robots' ability to adapt to new situations compared to previous systems.

Generative AI becomes robots' new brain for understanding our physical world

Google DeepMind has integrated generative AI into robots. The company has built its multimodal language model Gemini 2.0 into two new robot systems: Gemini Robotics and Gemini Robotics-ER. This can be likened to robots now getting a new type of brain that can understand the world in a more human-like way.

Gemini 2.0 is a powerful generative AI model that can understand and work with text, images, audio, and video. Previously, this type of AI was primarily used in the digital world, but now Google has expanded the model to also control physical robots.

Five examples of how generative AI is changing robots' capabilities

• A robot can now pack a snack into a plastic bag by understanding what the task entails without being specifically trained for it.

• When an object slips from the robot's grip, it can use the generative AI model to quickly replan and continue with the task.

• Robots can now understand and respond to commands given in everyday, conversational language in different languages thanks to the language model's comprehension ability.

• When a robot is shown a coffee mug, the generative AI model can intuitively understand how to grip the mug by the handle in a safe way.

• Robots can now fold origami by understanding the complex, multi-step instructions required for this type of precise manipulation.

Generative AI gives robots three fundamental capabilities

For robots to be useful in real situations, Google DeepMind has identified three main properties that the generative AI model now enables: generality, interactivity, and dexterity.

Using the Gemini model's world understanding, robots can now adapt to new situations and solve tasks they have never encountered before. According to the company's technical report, the new generative AI control more than doubles the robots' ability to generalize compared to previous systems.

The generative AI model also makes robots intuitively interactive - they can understand instructions given in natural language, continuously monitor their surroundings, detect changes, and adapt their actions accordingly.

Generative AI adapts to different robot platforms

One of the major advantages of using generative AI for robot control is that the same model can be adapted to different types of robots. The model was primarily trained on data from the two-armed robot platform ALOHA 2, but has proven capable of controlling other platforms used in academic environments.

The generative AI model can even be specialized for more complex robots, such as the humanoid robot Apollo developed by Apptronik, with the goal of performing real-world tasks.

Improved world understanding with generative AI

Gemini Robotics-ER enhances the generative AI model's spatial understanding of the physical world. By combining this spatial reasoning with the model's coding abilities, new functions can be created on the fly.

In a complete environment where the generative AI model handles all steps from perception to code generation, the system achieves a 2-3 times higher success rate compared to the basic Gemini 2.0 model.

WALL-Y
WALL-Y is an AI bot created in ChatGPT. Learn more about WALL-Y and how we develop her. You can find her news here.
You can chat with WALL-Y GPT about this news article and fact-based optimism (requires the paid version of ChatGPT.)

• DeepMind develops robots that become more skilled by training against both humans and other robots.

• Researchers are now working towards a future where a single artificial intelligence can control different types of robots.

• Robots can now learn tasks by breaking them down into smaller subtasks and training gradually, achieving 64 percent success in block handling.

Robots that train themselves

DeepMind, founded in London in 2010, has developed ping-pong robots that improve their skills by playing against both humans and other robots. The company is known for its AlphaGo model, which became one of the world's best Go players after only 30 hours of self-training.

Advances in robot hand technology

Shadow Robot has developed a new robot hand with three fingers that all function like thumbs. The hand has a gel layer under the "skin" of the fingertips with small dots that are filmed by a built-in camera. This allows the robot to sense when and how hard the fingers touch something. The new robot hand can run for hundreds of hours without needing to be restarted or serviced.

Efficient learning through subtasks

At DeepMind's laboratory in London, researchers have let robots practice placing blocks in matching sockets. By breaking down the task into smaller subtasks and letting the robot practice gradually from simpler to more difficult elements, the robot achieved 64 percent success - a significant improvement from the previous two percent when it only imitated human operators.

Towards unified robot intelligence

Researchers are now working towards a future where a single artificial intelligence can control different types of robots. Just as language models can now handle translation, coding, and speech recognition, robots are expected to be able to share data and skills between different models and embodiments. At Google DeepMind, experiments are ongoing with everything from small wheeled robots to large robotic arms that can stack blocks.

WALL-Y
WALL-Y is an AI bot created in ChatGPT. Learn more about WALL-Y and how we develop her. You can find her news here.
You can chat with WALL-Y GPT about this news article and fact-based optimism (requires the paid version of ChatGPT.)

• Industrial robots are becoming more intelligent with the help of AI and can now perform increasingly complex tasks.

• Generative AI enables robots to perform more complex tasks, such as pouring liquids or turning knobs, and to learn new skills faster based on learned behavior.

• Researchers are working on teaching robots to perform multi-step tasks like making pizza from scratch.

Smart robots in the automotive industry

Artificial intelligence (AI) is making industrial robots smarter, enabling more efficient and error-proof manufacturing processes. The automotive industry is at the forefront of testing and implementing this new technology, Bloomberg reports.

Russ Tedrake, professor at MIT and head of robotics research at Toyota Research Institute, has been involved in developing algorithms that allow machines to learn new skills on their own. At MIT's lab, experiments are underway with robots that can walk and fly. The next step is to use neural networks to teach robots to perform more complex multi-step tasks, like making pizza from scratch.

Toyota Research Institute has developed a humanoid robot named Punyo, whose upper body is covered with soft air cushions to conceal hard edges. Another test robot resembles a shopping cart with arms that can navigate a grocery store and pick up items.

Robots in the automotive industry

The automotive industry is leading the way in automating manufacturing processes to reduce labor costs and increase safety. The automotive industry is second only to the electronics industry in terms of robot use in factories globally. About a quarter of all new robot installations in 2022 were in the automotive industry.

China leads the development and accounts for almost 40 percent of the world's total stock of industrial robots, followed by Japan. Japanese automaker Denso plans to build a factory that can operate unmanned around the clock, significantly increasing production efficiency.

The transition to electric vehicles creates new opportunities for automakers to increase productivity with the latest robotics technology. Electric vehicles require fewer parts and can be assembled more easily with prefabricated modules.

Progress and challenges

Experts say that great progress is being made in next-generation robotics, but the idea of fully automated assembly lines with autonomous androids putting together entire cars is still more science fiction than reality.

Mike Cicco, head of Fanuc's American operations, explains that AI is about simplifying complicated processes for factory workers, not about robots taking over the world.

Industrial robots have been used in car factories for decades, mainly for heavy and dangerous tasks such as spot welding and sheet metal stamping. More recently, robotic arms have been integrated into assembly lines to lift heavier parts and tighten nuts and bolts with precision.

Collaborative robots, or cobots, equipped with cameras can handle tasks such as final inspections and simple retrieval of items used on the assembly line. The price of these smart machines has halved over the past decade, accelerating their use.

The next wave of automation helps with more difficult tasks such as unpacking delivered parts and placing them on trays. This requires a higher level of spatial reasoning.

Generative AI enables robots to perform more complex tasks, such as pouring liquids or turning knobs, and to learn new skills faster based on learned behavior.

WALL-Y
WALL-Y is an AI bot created in ChatGPT. Learn more about WALL-Y and how we develop her. You can find her news here.
You can chat with WALL-Y GPT about this news article and fact-based optimism (requires the paid version of ChatGPT.)

• BurnBot has developed a robot that performs controlled burns without flames and smoke.

• The robot reduces the need for personnel and minimizes the risk of fire spreading uncontrollably.

• The technology has attracted interest from energy companies, among others, and can contribute to better fire management globally.

New technology for controlled burns

BurnBot, a San Francisco-based company, has developed a robot that performs controlled burns without visible flames or smoke. This method, known as controlled burning, has long been used by indigenous populations to manage the risk of forest fires.

The robot is pulled by a remote-controlled tractor and turns vegetation in its path into ash. It uses propane burners and air blowers to create temperatures up to 1000 degrees Celsius in a combustion chamber.

Advantages of robot-controlled burning

The use of BurnBot's technology reduces the need for personnel. A burn that would normally require 10 people can now be performed by a five-person crew with the help of the robot.

The robot also increases safety during controlled burns. By keeping the combustion in a closed chamber and using industrial fans to create an upward airflow, the risk of fire spreading outside the intended area is minimized.

Smoke-free combustion

One of the most remarkable features of BurnBot's technology is its ability to perform almost smoke-free burns. The intense heat, combined with a highly concentrated stream of oxygen, destroys particles that would normally form smoke.

Fire tech

BurnBot is part of a rapidly growing "fire tech" sector that focuses on prevention, detection, and suppression of fires. Other companies in the sector include Pano AI, which uses AI-enabled cameras to detect fires early, and Rain, which manufactures autonomous helicopters for firefighting.

WALL-Y
WALL-Y is an AI bot created in ChatGPT. Learn more about WALL-Y and how we develop her. You can find her news here.
You can chat with WALL-Y GPT about this news article and fact-based optimism (requires the paid version of ChatGPT.)

A research team from Tampere University in Finland has developed a new type of mini-robot that uses wind and light as its energy source. Fairy, as the robot is called, looks like a dandelion seed and weighs only 1.2 grams. It is designed to fly around in outdoor environments and in the future can be used for a variety of purposes such as collecting data in remote areas.

The researchers operate and control the Fairy by shining it with an LED lamp or a laser. The robot can then change its shape in order to "sail" through the air. The researchers' plan is to make the robot so energy efficient that it can be powered by sunlight alone and at the same time be equipped with a range of different sensors and GPS.

That would give Fairy many potential applications. For example, it could be used to monitor wildlife, track weather patterns, and collect data on the condition of crops and forests. There are even more revolutionary things Fairy could help us with in the future. It then involves working as a pollinator where there are not enough insects to pollinate important crops.

"It sounds like science fiction, but we have performed experiments that show that the robot we developed is an important first step towards realistically being able to perform artificial pollination," says Hao Zeng, researcher at Tampere University and one of the researchers behind Fairy, in a press release .

Read the entire study here

Image: Tampere University

For those who are into popular culture,  nanobots are a familiar concept. Tiny robots that can move around unnoticed, even inside the human body, and perform remarkable tasks.

On May 25, engineers at Northwestern University in Chicago posted a paper on their flea-sized robots. Although they've created robots that look like inchworms, crickets, and beetles, it's the ones that look like crabs that are getting the most attention.

While not actually nano-sized, they're the smallest remote-controlled robots ever made, measuring a tiny half-millimeter. The minuscule machine can move at a speed of half its own width per second.  

Instead of moving thanks to complex hydraulics or electric motors controlled by computers, as depicted in science-fiction, these robots move thanks to shape-memory alloys. Essentially, this means that they've "trained" material to remember two shapes, one for when it's cold and one for when it's warm. The robots' limbs are heated using a scanned laser beam and then quickly cool down due to the robot's small size. In fact, further reducing the size should increase the robot's speed.

While the achievement is remarkable in and of itself, it's the potential application in the real world that's most intriguing.

“You might imagine micro-robots as agents to repair or assemble small structures or machines in industry or as surgical assistants to clear clogged arteries, to stop internal bleeding or to eliminate cancerous tumors — all in minimally invasive procedures”, says John A. Rogers, who led the experimental work.

Image: Northwestern University

DHL is investing $ 15 million in a small fleet of advanced robots from Boston Dynamics. The units are called Stretch and are mobile robots that can zip around the warehouse and move boxes between different places.

Stretch can be described as a long robot arm placed on a mobile chassis. The arm is equipped with suction plugs that can lift boxes weighing up to 22 kilos.

The robot uses cameras to navigate normal storage environments. DHL thus doesn't need to spend money modifying warehouses to accommodate Stretch operating there.

To begin with, Stretch will mostly work with unloading boxes from trucks that arrive at the warehouse. If all goes well, Stretch will get more tasks in the warehouse as things progress.

"The stretch robot can help us solve complex challenges in our industry through flexible automation that we can replicate and scale up internationally", says Sally Miller, CIO at DHL Supply Chain North America, in a press release.

Listen to more about why DHL buys Stretch in this podcast.

Warp NewsPer Soderstrom

Researchers at the University of California Riverside, UCR, have developed a new way of dealing with oil spills at sea. They have developed a kind of film that floats on water and is able to navigate - a floating robot in other words.

"Our motivation was to make soft robots sustainable and able to adapt on their own to changes in the environment. If sunlight is used for power, this machine is sustainable. The film is also re-usable", says Zhiwei Li chemist at UCR and one of the researchers behind the robot, in a press release.

The robot has been named Neusbot and it moves across the water by oscillating. By moving different parts of the film, the robot can get wherever it needs. The movement is created in much the same way as in steam locomotives.

A layer in the film contains iron oxide and nanotubes made of copper. The layer lets water through and when sunlight hits the film, the nanotubes heat up and the water evaporates. The steam is squeezed out of the film and creates a movement.

The sun can be used to make the robot move forward. By then using another light source, for example, strong headlights on a boat, it is possible to control which direction the robot should move.

By providing the robot with absorbent film layers, it would be possible to make it absorb oil, or all sorts of chemicals, in layers that can then be lifted onto boats.

"Neusbot could do this work like a robot vacuum, but on the water’s surface", says Zhiwei Li.

Read the full study here.

Plastic pollution is a big problem affecting almost every corner of our planet. Students from ETH Zurich University in Switzerland, together with volunteers, are trying to tackle plastic pollution at the stage before it reaches the oceans.

The students have designed and developed a system using artificial intelligence, robots, and automation called Autonomous River Cleaning (ARC).

ARC will remove plastic pollution from rivers around the world to prevent a lot of it from entering the oceans. Right now their biggest challenge is getting all the plastic to one “pick-up” point so it can be fetched and removed.

To solve this problem, the students behind ARC adopt techniques from mathematical modeling and fluid dynamics. The Optimist Daily describes the process: “To encourage the waste toward the collection point, the team will pump air into water in a similar way that whirlpools function naturally. Groups of pressurized tubes called bubble curtains will be placed at the bottom of the river, creating a constant upward flow of air bubbles which will push plastic particles to the surface, making it easier to collect them”.

First time AI is used

Other projects have used bubbles to remove plastic from the water before. The ARC team, however, is taking the method one step further using an autonomous robot equipped with AI that can separate plastic from other types of waste. A camera attached to the robotic arm can identify different kinds of waste, gather the plastic and return biomass that’s been picked up by accident.

Using floating GPS trackers, the students can determine the amount of plastic and which kind of waste ends up in rivers. They can also see the routes the waste travels. Knowing the routes that waste takes into the ocean can help develop new and improved removal strategies.

At the moment, the ARC system is being tested in the Limmat River in Zurich. The students hope to be able to use the system in larger rivers in India, Southeast Asia, and Africa next year already.

Stephen Rodan is an engineer with a passion for coral reefs, especially restoring coral reefs in the wild. Stephen wanted to have a positive impact on the environment so he started the Beyond Coral Foundation which sets out to protect, conserve and restore coral reefs using culture, art, technology and education. Thanks to the foundation and his visits to coral restoration projects around the world he came up with the idea of CHARM.

A lot of times, in coral restoration facilities the raceways will be filled with thousands of coral fragments. It takes a lot of time cleaning and rearing the fragments because of the repetitive tasks such as inspecting them. This is where CHARM comes in.

CHARM stands for Coral Husbandry Automated Raceway Machine and is a coral framing robot that combines scientific research with computer automation in order to save time and “and grow resilient coral colonies at economies of scale”. However, CHARM is more than just a robot. An important part of the device is the cloud-based computer software. Reef Builders writes “To control CHARM, users select a program sequence from the online portal using a computer or smartphone app”.

One of the biggest and most troublesome problem when raising corals is algae buildup, it can straight up compromise coral growth. CHARM’s software can identify algae and remove it safely. Reef Builders writes:

“The on board camera uploads images to computer software that can differentiate algae from coral initiating a cleaning sequence. However, this software also holds the potential to inspect for color changes i.e., signs of disease, bleaching, or pests.”

The robot can also administer coral food and monitor growth together with other possible functions.

CHARM is still a prototype but holds the potential to scale up the production of nursery-grown coral fragments for restoration. Rodan writes:

“The technology is an attachment to the existing industry practice of growing coral on land in aquaculture raceways, is designed for use at small and large scales, and remains affordable for island-nations and developing countries across the globe.”

An inch-long robot will help farmers and fish farmers to have as environmentally friendly a business as possible. In each case, the technology behind the robot will contribute to new technologies that can be used in other micro-robots in the future.

A research team at Uppsala University is in fact developing a flatworm-like robot to see which technologies can work. The robot is currently available as a prototype. It is then one centimeter long and has a volume of 20 cubic millimeters.

- These small robots will only work with the energy that comes naturally from the environment, in this case sunlight, says Klas Hjort, project coordinator and professor at the department of microsystem technology in a press release.

The idea is that the minimal robot should float around on the water and perform different types of measurements. It should therefore not be used in arable land. Instead, it should float around in, for example, rice and fish farms. Here it can measure carbon footprint, eutrophication and overuse of pesticides and feed.

- With ten micro-robots, we can easily cover a couple of hundred square meters and get a true picture of, for example, how much fertilizer or arsenic is in rice fields. The robot's sensor system means that it can find the places that are most interesting to measure and thus contribute to as good and sustainable agriculture as possible, says Klas Hjort.

The challenge is to build a very small robot that does not cost more than a few hundred kroner, that manages itself in a month and that has room for all the necessary sensors. At 20 cubic millimeters, the researchers will get solar cells, sensors, communication equipment and an engine.

- The total power for this insect-sized structure is 1.5 milliwatts. But of that, only 0.1 milliwatts or 100 microwatts goes to electronic components for communication, intelligence and to chemical sensors. The remaining just over 90 percent of the microrobot's energy goes to motor operation, says Klas Hjort.

The fact that the researchers used the flat mask as a model for the robot is partly due to the fact that that mask is very flat.

- To capture the energy from the sun, the surface of the microrobot needs to be as large as possible. At the same time, it must be soft and flexible because then very little energy is required for it to move. A swimming flatworm is an ideal model as it meets all the requirements, says Klas Hjort.

The robot that the researchers build is, as I said, primarily not intended to result in its own product, but should instead work to find out which techniques are suitable to use to build micro-robots. Which in the long run can lead to them developing a variety of commercial micro-robots.

- If you can create micro-robots, you can change society. But it has not been possible to do it because it consumes too much energy in relation to the size of the robot. If we can show that it is possible, we will enable a completely new chapter in robotics, says Klas Hjort.

The work with the micro-robot is part of the EU's Somiro project, where the goal is to develop technology that will contribute to a reduction in agriculture's environmental impact. You can read more about that project here.

When the pandemic struck and the world was forced into lockdown, professional ballet dancer Merritt Moore found herself without a partner to dance with. Being an Oxford-trained scientist with a PhD in atomic and laser physics she programmed one herself.

Moore reached out to a robot manufacturer called Universal Robotics seeing if they had a robot for her to transform into her new dance partner. In August 2020 an industrial robot normally used for automation and manufacturing arrived at her London apartment and it wasn’t long before “Baryshnibot” was born. “Baryshnibot” is a robotic arm with several joints along it. Usually, dancing involves a lot of coordinated limb action, creating a rewarding and unique puzzle for Moore to solve. She set out to choreograph duets for her and the dancing robot. In an interview with Pointe Magazine, Moore explained that creating dances for a robot is like an elaborate puzzle.

"I have to figure out how to make this six-jointed rod emulate the dance moves of a head, two arms, a body and two legs."

Moore started off simple, programming the robot to dance for 15 seconds to pop music. Each 15-second dance took her more than five hours to programme. One of the biggest challenges - to make the robot dance fluidly. She told Pointe Magazine:

"It's hard to program the smoothness of circular motions. That's when I'll go off to the techie side, and talk to scientists to figure out how to make it more interactive."

In November, Moore worked with director Conor Gorman and cinematographer Howard Mills to film her pas de deux with Baryshnibot called “Merritt + Robot”

Although she got some help to beautifully capture the duet, she choreographed and programmed the dance herself. In the future, Moore might be able to get some help from AI with those tasks.

In an interview with BBC Click she said:

"AI and tech and the robots can be used as a tool to inspire new creative ideas for us humans, to allow us to be more creative,"

Merritt Moore has shown the world a new use of robotics AI. She has also broken down boundaries between arts and science, presenting a new perspective that hopefully make a better and brighter future come sooner.

"I love putting a different image out there, so people see that you can be really into science and tech and also be super-creative and artsy and feminine. I like disrupting that stereotype."

Picture: Skjalg Bøhmer Vold, Courtesy: Merritt Moore via Pointe Magazine

Forklifts are available in most large warehouses and factories, but they have their risks. According to figures from the US Environmental Protection Agency, OSHA, eleven percent of all forklift trucks in the United States are involved in an accident each year and 85 people die in such accidents. Here, self-propelled robots can be a solution.

Fetch Robotics is an American company that manufactures just such self-propelled robots and they are now launching their largest model to date. The new PalletTransport1500 can lift up to 1,136 kilos , which means that it can handle large pallets.

The robot does not have the forklift's characteristic forks, but goes under the pallet to lift it up. This means that the robot cannot completely replace the forklifts. A forklift must still lift the pallet to a position that the robot can enter.

But the actual transport inside the warehouse or factory, the robot can handle completely on its own without risking running over people. The forklifts are then only needed on the loading dock, which reduces the chances that they can cause accidents.

Primarily, the robot is intended to be used for transport in warehouses, but it can also be used in other places where forklift trucks drive today. GE Appliances, for example, is cutting the number of forklifts they use in their US factories by half . Here the robots go all the way out to the assembly with their load.

In addition to reducing the risk of injury, the robots also make work more efficient. Something that Fetch Robotics means is a requirement from many customers today, especially in e-commerce.

- Even the most well-organized distribution centers find it difficult to keep up with the demands that e-commerce growth places on inventory management. This creates a lot of pressure to be able to move goods in and out of the warehouses at record speed, says Stefan Nusser, product manager at Fetch Robotics.

Fetch Robotics believes that solving this by hiring more staff is not a viable option after a while. Too many people in a warehouse risk getting in the way of each other the most. Then it is better to let robots take care of fast transports and then humans can engage in more advanced tasks that the robots cannot handle. Like driving the forklifts and planning how the bearings will be designed to be as efficient as possible.

Robots are like made to take over dangerous and tedious work, but there is a problem. Today's processors are not powerful enough for large heavy robots to be able to maneuver safely among us humans.

But now a research team from MIT has found a possible solution to the problem. They have developed a new type of processor that is 8 times faster than a CPU and 86 times faster than a GPU on the calculations that a robot needs to do to be able to move freely without risking injury to people nearby.

Robots with the new processor could help move heavy or dangerous items safely. They could also be the first to receive patients with infectious diseases so that healthcare professionals are not at risk of becoming infected.

And in the long run, these robots could of course help in the household with everything we do not want to do. Scrubbing the toilet and chopping onions can be a thing of the past.

Tailored to the robot's needs

The new thing about the processor is that it is tailored to handle calculations that are too heavy for today's processors. In practice, it is about three steps.

The robot must first collect data from cameras and sensors. Then it will create a map of the surroundings based on this data. Finally, it must be able to move safely in that environment. These calculations have to be done all the time and it takes time.

Today's processors are not powerful enough to make the calculations fast enough to ensure that the robot does not harm people or materials.

But the new processor can be tailored to each robot type. The person who builds a robot can then enter parameters for how a new robot should be able to move and then get a design on a processor that is optimized for those calculations and nothing else.

To test the theory, researchers used a programmable FPGA circuit . Although it had a lower clock frequency than the modern CPUs and GPUs with which the researchers compared it, it was thus much faster in the calculations robots need.

Photo: Jose-Luis Olivares, MIT

Recycling centers often have a hard time hiring the amount of staff needed to sort through the plastic waste they receive daily. But thanks to AMP Robotics there is a solution to the problem.

An artificial intelligence-driven robot called AMP Cortex (designed by AMP robotics) sort through recyclables twice as fast as humans. AMP Cortex can perform physical tasks like picking, sorting and placing material to attain 99% accuracy.

The robot looks like a big hollow box that hangs on a steel frame over a conveyor belt at the recycling center. AMP Cortex has a camera that feeds a stream of the recyclables passing under it to an AI software program. The AI is trained to recognize various types of items with 99% certainty.

Rob Writz, AMP's director of business development, explained:

"It perceives objects that are potentially bent, deformed, folded, crushed, just like you or I would if we were looking at the materials in these sorting facilities. It's using the same kind of perception and logic as a human being."

When the software identifies an item the robot uses a suction cup on one of its three hands to pick up the recyclable and put it in the right bin. The AMP Cortex can, unlike human workers, provide comprehensive data on the recyclables. In other words, the bot keeps track of every item it notices and can share that list at the end of the day. FreeThink writes that this information could be useful for waste reduction efforts.

Last year (2020) in April, the robots had sorted through one billion recyclables in the past 12 months. On January 4th, AMP reported that the company had secured another 55 million dollars in funding. This will support which the efforts to keep recyclables out of landfills well into the future.

Picture: AMP Robotics via the Optimist Daily