Every day, hundreds of thousands of kits are prepared in warehouses before components ever reach an automotive production line. While robots have become commonplace in modern manufacturing, many upstream logistics activities still rely heavily on human operators performing repetitive pick-and-place and kitting tasks.

What if robots could learn these operations the same way humans do: by simply watching a demonstration?

That question was at the heart of I-GENIUS, a research project coordinated by MYWAI within the European ARISE initiative with Centro Ricerche FIAT (CRF) and the University of Genoa’s Department of Mechanical, Energy, Management and Transportation Engineering (DIME).

The project explored new approaches to Human-Robot Interaction, combining AI, computer vision, and robotics to enable machines to acquire manipulation skills from minimal human guidance.

One of the project’s key outcomes was VILMA (Visual Imitation Learning for Manipulation Activities), an AI-powered toolkit integrated into MYWAI’s EDGE AI middleware platform. VILMA helps enable robots and humanoids to learn complex manipulation tasks from one-shot human demonstrations, aiming to significantly reduce programming effort while improving flexibility in dynamic industrial environments.

The technology was evaluated in a large-scale automotive warehouse use case developed together with CRF and reproduced within the robotics laboratories at DIME.

Today, MYWAI is bringing this technology to a broader community of developers, makers, and robotics innovators by porting the VILMA Toolkit to new Arduino products powered by Qualcomm Dragonwing processors, including both the Arduino^® UNO Q and VENTUNO Q boards.

This demonstration showcases the potential for edge-native robotics applications that use imitation learning techniques on hardware platforms built to support compact form factors and efficient power consumption.

At the iGenius final presentation, the founder and CEO of MYWAI, Fabrizio Cardinali, stated: “The dual-brain architecture of the UNO Q and VENTUNO Q platforms is an ideal foundation for MYWAI’s next generation of Edge AI robotics. After validating distributed intelligence concepts within the ARISE I-GENIUS project, we are now leveraging these platforms to bring World Action Models closer to the edge through the latest release of the MYWAI EdgeAI Management Platform and its mobile tracker, HEDGELOG. By combining One-Shot Video Imitation Learning with edge-native AI execution, we aim to enable robots and intelligent industrial machines to acquire, distribute, adapt, and execute complex manipulation skills with unprecedented flexibility and scalability.”

Watch the full demonstration of the I-GENIUS project and see VILMA in action in this video.

The MYWAI VILMA agent

VILMA is a visual imitation learning toolkit that helps enable robots to learn manipulation tasks from human demonstrations. It is designed to process RGB-D recordings or MP4 videos to extract hand and object trajectories, generate reusable robot skills using Dynamic Movement Primitives (DMPs), and produce robot-ready trajectories for playback. It is constructed to serve as the demonstration learning module, supporting rapid robot programming, skill reuse, and deployment.

The AI pipeline

One-shot demonstration acquisition

The one-shot demonstration acquisition step is set to record a human performing the task or retrieve an existing demonstration from a selected MYWAI equipment event. It stages the video, RGB frames, depth data, and camera parameters, and allows the user to select the target object for tracking. This information provides the inputs required by the remaining pipeline stages. 

Hand detection

Using MediaPipe, this stage is structured to detect 21 hand landmarks in each RGB frame and combine their 2D positions with depth data to calculate 3D camera coordinates. For demonstrations loaded from MYWAI, the staged RGB and depth data are processed through the same pipeline. Kalman smoothing and previous-position retention improve tracking robustness, and the resulting trajectories can be saved back to the MYWAI event. 

Object detection

Using a YOLO model, this stage is designed to detect or track the object selected through the local or MYWAI interface. It combines the bounding-box centre with depth information to calculate the object’s 3D position, applies Kalman smoothing, and saves the trajectory and annotated frames. These results can then be included in the pipeline artifacts stored in MYWAI. 

Trajectory and segmentation

This stage is constructed to load the smoothed hand and object trajectories, estimate the grasp point from the hand’s proximity to the object, and detect the release point from the object’s movement and stabilization. It uses the hand trajectory as the main motion path and divides it into reach, grasp, move, release, and post-release phases. The trajectories, event indices, and segmentation metadata can be packaged as MYWAI event data, a time and space data fusion format developed by MYWAI for its AI-IoT management platform particularly geared towards Multimodal AI and, next, towards World Action Models. 

DMP generation

The DMP-generation stage is designed to learn separate Dynamic Movement Primitive models for the reach and move phases. It evaluates different regularization values, selects the model that provides the best accuracy and smoothness, validates the reproduced motion, and saves the trained models and trajectories. These DMP artifacts can be uploaded to MYWAI with the other pipeline results for later retrieval and reuse. 

Demonstration

The demonstration stage is structured to convert the generated DMP trajectory into Cartesian robot positions using the configured scale, offset, and rotation, then apply inverse kinematics to calculate the joint trajectory. The robot model may be loaded from the selected MYWAI equipment, and the resulting motion is displayed through the MYWAI 3D Viewer, synchronized with the recorded video and its grasp and release events. 

DMP adaptation with new goal and new object

The adaptation stage is set to load the learned skill – either from the current pipeline or a restored MYWAI event – and detect a new target object using RGB and depth data. It calculates the 3D offset between the original and new objects, redirects the reach and move trajectories toward the new pick and release positions, and preserves the demonstrated motion characteristics. The adapted trajectory can then be visualized with the MYWAI 3D Viewer or sent to the robot. 

Live streaming adaptation and UNO Q and VENTUNO Q support

The Live stream phase represents the deployment and real-time inference stage of the VILMA Agent. While the initial learning phase is conducted on the MYWAI platform to generate Dynamic Movement Primitives (DMP), the Live stream phase focuses on shipping these DMPs along with a fine-tuned YOLOv8 model, supported today on UNO Q and VENTUNO Q.

Architecture and components

As illustrated in the system schematic below, the architecture is designed as a distributed setup divided into an Edge AI Layer for intelligence and a Communication Layer for hardware interfacing.

Let’s break down how the live stream pipeline is working considering the VENTUNO Q version.

1. Edge AI layer (VENTUNO Q)

Running on VENTUNO Q, this layer is structured to handle high-level decision-making.

• VILMA Control Loop: The primary application logic responsible for the overall control loop. It It is designed to orchestrate object detection and performs DMP Adaptation to translate learned human motions into the current physical environment.
• Video Object Detection Brick: This component runs on VENTUNO Q to manage the inference flow. It receives the incoming video feed and communicates with the inference service.
• Docker: YOLOv8 Inference Service is formed as a containerized service that runs the quantized YOLOv8 model. This model is designed to be fine-tuned and deployed via the Edge Impulse platform using the “Bring Your Own Model” feature.

2. ROS 2 communication layer (Workstation)

A separate workstation connected directly to the devices manages the high-bandwidth data streams and robotic control via ROS 2.

• ROS 2 Streaming Node: Interfaces with the ZED Camera/Depth Sensor to capture raw visual data, publishing it as a /camera_feed to VENTUNO Q.
• ROS 2 Command Node: This node acts as a wrapper around the Fairino Python SDK. It is engineered to serve as the receiver for the /learned_trajectory sent from the edge device, utilizing the SDK to directly control the robot and help ensure it accurately follows the planned trajectory.

3. Physical hardware

External hardware is connected to complete the runtime VILMA ecosystem, namely:

• ZED Camera: The stereo camera which is engineered to capture the image and depth data required for the vision system.
• Fairino FR10 Robot: The robotic arm that is constructed to execute the pick-and-place tasks based on the trajectories computed by VILMA.

Component communication and data flow

The communication between these components is designed for low-latency execution as shown in the schema above:

• Vision Input: The Workstation is engineered to stream the /camera_feed (image and depth) to VENTUNO Q.
• Edge Inference: The VILMA Control Loop is designed to utilize a WebSocket stream to send frames to the Docker YOLOv8 Inference Service. The service returns the detected object bounding box to the control loop.
• Motion Adaptation: The system is structured to take the detected object positions and adapts the human-learned DMP to calculate a precise pick-and-place trajectory.
• Robotic Execution: The resulting /learned_trajectory is published back to the Workstation’s ROS 2 Command Node, which drives the Fairino FR10 robot to complete the task.

This modular approach is designed to allow the heavy vision processing and motion adaptation to happen on the edge (VENTUNO Q) while leveraging the robust ROS 2 ecosystem for robot communication and sensor streaming.

To learn more about the project and MYWAI’s EDGEAI Platform and Middleware, visit myw.ai.

Qualcomm branded products are products of Qualcomm Technologies, Inc. and/or its subsidiaries. 

Arduino, UNO, and VENTUNO are trademarks or registered trademarks of Arduino S.r.l.

The post MYWAI™ VILMA™ is designed to bring human-like learning to robots via one-shot demonstration appeared first on Arduino Blog.

Since 2017, the Vespa-maker Piaggio's Fast Forward division has been exploring a unique way to cut down on car trips: Building robots that can follow you around and carry your stuff. We called its original Gita (pronounced "jee-tah," Italian for "trip") robot an attractive rolling porter, but its size and initial $3,250 price made it more of a robo-curio than something you'd actually want to buy. The company followed that up with the Gitamini, a smaller and slightly cheaper $1,850 model, but even that remains niche.

Most people just don't have thousands of dollars to spend on a cute follow bot — but the calculation may be different for Star Wars fans. Squint a bit, and Piaggio's devices don't look that different from BB-8, R2-D2 and the multitude of other droids. So it's not a huge surprise to see Piaggio Fast Forward debut a special Star Wars edition robot: the G1T4-M1N1. It's basically just a $2,875 version of the Gitamini (which now sells for $2,475) with Star Wars decals and sound effects.

Devindra Hardawar for Engadget

G1T4-M1N1 boasts all of the same specs as the Gitamini: It can carry up to 20 pounds in its top compartment; it lasts for 7 hours, or 24 miles; and its speed tops out at a jaunty 6 MPH. But like a true Star Wars droid, it's more talkative than its older sibling. Lucasfilm offered up some of its unique sound effects for the bot, so you might hear it sound extra flustered as it's rushing to catch up to you, or turning quickly around a corner. It's a big leap beyond the Sphero-powered BB8 toy from 2015.

I had a chance to spend a weekend with G1T4-M1N1 at my home, but unfortunately I couldn't take it outside or use it in public, where it's really meant to shine. Instead, it was stuck puttering around my home and deck. While that delighted my kids — who both jumped in the air the moment G1T4-M1N1 awoke with a loud boot-up sound — they were still frustrated they could't pile their toys into its compartment and walk it to the neighborhood park.

From my brief time with it, I can confirm G1T4-M1N1 followed me around like an eager droid, ready to carry whatever I needed. But it also had some trouble navigating around my cluttered floors — if it hit any major obstacle or corner, it would go to standby mode until I tapped the capacitive button on its noggin. Unlike my Roomba vacuums, it doesn't back up and try again. I also had to be extremely careful with the G1T4-M1N1 around my deck stairs, since it continually tried to fall to its death like a robotic lemming.

According to Piaggio Fast Forward CEO and co-founder Greg Lynn, the company expects bot owners to watch out for major "cliff" drops like stairs. Gitamini and Gita have the ability to pause before jumping over large curb drops, he says, but I couldn't see that in action.

Devindra Hardawar for Engadget

Between its trouble with stairs and obstacle avoidance, I couldn't imagine ever using any Gita robot indoors without headaches. But I could see it being useful if I still lived in Brooklyn, where it could accompany my kids and I to the park while lugging their gear. (My aching back sure would appreciate that, anyway.) I could also see parking it outside a local grocery store and having it follow me home afterwards.

Even in that dream scenario, though, I'd worry about someone trying to steal the G1T4-M1N1. While you can lock it down from an app, which closes the storage compartment and turns off any following capability, you still have to remember to do that manually. According to Lynn, he's seen curious strangers accidentally enable the Gita bots' following feature more than he's encountered stories of potential theft. Still, the worry is there, especially since the G1T4-M1N1 and Gitamini weigh 28 pounds. It's not hard to imagine some intrepid thief driving up to a parked G1T4-M1N1 and throwing it in their car. (Unfortunately, there's no way to physically chain it like a bike, either.)

I'm not making any final judgements on the G1T4-M1N1, or its siblings, since I couldn't test it properly. But it's clearly not something meant for everyone — not at that price, and not with the navigation problems I encountered. But I could see a future where assistive robots like these could be helpful. We currently have robots delivering food on the streets of some cities, and I've seen them helping waiters move food around local restaurants. At some point, we'll see cheaper and more useful home bots beyond the Roomba and its ilk. And whatever comes next will likely owe a debt to the ground being tread by Gita's large wheels.

Dyson helped pioneer the cordless vacuum space, and now it’s testing the waters of the robot vacuum arena. The Dyson 360 Vis Nav has been available in other parts of the world for a little while, but it recently came stateside to the tune of $1,200. The company claims a 360-degree vision system, complete with cameras and LED lights, along with suction power akin to that of its stick vacuums sets the 360 Vis Nav apart from competitors. But even if that’s true, Dyson is undeniably late to the party. Robot vacuums have gotten a lot better in the past three to five years, and there are dozens on the market — including some that offer much more in the way of features for a similarly exorbitant price. That said, did Dyson pull an Apple and shake up a product category despite its late entry to the stage? I spent some time with the 360 Vis Nav to find out.

Setup and design

Unboxing and setting up the 360 Vis Nav is similar to any other robot vacuum. In addition to the device, the box contains a charging cable and the vacuum’s base, a rectangular, purple unit with two black-and-white checkered spots at either corner. While it’s refreshing to see a gadget that doesn’t wear the typical black or gray uniform, the 360 Vis Nav and its dock stick out like sore thumbs among the other items and furniture in my home. There’s no semblance of “blending into the decor” with this robo-vac.

Once docked for the first time, the Dyson 360 Vis Nav’s small onboard display prompts you to choose a preferred language and get connected via the mobile app. The screen doubles as a button, too, so you can press fully down on it to select options and move forward in the setup process. As I learned while using the machine, the display also shows helpful maintenance tips and accompanying graphics when you need to clean the robot’s sensors. The My Dyson app (on iOS and Android) provides all of the same information and more, and will guide you to connect the machine to your home’s Wi-Fi network, update the firmware if necessary and finish the prerequisites before you get to cleaning. Aside from waiting a few minutes for my review unit to update its firmware, the entire process took me only about 15 minutes.

To get the lay of the land, let’s revisit the few things Dyson claims set the 360 Vis Nav apart from other robot vacuums. The D-shaped design isn’t one of them, but it certainly helps the machine’s side-edge actuator when cleaning room corners. Dyson claims the actuator only opens up when cleaning corners like this, and it uses suction (rather than sweepers like other robo-vacs do) to capture debris from these hard-to-reach places. Along with that, the machine has a 360-degree vision system that helps it map our your home and clean around furniture and other objects, plus a sensor that detects the amount of dust present so it can kick up the suction when necessary and create a heat map of the dirtiest parts of your home. Those are the main differentiators, along with the claim that the 360 Vis Nav essentially has a similar level of suction power as one of Dyson’s cordless stick vacuums.

Performance

So how did all of that come together in practice? Pretty well, as it turns out. For the initial go-around, I had the Dyson 360 Vis Nav clean the main floor of my home, rather than map it out first. I did this mainly because I like chaos, but also because I wanted to see how the machine would navigate around coffee tables, couches and other furniture, plus small things like cat toys left in its path. Dyson states very clearly in the setup process that you should remove all small obstacles out of the way of the 360 Vis Nav before it cleans — I picked up a couple of reusable bags languishing from our last grocery run and the smaller cat toys, but I left some charging cables snaking on the ground because, let’s be real, most people aren’t going to clean before sending the robot they bought to clean for them out to do its job.

I was quickly struck by how many times the Dyson 360 Vis Nav came to a complete stop, “looked” around and kept moving during the first cleaning. It did a decent job avoiding large pieces of furniture like chairs and couches because — when it did get close to pieces like that — it seemed to sense it a few inches before hitting it, so it could then stop and redirect itself. It was most confused by a coffee table with a supporting beam that runs on the floor in between two legs. The 360 Vis Nav has adjustable wheels that allow it to “climb” on top of things like thicker rugs when it’s cleaning, and I think it confused this roughly 0.5-inch supporting beam for a piece of decor. The robo-vac tried so hard to climb over it; it was borderline concerning, but I was rooting for it the whole way. It actually did manage to climb over the beam, get back down onto the floor and keep cleaning, so kudos to the little guy for never giving up.

After that thrilling show, I left the 360 Vis Nav to its own devices. It proceeded to clean the main floor of my home for almost an hour, audibly notching up the suction when it detected a messier area. Notably, it seemed better at avoiding furniture than other robot vacuum cleaners I’ved tried. Yes, it did knock into a few things, but the number of times that happened was slim to none. The first run I did was in Auto mode, the default cleaning setting and one of four you have to choose from, and while the machine was a tad loud, it’s nothing that will drive pets or young children from a room. It’s loudest when the machine automatically kicks the motor into high gear upon detecting a high-dust area, which it did a few times in the highest trafficked areas of my kitchen and on top of a few area rugs. Like any other robot vacuum worth its salt, the 360 Vis Nav automatically returned to its dock to recharge when the battery got low.

I deemed that first cleaning job a success, and the case was the same when I did the first mapping run. The 360 Vis Nav is much quieter when it’s puttering around your home not sucking up dirt, but rather just using its vision system to create a map of your home and all its furniture and obstacles. The map it created of my main floor was accurate, and the Dyson app makes it simple to add boundaries and create rooms that you can label. Once you do this, you can create no-go zones and other restrictions like cleaning without the brush bar. And like most other vacuums with smart mapping, you can tell the 360 Vis Nav to only clean certain rooms with each job if you please.

Like most other robot vacuums, the companion app experience focuses on home maps, schedules and basic on-off controls. Dyson lets you choose from Auto, Quick, Quiet and Boost cleaning modes before you start a job, and if you have rooms designated on your map, you can customize cleaning modes for each room every time. For example, I could tell the machine to clean my kitchen in Auto mode and then the den in Quiet mode. The app presents a heat map of your home after every job that highlights the areas with the most dust, which is interesting to look at but didn’t provide me with a ton of groundbreaking information. I was not surprised that the dirtiest spots on my main floor were almost always next to the front door and my deck door.

Ultimately, the two things that impressed me the most about the Dyson 360 Vis Nav were its suction power and its obstacle avoidance capabilities. I live in a one-cat household, but if you saw how much cat hair my cordless stick vacuum sucks up every time I use it, you’d think I live with a few more felines. There’s always more fur in the bin after I clean the upper floor of my home, since it’s carpeted. After leaving the 360 Vis Nav on my upper floor to clean for almost an hour, I was surprised to see that the contents of its bin looked nearly identical to that of my cordless stick vac after cleaning up there. Even the best robot vacuum cleaners I’ve used in the past never sucked up this much pet fur in one go-around on carpeted floors — it’s impressive.

As for obstacle avoidance, I didn’t expect much from the 360 Vis Nav in this department because the instructions tell you, more than once, to clear your floors of any obstructions before cleaning. Most of the time, I had at least a few pairs of shoes, a couple of charging cables and other small objects on the floor while the machine was cleaning, but I never once got an alert that it was stuck. Some robot vacuums I’ve tested have sent me close to a half dozen alerts during a cleaning job, and I’d have to remove an object they were choking on before they could start up again. That never happened with Dyson’s machine.

The competition

With the Dyson 360 Vis Nav coming in at an eye-watering $1,200, the competition squarely within that price range is pretty slim. The $1,400 iRobot Roomba Combo j9+ is arguably its most direct competitor and it offers quite a bit more for that extra $200. (We’ve seen the Combo j9+ drop to as low as $1,000 in the past, so you might be able to pick it up for even less than Dyon’s machine.) As a “combo” device, the j9+ vacuums and mops without the help of a human (mostly) and it’s smart enough to know when it needs to switch from sucking up dirt to mopping floors using its built-in scrubbing pad and water reservoir. It also comes with a self-emptying base that can hold up to 60 days worth of dry debris and refill the water reservoir with clean supply. While auto-mopping is more of a nice-to-have than a requirement on a robot vacuum, it’s painful that Dyon’s $1,000+ machine doesn’t come with a clean base — a piece of hardware included in some models half of its price.

Wrap-up

When you understand the lay of the robot-vacuum landscape, the Dyson 360 Vis Nav seems almost quaint — and a bit confusing. Robot vacuums aren’t new, they’ve been around for a long time — long enough where you can get a pretty decent one for less than $300. For Dyson’s machine, spending $1,200 gets you a damn good robot vacuum with possibly the best suction power I’ve experienced on one, impressive obstacle avoidance, good home mapping and a clean app experience. But none of those features are unique to the 360 Vis Nav. Its suction power and obstacle avoidance may be superior, but is that enough to justify the cost? If you’re already willing to spend $1,000 or more on a robot vacuum, you have other options that will give you similar features and more, including mopping and self-emptying capabilities. That said, there are two groups of people who I’d recommend the 360 Vis Nav to: those who are willing to give up advanced features in exchange for the most suction power possible (and have cash to burn), and Dyson diehards.

If you’ve been craving some help with cleaning, this Memorial Day Roomba sale may be up your alley. Wellbots has the iRobot Roomba Combo J9+, Engadget’s favorite vacuum-mop combo, for $425 off. The machine not only cleans floors and carpets; its redesigned dock can automatically empty debris and refill it with mopping liquid. Use the exclusive code ENG425 to get iRobot’s top-of-the-line combo cleaner for $974.

Although that’s still a hefty price for a cleaning robot, you’re getting a premium product in return. The Roomba Combo J9+ has an upgraded motor and a four-stage cleaning system that takes multiple passes across your floors and carpets. This latest model, which arrived only last fall, has dual rubber brushes for optimal suction and pressurized scrubbing.

The robot has an updated Clean Base that automatically refills its water tank while looking more like a nice home appliance than a gadget’s charging dock. It automates as much of the setup process as possible, leaving you only to attach its mop pad and add water and cleaning solution. As for upkeep, you merely swap out its mop pads when needed and clean the Roomba’s bristles and bin.

Wellbots also offers the standard Roomba J9+ — the mop-less variant — for $325 off with the code ENG325. This model has all the vacuuming features from the more expensive Combo version. That includes stronger suction, multi-surface rubber brushes and a three-stage cleaning system. Usually $899, the coupon brings its price down to $574.

If you’re looking for a vacuum-mop combination robot for cheaper, the previous-generation Roomba Combo J7+ is $225 off with the code ENG225 (usually $999). This model skips out on some of the high-end features in the J9+ but still offers obstacle avoidance and a four-stage cleaning system. It can return to its base when it’s full and empty itself for up to 60 days before you have to empty its bin.

Follow @EngadgetDeals on Twitter and subscribe to the Engadget Deals newsletter for the latest tech deals and buying advice.

The Shark AI robot vacuum and mop combo unit has dropped to a record low of $230 at Amazon. This is a massive discount of 52 percent, which knocks nearly $250 off the price. It does look to be one of those deals that ends when the stock runs out, so you might want to act quickly instead of popping it on your wishlist.

This is a version of our very favorite robot vacuum. The only difference between this robovac and the one that took the top spot is that this edition doesn’t come with a self-emptying base. That model, however, is also currently on sale for $370.

If you can get over the absence of the base, the Shark AI vacuum is a fantastic cleaning partner. It ticks all of the boxes. The suction power is fantastic, the obstacle avoidance is on point and it works with voice assistants like Alexa. It’s also a great mop, cleaning hard surfaces with 100 scrubs per minute.

We also enjoyed using the mobile app, which easily creates accurate maps of your space. The app can also be used to operate the vacuum when not at home, so you can return from work to find a freshly-cleaned living area. You can even set “no-go” zones for the robot to avoid, so it won’t mop carpet. The best robot vacuums are the ones that do their job without requiring constant surveillance, and this lil robovac consistently manages to get itself out of tight spots to continue cleaning. The AI in the name refers to obstacle avoidance. 

There’s only one major downside, besides the lack of a base. This robovac/mop hybrid can get pretty loud when vacuuming, so you may want to schedule the book club meeting for somewhere else. Other than that, this is pretty much the perfect mid-range robot vacuum. 

Follow @EngadgetDeals on Twitter and subscribe to the Engadget Deals newsletter for the latest tech deals and buying advice.

When Boston Dynamics announced on Tuesday it was retiring the hydraulic version of Atlas, there were a few hints that the company wasn't done with humanoid robots entirely. Sure enough, one day later, Boston Dynamics has unveiled an all-electric model.

Atlas was originally envisioned as a search-and-rescue robot and Boston Dynamics claims the latest model is designed for real-world applications. It calls Atlas "the world’s most dynamic humanoid robot" and it certainly looks limber. 

A video shows Atlas lying prostrate and flipping its feet over to push itself up into a standing position. The robot then turns its head 180 degrees, followed by its torso. The rotations of the legs and the rest of the body are a little unnerving, but it's an impressive display of balance and flexibility. 

The electric Atlas appears sleeker than its predecessor, which looked slightly like a person wearing an exosuit. Rather than having a face with human features, Atlas' featureless head looks a bit like a ring light.

Boston Dynamics says parent company Hyundai's next generation of automotive manufacturing tech is the "perfect testing ground for new Atlas applications." It plans to show off what the robot can really do over the coming months and years, and to put Atlas through its paces with a small group of partners at first.

The company is looking into new gripper systems to make sure Atlas is suitable for a range of commercial needs while building on the previous parkour-capable model's ability to lift and move a variety of heavy and irregular objects. It claims that the new Atlas will be stronger than before and it's confident that it can commercialize a humanoid robot.

"Atlas may resemble a human form factor, but we are equipping the robot to move in the most efficient way possible to complete a task, rather than being constrained by a human range of motion. Atlas will move in ways that exceed human capabilities," Boston Dynamics wrote in a blog post. "Combining decades of practical experience with first principles thinking, we are confident in our ability to deliver a robot uniquely capable of tackling dull, dirty and dangerous tasks in real applications."

Boston Dynamics is hardly the only company working on a humanoid robot. Tesla, of course, has one in the pipeline, while Menteebot, which can be controlled using natural-language voice commands emerged just this morning. 

However, Boston Dynamics has been working on robots with this form factor for well over a decade, far longer than most. As things stand, it may be best positioned to get a humanoid robot into workplaces and even homes. Before that though, you might expect to see some videos in which the electric Atlas shows off some slick dance moves.

Over the past two-and-a-half years, a group of students from ETH Zurich have been developing a robot with three spindly legs that was designed to be able to hop like an insect in microgravity. That's right — the curious little machine was built for space, specifically for the exploration of small celestial bodies like asteroids and moons. SpaceHopper, as the robot is called, could thus provide us more information to advance our understanding of life's origin, of the origin of water on our planet and of asteroids as potential providers of valuable resources. 

It has no preferred orientation, so it can go in any direction, and it has nine motors that give it the capability to jump long distances in low-gravity environments. The robot can even self-right after landing, ensuring the safety of any scientific payload it may carry. Since SpaceHopper was made for use on asteroids and moons, which have very little gravity compared to Earth, it has to be tested under conditions similar to those environments first. To see if it will actually work as intended, the students and the European Space Agency have recently taken the robot on a parabolic flight that creates a zero gravity environment when the aircraft freefalls. Apparently, they had no idea if SpaceHopper would be able to move as they intended in zero gravity scenarios and seeing that it actually worked was a "massive weight off [their] shoulders."

You can watch SpaceHopper flail about in the test flight below:

The whole world is ragging on a barely functional $700 AI pin at the moment, but what if similar tech was squeezed into a gigantic robot that lives in your home? That’s a worst case scenario for the recently-introduced Menteebot, a human-sized robot that’s stuffed to the brim with AI-adjacent bells and whistles.

It’s being advertised as the “personalized AI-based robot you can mentor.” It can run, walk sideways and even turn, all “with the same balance and control as a human.” Manufacturer Mentee Robotics also says it’ll adjust its gait when lifting heavy objects. It should be able to lift these heavy objects with ease due to the fact that it’s, well, absolutely gigantic. Many of the models also have no head, which certainly doesn’t recall any old-time myths about a scary demon on a horse.

Now, we’ve had humanoid robots for a while. There was Honda’s Asimo, which has been sadly discontinued, and the army of nightmare creatures that Boston Dynamics is busy cooking up. Agility Robotics has been building out its robot assistant Digit and Elon Musk, who never makes false promises ever swear to God, says that Tesla is working on a humanoid robot called Optimus.

There’s one major difference between the aforementioned bots and Mentee’s creation. Menteebot is stuffed with AI algorithms, natural language processing models and software that unlocks “advanced training techniques.” The company says that this means the robot is “not bound to a limited set of commands” and that it can even hold conversations with humans. As a matter of fact, users issue commands to the robot via natural language. 

It’s a robot with two arms and two legs that can, in theory, do many of the same things we do. The company says that we can train it to do these things. This seems to sort of work like another controversial piece of AI tech, the Rabbit R1. To teach Menteebot a new task, you run a simulated version of the bot through a digital version of the task. The software completes the task over and over until it figures it out. Then the robot should be able to complete the task in the real world. This seems like an extremely lofty promise, but we’ll wait to see the final result. Here’s hoping it doesn't hallucinate and do whatever the heck it wants like other bits of AI tech. 

Menteebot does look quite agile. There are tons of videos of the robot being put through its paces. It can run and the arms and hands “present a full range of motion and enough accuracy to perform delicate tasks.” To that end, there’s a video of it gently handing a piece of dinnerware to a person.

While it’s highly unlikely this robot will live up to the initial promotional materials when it arrives in 2025 (just look at the initial promises Humane made for the AI pin), it still seems pretty darned cool. There’s no announced price, but it’s certainly going to be a whole lot more than the aforementioned $700 pin. This is an agile humanoid robot that weighs over 150 pounds. 

Menteebot will be available in two flavors. There’s the residential bot, which is forced to do household chores, and the commercial bot, which is forced to do manual labor. No matter which you choose, for heaven’s sake, be extra nice to the thing. Don’t boss it around. Let it sit at the dinner table. Keep it away from the vast majority of sci-fi. It can watch Star Trek: The Next Generation, maybe, as Data seems like a decent enough role model.

Nearly 11 years after it first showed off its current humanoid robot, Boston Dynamics is retiring Atlas. The DARPA-funded robot was designed with search and rescue missions in mind, with the idea that it would be able to enter areas that were unsafe for humans to carry out a range of tasks. However, Atlas became a bit of a star thanks to videos showing off its slick dance moves and impressive feats of strength, agility and balance. Fittingly, Atlas is trotting off into the sunset with one final YouTube video.

"For almost a decade, Atlas has sparked our imagination, inspired the next generations of roboticists, and leapt over technical barriers in the field," the YouTube description reads. "Now it’s time for our hydraulic Atlas robot to kick back and relax."

Boston Dynamics' farewell to Atlas doesn't just show some of the cool things the robot can do. It's a bit of a blooper reel as well. Along with hurling a toolbag and leaping between platforms, Atlas slips, trips and falls a bunch of times in the clip — oddly enough, that makes it seem more human.

Boston Dynamics of course has more commercially successful robots in its lineup, including Spot. It's likely not the end of the line for the company's humanoid robots entirely, though. "Take a look back at everything we’ve accomplished with the Atlas platform to date," reads the description on the farewell video. Those last two words suggest Boston Dynamics isn't quite done with that side of robotics yet.

Engadget has contacted the company for details about its future humanoid robot development plans. For now, it seems Atlas could be looking for a Wednesday afternoon dance partner at a robot retirement home.

Robot vacuums are handy little devices that can help folks save a ton of time and energy. However, some of the more well-known options are often a bit pricey, especially when a mopping function comes into the mix. As it happens, iRobot has revealed a relatively budget-friendly 2-in-1 robot vacuum and mop. It says the $275 Roomba Combo Essential actually outperforms the Roomba 600 Series thanks to 20 times more suction power, and the addition of a mop and smart navigation.

According to iRobot, this model offers 25 percent better performance at picking up dirt from hard floors than the Roomba 600 Series. It's also said to have a longer battery life at up to 120 minutes, the ability to clean in neat rows, customizable suction and liquid settings, Clean Map reports and intelligent settings such as suggested cleaning schedules.

Although you can set up cleanings in advance, you can start one at any time with an Alexa, Siri or Google Assistant voice command. Alternatively, you can press the Clean button or tap a button in the iRobot Home app to put the device to work right away.

iRobot

The device has a four-stage cleaning system that includes adjustable suction and liquid settings, a v-shaped multi-surface brush, an edge-sweeping brush and a pump-fed microfiber mop pad. For vacuum-only operation (i.e if you're looking to remove dirt from a rug), you'll need to remove the mop pad first. That adds a little bit of friction to using this model. There's another trade-off in that this isn't a self-emptying Roomba — you'll need to empty out the dirt storage bin manually more often.

Still, this seems like a solid Roomba at an eye-catching price. It's available in Europe, the Middle East and Africa now, and iRobot will start selling it in the US on April 7 and Canada on April 12. The Roomba Combo Essential will reach Asia Pacific markets later this month. Folks in the US can trade in a Roomba 600 Series for a $50 credit toward a Roomba Combo Essential

In addition, iRobot is rolling out a model called the Roomba Vac Essential in North America. It has the same smart functions and other similar features as the Combo Essential, but there's no mop. That robot vacuum will cost $250 and it'll land in the US on April 7 and Canada on April 12.

Meanwhile, iRobot says it has reached a new milestone. Since debuting the Roomba in 2002, the company has sold more than 50 million robots.