Hundreds of robots zip back and forth across the floor of a colossal robotic warehouse, grabbing items and delivering them to human workers for packing and shipping. Such warehouses are increasingly becoming part of the supply chain in many industries, from e-commerce to automotive production.

However, getting 800 robots to and from their destinations efficiently while keeping them from crashing into each other is no easy task. It is such a complex problem that even the best path-finding algorithms struggle to keep up with the breakneck pace of e-commerce or manufacturing. 

In a sense, these robots are like cars trying to navigate a crowded city center. So, a group of MIT researchers who use AI to mitigate traffic congestion applied ideas from that domain to tackle this problem.

They built a deep-learning model that encodes important information about the warehouse, including the robots, planned paths, tasks, and obstacles, and uses it to predict the best areas of the warehouse to decongest to improve overall efficiency.

Their technique divides the warehouse robots into groups, so these smaller groups of robots can be decongested faster with traditional algorithms used to coordinate robots. In the end, their method decongests the robots nearly four times faster than a strong random search method.

In addition to streamlining warehouse operations, this deep learning approach could be used in other complex planning tasks, like computer chip design or pipe routing in large buildings.

“We devised a new neural network architecture that is actually suitable for real-time operations at the scale and complexity of these warehouses. It can encode hundreds of robots in terms of their trajectories, origins, destinations, and relationships with other robots, and it can do this in an efficient manner that reuses computation across groups of robots,” says Cathy Wu, the Gilbert W. Winslow Career Development Assistant Professor in Civil and Environmental Engineering (CEE), and a member of a member of the Laboratory for Information and Decision Systems (LIDS) and the Institute for Data, Systems, and Society (IDSS).

Wu, senior author of a paper on this technique, is joined by lead author Zhongxia Yan, a graduate student in electrical engineering and computer science. The work will be presented at the International Conference on Learning Representations.

Robotic Tetris

From a bird’s eye view, the floor of a robotic e-commerce warehouse looks a bit like a fast-paced game of “Tetris.”

When a customer order comes in, a robot travels to an area of the warehouse, grabs the shelf that holds the requested item, and delivers it to a human operator who picks and packs the item. Hundreds of robots do this simultaneously, and if two robots’ paths conflict as they cross the massive warehouse, they might crash.

Traditional search-based algorithms avoid potential crashes by keeping one robot on its course and replanning a trajectory for the other. But with so many robots and potential collisions, the problem quickly grows exponentially.

“Because the warehouse is operating online, the robots are replanned about every 100 milliseconds. That means that every second, a robot is replanned 10 times. So, these operations need to be very fast,” Wu says.

Because time is so critical during replanning, the MIT researchers use machine learning to focus the replanning on the most actionable areas of congestion — where there exists the most potential to reduce the total travel time of robots.

Wu and Yan built a neural network architecture that considers smaller groups of robots at the same time. For instance, in a warehouse with 800 robots, the network might cut the warehouse floor into smaller groups that contain 40 robots each.

Then, it predicts which group has the most potential to improve the overall solution if a search-based solver were used to coordinate trajectories of robots in that group.

An iterative process, the overall algorithm picks the most promising robot group with the neural network, decongests the group with the search-based solver, then picks the next most promising group with the neural network, and so on.

Considering relationships

The neural network can reason about groups of robots efficiently because it captures complicated relationships that exist between individual robots. For example, even though one robot may be far away from another initially, their paths could still cross during their trips.

The technique also streamlines computation by encoding constraints only once, rather than repeating the process for each subproblem. For instance, in a warehouse with 800 robots, decongesting a group of 40 robots requires holding the other 760 robots as constraints. Other approaches require reasoning about all 800 robots once per group in each iteration.

Instead, the researchers’ approach only requires reasoning about the 800 robots once across all groups in each iteration.

“The warehouse is one big setting, so a lot of these robot groups will have some shared aspects of the larger problem. We designed our architecture to make use of this common information,” she adds.

They tested their technique in several simulated environments, including some set up like warehouses, some with random obstacles, and even maze-like settings that emulate building interiors.

By identifying more effective groups to decongest, their learning-based approach decongests the warehouse up to four times faster than strong, non-learning-based approaches. Even when they factored in the additional computational overhead of running the neural network, their approach still solved the problem 3.5 times faster.

In the future, the researchers want to derive simple, rule-based insights from their neural model, since the decisions of the neural network can be opaque and difficult to interpret. Simpler, rule-based methods could also be easier to implement and maintain in actual robotic warehouse settings.

“This approach is based on a novel architecture where convolution and attention mechanisms interact effectively and efficiently. Impressively, this leads to being able to take into account the spatiotemporal component of the constructed paths without the need of problem-specific feature engineering. The results are outstanding: Not only is it possible to improve on state-of-the-art large neighborhood search methods in terms of quality of the solution and speed, but the model generalizes to unseen cases wonderfully,” says Andrea Lodi, the Andrew H. and Ann R. Tisch Professor at Cornell Tech, and who was not involved with this research.

This work was supported by Amazon and the MIT Amazon Science Hub.

The Reshoring of Chipmaking

Under Donald Trump and Joe Biden Alike, the US has been determined to "reshore" chipmaking

The United States has been on a mission to bring back chip manufacturing from overseas, a move touted as a key component in the country’s quest for technological supremacy. This effort, dubbed "reshoring," has seen billions of dollars in investments pouring into the sector. In Ohio, a vast Intel site is now reaping the benefits of this strategy, but the question remains: is this effort worth the cost?

The Rise of Reshoring

In 2020, the US government signed the CHIPS Act, a bill aimed at supporting the domestic semiconductor industry. The legislation provided significant funding for chip research and development, as well as incentives for companies to establish domestic manufacturing facilities. The strategy was simple: by encouraging companies to "reshore" their chip production, the US could reduce its reliance on foreign suppliers, improve national security, and create high-paying jobs.

Intel’s Ohio Mega-Site

In the heart of Ohio, a massive Intel facility is undergoing a major expansion. The site, which will ultimately span over 2,000 acres, will become one of the largest and most advanced chip manufacturing complexes in the world. With a price tag of around $20 billion, the project is a testament to the US government’s commitment to reshoring chip production.

The Infrastructure of Reshoring

Behind the scenes, the US government has been investing heavily in the infrastructure required to support chip manufacturing. The CHIPS Act has allocated billions of dollars to develop new facilities, upgrade existing ones, and establish research centers. The Department of Defense has also played a key role, providing funding for chip research and development.

But at What Cost?

As the Ohio Intel site takes shape, concerns are growing about the long-term sustainability of this massive investment. The site’s construction has already resulted in significant environmental impact, with local residents expressing worries about air and water pollution. Additionally, the billions of dollars being poured into the project could be seen as a distraction from other pressing issues, such as the country’s struggling education system and crumbling infrastructure.

Conclusion

As the US continues its efforts to reshore chipmaking, it’s essential to weigh the benefits against the costs. While the promise of high-paying jobs and improved national security is tantalizing, the environmental and financial implications of this project cannot be ignored. The US government must carefully balance its goals with the well-being of its citizens and the planet.

FAQs

Q: Why is the US trying to reshore chipmaking?
A: The US is trying to reshore chipmaking to reduce its reliance on foreign suppliers, improve national security, and create high-paying jobs.

Q: How much is the Ohio Intel site costing?
A: The Ohio Intel site is estimated to cost around $20 billion.

Q: What is the impact of the site’s construction on the local environment?
A: The site’s construction has raised concerns about air and water pollution among local residents.

Q: Is the government’s investment in the chip industry a good use of taxpayer dollars?
A: The effectiveness of this investment is a topic of debate, with some arguing that it distracts from more pressing issues, while others see it as a vital component of the country’s technological future.

As it trundles around an ancient lakebed on Mars, NASA’s Perseverance rover is assembling a one-of-a-kind rock collection. The car-sized explorer is methodically drilling into the Red Planet’s surface and pulling out cores of bedrock that it’s storing in sturdy titanium tubes. Scientists hope to one day return the tubes to Earth and analyze their contents for traces of embedded microbial life.

Since it touched down on the surface of Mars in 2021, the rover has filled 20 of its 43 tubes with cores of bedrock. Now, MIT geologists have remotely determined a crucial property of the rocks collected to date, which will help scientists answer key questions about the planet’s past.

Here, the Perseverance drills into the surface of Mars.

Image: NASA/JPL-Caltech/ASU/MSSS

In a study appearing today in the journal Earth and Space Science, an MIT team reports that they have determined the original orientation of most bedrock samples collected by the rover to date. By using the rover’s own engineering data, such as the positioning of the vehicle and its drill, the scientists could estimate the orientation of each sample of bedrock before it was drilled out from the Martian ground.

The results represent the first time scientists have oriented samples of bedrock on another planet. The team’s method can be applied to future samples that the rover collects as it expands its exploration outside the ancient basin. Piecing together the orientations of multiple rocks at various locations can then give scientists clues to the conditions on Mars in which the rocks originally formed.

“There are so many science questions that rely on being able to know the orientation of the samples we’re bringing back from Mars,” says study author Elias Mansbach, a graduate student in MIT’s Department of Earth, Atmospheric and Planetary Sciences.

“The orientation of rocks can tell you something about any magnetic field that may have existed on the planet,” adds Benjamin Weiss, professor of planetary sciences at MIT. “You can also study how water and lava flowed on the planet, the direction of the ancient wind, and tectonic processes, like what was uplifted and what sunk. So it’s a dream to be able to orient bedrock on another planet, because it’s going to open up so many scientific investigations.”

Weiss and Mansbach’s co-authors are Tanja Bosak and Jennifer Fentress at MIT, along with collaborators at multiple institutions including the Jet Propulsion Laboratory at Caltech.

Profound shift

The Perseverance rover, nicknamed “Percy,” is exploring the floor of Jezero Crater, a large impact crater layered with igneous rocks, which may have been deposited from past volcanic eruptions, as well as sedimentary rocks that likely formed from long-dried-out rivers that fed into the basin.

An image mosaic, taken by the rover’s Mastcam-Z, shows a portion of the Jezero crater floor, where Perseverance drilled cores of Martian bedrock.

Image: NASA/JPL-Caltech/ASU/MSSS

An image mosaic, taken by the rover’s Mastcam-Z, shows a region of the Jezero delta, where Perseverance drilled and collected cores.

Image: NASA/JPL-Caltech/ASU/MSSS

“Mars was once warm and wet, and there’s a possibility there was life there at one time,” Weiss says. “It’s now cold and dry, and something profound must have happened on the planet.”

Many scientists, including Weiss, suspect that Mars, like Earth, once harbored a magnetic field that shielded the planet from the sun’s solar wind. Conditions then may have been favorable for water and life, at least for a time.

“Once that magnetic field went away, the sun’s solar wind — this plasma that boils off the sun and moves faster than the speed of sound — just slammed into Mars’ atmosphere and may have removed it over billions of years,” Weiss says. “We want to know what happened, and why.”

The rocks beneath the Martian surface likely hold a record of the planet’s ancient magnetic field. When rocks first form on a planet’s surface, the direction of their magnetic minerals is set by the surrounding magnetic field. The orientation of rocks can thus help to retrace the direction and intensity of the planet’s magnetic field and how it changed over time.

Since the Perseverance rover was collecting samples of bedrock, along with surface soil and air, as part of its exploratory mission, Weiss, who is a member of the rover’s science team, and Mansbach looked for ways to determine the original orientation of the rover’s bedrock samples as a first step toward reconstructing Mars’ magnetic history.

“It was an amazing opportunity, but initially there was no mission requirement to orient bedrock,” Mansbach notes.

Roll with it

Over several months, Mansbach and Weiss met with NASA engineers to hash out a plan for how to estimate the original orientation of each sample of bedrock before it was drilled out of the ground. The problem was a bit like predicting what direction a small circle of sheetcake is pointing, before twisting a round cookie cutter in to pull out a piece. Similarly, to sample bedrock, Perseverance corkscrews a tube-shaped drill into the ground at a perpendicular angle, then pulls the drill directly back out, along with any rock that it penetrates.

To estimate the orientation of the rock before it was drilled out of the ground, the team realized they need to measure three angles, the hade, azimuth, and roll, which are similar to the pitch, yaw, and roll of a boat. The hade is essentially the tilt of the sample, while the azimuth is the absolute direction the sample is pointing relative to true north. The roll refers to how much a sample must turn before returning to its original position.

In talking with engineers at NASA, the MIT geologists found that the three angles they required were related to measurements that the rover takes on its own in the course of its normal operations. They realized that to estimate a sample’s hade and azimuth they could use the rover’s measurements of the drill’s orientation, as they could assume the tilt of the drill is parallel to any sample that it extracts.

To estimate a sample’s roll, the team took advantage of one of the rover’s onboard cameras, which snaps an image of the surface where the drill is about to sample. They reasoned that they could use any distinguishing features on the surface image to determine how much the sample would have to turn in order to return to its original orientation.

In cases where the surface bore no distinguishing features, the team used the rover’s onboard laser to make a mark in the rock, in the shape of the letter “L,” before drilling out a sample — a move that was jokingly referred to at the time as the first graffiti on another planet.

By combining all the rover’s positioning, orienting, and imaging data, the team estimated the original orientations of all 20 of the Martian bedrock samples collected so far, with a precision that is comparable to orienting rocks on Earth.

“We know the orientations to within 2.7 degrees uncertainty, which is better than what we can do with rocks in the Earth,” Mansbach says. “We’re working with engineers now to automate this orienting process so that it can be done with other samples in the future.”

“The next phase will be the most exciting,” Weiss says. “The rover will drive outside the crater to get the oldest known rocks on Mars, and it’s an incredible opportunity to be able to orient these rocks, and hopefully uncover a lot of these ancient processes.”

This research was supported, in part, by NASA and the Mars 2020 Participating Scientist program.

Why Spotify’s AI Music Problem Is a Bigger Issue Than You Think

A screenshot of the fake Annie album, taken October 12th.

Some kind of AI slop had been uploaded to HEALTH’s artist page on Spotify, one of three fake albums that would appear under their name that weekend. The band’s X account made some jokes about it, the albums were eventually removed, and I went back to minding my own business. Then, the next weekend, I saw a new Annie album had dropped.

The Problem

That album was more plausible — Annie had just released a new single, "The Sky Is Blue" — but when I clicked in, I couldn’t find it on the list of the song titles. Confused, I played the album and heard birdsong and a vaguely New Age-y instrumental. That… did not sound like Annie.

"That was upsetting to me, because if you have ears, you can definitely hear it’s not our music," said Marcos Mena, Standards’ lead songwriter and guitarist. "It’s definitely a bummer because we did have a new album come out this year, and I feel like it’s detracting from that."

What’s Going On

To understand how this works, you need a sense of the mechanics. Streaming platforms like Spotify don’t work like your Facebook page — Mena and other artists aren’t logging in and adding albums to their accounts directly. Instead, they go through a distributor that handles licensing, metadata, and royalty payments. Distributors send songs and metadata in bulk to the streaming services. The metadata part is important; it includes things such as the song title and artist name but also other information, such as the songwriter, record label, and so on. This is crucial for artists (and others) to get paid.

Why It’s a Problem

The way it should have worked, getting the plumbing right, is that all those albums should have been flagged as new artists, and then it wouldn’t matter. But allowing too many junk bands through creates problems with the streaming services.

As for the distributors, the thing to keep an eye on is UMG’s lawsuit. A pretrial conference is scheduled for January. The outcome of the suit could potentially change how distributors filter the music people try to upload through their platforms — because if lawsuits are more expensive than content moderation, there’s likely to be more content moderation. That could improve things for Spotify, which is downstream of them.

Conclusion

AI music poses the same threat to Spotify, McDonald says. He points out that I had been waiting for the Annie album, excited for it, even. And then instead, I got duped into garbage. "There’s all these mechanisms around assuming this stuff is correct," he says. But right now, those mechanisms are broken — and people who truly care, like artists themselves, don’t have their hands on the controls.

FAQs

Q: Why is this a problem?
A: This is a problem because it’s affecting artists and their music. Fake albums are being uploaded to their accounts, which can confuse fans and detract from their actual music.

Q: Who is responsible for this?
A: The distributors and Spotify are responsible for this. They need to do a better job of filtering out fake albums and ensuring that only legitimate music is uploaded to the platform.

Q: What can be done to fix this?
A: UMG’s lawsuit could potentially change how distributors filter music. Additionally, Spotify and distributors need to do a better job of content moderation to prevent fake albums from being uploaded in the first place.

Q: Is this a new problem?
A: No, this is not a new problem. Fraudsters have been using AI to create fake music for years. However, the rise of AI-generated music has made it easier for fraudsters to create convincing fake albums.