KeysQuestions4Vehicle1<\/p>\n
The PACMAN controller running on a real tokamak isn’t a chatbot bolted onto a press release \u2014 it’s a machine learning system making 50 hardware decisions per second on live fusion plasma, and that gap between human reaction time and plasma instability timescales is exactly why this matters.<\/p>\n
The system called a magnetic tear 200 milliseconds before it formed, which sounds like a modest number until you realize the instability itself grows in thousandths of a second and a human operator responds in full seconds.<\/p>\n
If fusion ever becomes a real power source, the control architecture being tested right now in San Diego is the invisible layer that determines whether it works at all \u2014 and this is one of the first times a model has been trusted with the actual heating hardware, not just the simulation.<\/p>\n
Ask The Box \u2192<\/p>\n
Considering the Accord?<\/p>\n
Accord vs rival?Best trimReal MPG<\/p>\n
AI-generated summary from reviews, comparisons and articles by AutoNotion<\/p>\n
AI keeps getting announced into jobs it doesn\u2019t really do. Somebody\u2019s usually bolted a chatbot onto a product nobody wanted to have a conversation with, and the press release says it\u2019s a breakthrough. So when a national lab says it handed the heating systems on a working fusion machine over to a machine learning model, that\u2019s a different kind of claim, and I want to know exactly what that model was allowed to touch.<\/p>\n
It\u2019s quite a lot, as it turns out. Researchers at the Princeton Plasma Physics Laboratory<\/a> and Princeton University built a control framework called PACMAN, short for Prediction And Control using MAchiNe learning, and ran it on the DIII-D National Fusion Facility<\/a> tokamak in San Diego. The lab says the whole loop comes around about every 20 milliseconds, roughly 50 times a second, over and over for the length of a shot.<\/p>\n Let\u2019s get the dates out of the way, because they matter here. PPPL put its announcement out on September 2. The paper behind it<\/a> went into the journal Nuclear Fusion on July 2, and a preprint has been sitting on arXiv<\/a> since November 2025. The experiments are older than all three, and the lab doesn\u2019t publish the dates of individual shots. The paper names one of them, shot 204975, and there\u2019s no calendar date attached to it. It\u2019s a July result with a September press release.<\/p>\n The loop runs about 50 times a second<\/p>\n A tokamak holds plasma, an electrically charged gas, inside a magnetic field, and plasma doesn\u2019t sit still. Disturbances grow in thousandths of a second. A focused human operator reacts in seconds, according to Andy Rothstein, one of the two lead authors, and the plasma isn\u2019t going to wait around that long.<\/p>\n The simulation codes that describe plasma properly take days or months to run, so they\u2019re no use at all during an experiment that lasts a few seconds. Machine learning models are fast and approximate instead of slow and exact. That tradeoff is the whole reason they\u2019re in the control room.<\/p>\n PACMAN does four things in order and then does them again. First it grabs every real-time measurement the machine produces, temperatures, densities, magnetic signals, checks them for errors and packs them into a single bundle. Then the AI models take the inputs they want and put out a call on where the plasma is headed. Controllers turn that call into hardware requests, like telling a heating beam to go up. Last comes the stage that settles arguments between controllers, clamps every command against the machine\u2019s hardware limits and sends it out. If that sounds like a lot of plumbing for an AI story, that\u2019s fair, and the plumbing is what nobody had built before.<\/p>\n That last stage doesn\u2019t get talked about much, and it\u2019s arguably the most important piece of the whole design if you\u2019re the one signing off on letting a model near the hardware.<\/p>\n So what\u2019s a tearing mode?<\/p>\n The magnetic field in a tokamak is supposed to hold the plasma in smooth nested layers, a bit like the rings of an onion. A tearing mode is what happens when those layers tear and reconnect into islands. Heat leaks across the islands, confinement drops, and a bad one can end the shot early.<\/p>\n Autonoci\u00f3n Newsletter<\/p>\n No spam. Unsubscribe anytime. Privacy policy (opens in new window)<\/a><\/p>\n So why not just catch it the old way? Because a conventional controller can only spot a tearing mode once it\u2019s started, and suppressing one after the fact costs you performance. In one of these experiments, the model flagged the mode roughly 200 milliseconds before it showed up.<\/p>\n Then the controller does something physical about it. When the predicted probability crosses 10 percent, the first gyrotron stops driving current broadly off-axis and starts tracking the q=2 surface. That\u2019s the layer of the magnetic field where these islands like to form, and putting current drive right there makes it harder for one to grow. Past 20 percent, a second gyrotron switches to the same target.<\/p>\n Aiming heat at a problem before the problem exists is a pretty different way to run a machine than waiting for the alarm to go off.<\/p>\n PACMAN steered all six microwave heaters at once<\/p>\n A gyrotron is a tube that produces a very high power microwave beam, and DIII-D has six of them. The beams don\u2019t go in fixed. Each one bounces off a steerable mirror, so move the mirror and you\u2019ve moved where the heat and the driven current land inside the plasma. The tubes are a serious business in their own right, and we\u2019ve written about the Japanese supplier selling them into DIII-D<\/a> that skipped building a reactor of its own entirely.<\/p>\n PACMAN steered all six at the same time, swinging the mirrors over and trimming the power on each one while the shot ran, to hit goals the researchers had set beforehand. \u201cThere was no algorithm to find that optimal solution before,\u201d said Hiro Farre Kaga, the other lead author.<\/p>\n A different experiment used beams, microwaves and gas together to hold 33-point profiles of plasma density and rotation on target. That one ran on a 20 millisecond cycle, about 4 milliseconds to squash the plasma state down into a small mathematical model and up to 10 more for the optimizer working out what to do about it.<\/p>\n ACTIVE<\/p>\n Control loop<\/p>\n 20 ms<\/p>\n Typical time for the full framework to read the machine and act on it. An operator reacts in seconds.<\/p>\n Tearing mode warning<\/p>\n 200 ms<\/p>\n How far ahead the model called the instability in one experiment, before it formed.<\/p>\n Microwave heaters<\/p>\n 6<\/p>\n Gyrotrons steered together, mirrors retargeted and power adjusted mid-shot.<\/p>\n Experiments<\/p>\n 5<\/p>\n Separate runs described in the paper, with cycle times from 2 to 50 milliseconds.<\/p>\n Even so, it isn\u2019t driving everything<\/p>\n The authors are blunt about where PACMAN doesn\u2019t belong. The framework, they write, \u201cwould be inappropriate\u201d for vertical displacement events, where the plasma slips vertically and disrupts in under a millisecond. A loop that comes around every 20 milliseconds turns up to find the shot\u2019s already over.<\/p>\n There are other guardrails. If two controllers ask for conflicting settings on the same hardware, the output stage sends nothing at all and tells the operators. That\u2019s the safe answer, and the authors are honest that doing nothing can occasionally be more dangerous than doing something imperfect. If a diagnostic fails, the model that needed it doesn\u2019t run that cycle. Beam power, microwave power, gas valve voltages, mirror angles and coil currents all get clamped to fixed minimums and maximums before any of it reaches the tokamak. Humans still set the targets before each shot, and physicists retune the controllers between them.<\/p>\n And no, this wasn\u2019t one heroic shot with everything switched on together. Each experiment was its own run at its own speed, from a 2 millisecond cycle for the model predicting edge energy bursts up to 50 milliseconds for the reinforcement learning controller. The paper describes five of them and mentions four more it doesn\u2019t detail.<\/p>\n So does any of it travel?<\/p>\n That\u2019s the actual pitch. Building the framework and getting the first model into it took months of work. The second one was in within a couple of days, according to Rothstein, and he says it went in far more cleanly than the first.<\/p>\n Egemen Kolemen, the Princeton professor who supervised the work, argues that the modular setup means a model can be dropped in, pulled back out or stacked alongside others while everything else stays where it is, and that the same framework could move to tokamaks of other sizes and shapes, including machines nobody has designed yet. I\u2019d file that under reasonable goal rather than delivered result. Every tokamak you\u2019d want to put this on has its own diagnostics, its own actuators and its own control system, and the machine being assembled in southern France<\/a> has spent three years fixing a heat shield.<\/p>\n You\u2019re not getting electricity out of any of this either, and DIII-D isn\u2019t built to produce any. What PACMAN did is steer hardware faster than a person can, inside limits a person wrote down first.<\/p>\n PPPL says the framework is ready to take new models as DIII-D works toward its next control targets, and the team wants to run the models and controllers in parallel to speed the loop up further. The paper went up in Nuclear Fusion on July 2 with Rothstein and Farre Kaga as co-lead authors, and the lab announced the results on September 2.<\/p>\n","protected":false},"excerpt":{"rendered":"KeysQuestions4Vehicle1 The PACMAN controller running on a real tokamak isn’t a chatbot bolted onto a press release \u2014…\n","protected":false},"author":2,"featured_media":626013,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[24],"tags":[61,60,248,82],"class_list":["post-626012","post","type-post","status-publish","format-standard","has-post-thumbnail","category-physics","tag-ie","tag-ireland","tag-physics","tag-science"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/posts\/626012","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/comments?post=626012"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/posts\/626012\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/media\/626013"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/media?parent=626012"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/categories?post=626012"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/tags?post=626012"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}