Globe Magazine

Why an MIT robot is collecting poop from our sewers

Luigi’s job is to gather what we flush so that researchers can learn about our habits and our health.

Mariana Matus (left) and Newsha Ghaeli with sewer-bot Luigi in the MIT lab where Matus analyzes its samples.
webb chappell
Mariana Matus (left) and Newsha Ghaeli with sewer-bot Luigi in the MIT lab where Matus analyzes its samples.

NEWSHA GHAELI didn’t come to MIT to study sewage, but on a chilly November morning, she’s staring into an uncovered manhole near campus, watching water churn below her colleague Luigi as Luigi is raised from the sewer by another MIT researcher, Shinkyu Park. Ghaeli and Park occasionally wince as the scent of rotten eggs penetrates their protective masks. “This is worse than anywhere we’ve ever sampled,” Ghaeli says. Luigi doesn’t notice — Luigi is a robot.

On-board sensors tell Luigi — a plastic-encased tubular tangle of wires, batteries, and electronics — when it’s below the waterline, the water temperature, whether the water is flowing, and how much is in the 250-milliliter bottle (about 8.5 fluid ounces) it carries. The bottle is sterile and empty when Luigi is lowered on 60-pound fishing wire the 10 feet or so into the sewage stream. But when Luigi returns, the bottle is filled with a gray-brown liquid, a broth of flushes and drains from neighboring homes and businesses. And it is covered with clumps of an unwanted substance: toilet paper.

Ghaeli grimaces. “There’s a lot of toilet paper here,” she says as she gingerly wipes away the wet muck from the bottom of the robot. Toilet paper is enough of a problem that Ghaeli and Park’s research team, dubbed “Underworlds,” built a fake flowing toilet to help determine which kind of mesh cover would best keep it from gumming up Luigi’s works, but still let in tiny bits of solid fecal matter for useful data. “We’ve designed multiple versions,” says Ghaeli. “The one we have currently works nine times out of ten.”


Ghaeli, an architect, came to MIT to do research at Carlo Ratti’s Senseable City Lab, which uses technology to explore urban environments in new ways. Past projects have measured levels of greenery in cities, rethought traffic flow, and explored the supermarkets of the future, all more glamorous than Underworlds’ sewers. But sewers offer a look at the secret life of cities. They could reveal neighborhoods with high rates of infectious diseases or even obesity, levels of illegal drug use, and pollutants. Today’s odiferous work is part of a semester-long effort to answer questions like “Around exam time, are [MIT] students eating bad food? Are they drinking a lot of coffee or taking other things to help them study?” Ghaeli says. The team is also watching how cortisol, adrenaline, and other bio-markers for stress change as the semester unfolds. City planners could use data like this to understand how physical environs, such as proximity to highways or public parks, affect residents’ stress levels.

Get Today's Headlines in your inbox:
The day's top stories delivered every morning.
Thank you for signing up! Sign up for more newsletters here

Ghaeli prepares the sample bottle and stores it in a cooler. Next stop: a nearby lab where co-project lead Mariana Matus will analyze the sample. Matus works in the MIT lab of biological engineer Eric Alm, who created the Underworlds project with Ratti in 2014. Alm says he and Ratti originally conceived of Underworlds as citywide surveillance, gathering information to benefit public health. They didn’t know if it would work, but the Kuwait Foundation for the Advancement of Sciences funded it despite the risk. “The short-term benefits were unclear,” says Alm. They told the foundation they would build a platform that could vastly increase what they could learn about public health, delivery date “at some point in the future.” That future is coming much more quickly than they anticipated.

MIT’s Ghaeli peers into a Cambridge manhole in 2015; at her feet is Mario, the prototype sewer-bot.
MIT Senseable City Lab
MIT’s Ghaeli peers into a Cambridge manhole in 2015; at her feet is Mario, the prototype sewer-bot.

THE IDEA THAT SEWAGE WATER can provide public health information stretches back more than 150 years, to a time when London was in the noose of a cholera epidemic. Scientists at the time believed cholera was caused by a “miasma” in the air, but after a particularly bad 1854 outbreak killed more than 600 people, a London doctor named John Snow demonstrated that each case of the disease could be traced to one particular London water pump. The outbreak ended after Snow persuaded city officials to remove the pipe’s handle and thus access to its water. Today, Snow is considered the father of epidemiology, and his discoveries led to our modern sanitation systems.

Poop is back on the map because scientists have learned trillions of microbes live in our guts, meaning we excrete information. Researchers have used sewer water to gauge whether polio vaccines are working to wipe out the virus and even to determine if there’s a difference in the collective microbial signature of cities that are, on average, fatter or thinner. They’ve also done non-microbial work looking for toxic chemicals. But these studies have largely been conducted at waste-water treatment plants. The Underworlds team thought it would be more useful to measure sewage near its neighborhood sources. To prove it, they ran a research project in 2015, looking at sewage in Cambridge and Chelsea.

One of the first questions to resolve was when to take samples. At a treatment plant, the flow is largely the same at any time of day. But at a neighborhood manhole, it depends on time of day and the number of people flushing. The scientists also have to be able to distinguish toilet flushes from runoff from showering and dish washing, and then figure out just how many people are flushing in the time leading up to collection of a sample.


Once they could, the difference was clear. A sample collected at a manhole “looks like a mixture of people,” says Matus, a PhD candidate in computational and systems biology. At a treatment plant, “it looks more like an aquatic environmental sample,” like a river or lake, “than a mixture of human poo.”

Matus says they were able to show that 65 percent of the microbes under the manholes were human-related, as compared with up to 20 percent by the time the sewage reaches the Deer Island treatment center. It can take two days for sewage to go from toilet or tap to the treatment center, time during which microbes from our innards mix with animal microbes and those that proliferate in the tunnels, where oxygen levels are far richer than in our guts. The mix of microbes matters for research: Though scientists haven’t yet teased out the relationships, the microbial communities in human guts seem to be different in patients suffering from diabetes, autism, or colorectal cancer than for those without these conditions. But the differences might depend on the ratio of different kinds of microbes, rather than their mere presence or absence. Capturing an accurate snapshot of these microbe communities may be crucial for finding answers.

Chemicals also change in the sewers: At the manhole sites, the researchers could find acetaminophen, the main ingredient in painkillers like Tylenol, both in its chemical state and in a metabolized form that meant it had been swallowed by a human. But by the time the flow reached the sewage treatment plant, the metabolized version of acetaminophen — the one that had definitively been processed through the human body — had disappeared entirely. Alm says this shows that public health officials who rely only on data from treatment plants could miss out on important information about residents.

Sampling in neighborhoods is much more complicated, and expensive, than taking the pulse of an entire city via a waste treatment center. The Underworlds team needed to devise a system that would work exactly the same in any location, could be managed by someone with no background in microbiology, and didn’t cost that much. This being MIT, they thought: a robot.

 Analyzing sewage requires gear for biohazard protection.
Webb Chappell
Analyzing sewage requires gear for biohazard protection.

CREATING A SEWER-BOT wasn’t just an engineering challenge. The project required people with expertise in civil engineering, architecture, urban planning, biology, data science, computer science, and artificial intelligence. They named their first robot Mario, after the world’s most famous plumber, Nintendo’s Super Mario. At 18 pounds, Mario was bulky and made entirely of custom parts.


Luigi (Super Mario’s taller twin brother) weighs in at a comparatively svelte 4.5 pounds. The Luigis in existence are made of off-the-shelf components, which lowers costs dramatically. The Underworlds team has been using the Luigis for the project on college student habits and to gauge biochemical signatures at a specific hour in neighborhoods in Boston, Cambridge, and Kuwait.

In the future, Cambridge wants to see whether different viruses, drugs, or pharmaceuticals are prevalent in different neighborhoods. Boston officials would like to know whether humans or animals are the source of fecal matter discharged from overflow tunnels into the Charles River. The answer would mean either an expensive fight against illegal sewage hookups or a more routine effort to get residents to pick up after their pets.

Sam Lipson has watched this project with interest almost since it started from his post as Cambridge’s director of environmental health. He served as a sounding board for how this incipient system might be used as well as for the conflicts that might arise. Lipson says one of his first questions involved privacy. Would there be any way to trace information found in the sewage back to a particular person? Matus had a ready answer: The DNA in our waste is nearly all microbial, not human, and so, she says, “there’s not enough data to assemble even one [human] genome.” Plus, even though manholes provide samples relatively close to the source, they still include the waste from as many as 4,000 households in each bottle.

Lipson also wanted to know who owns the data. “Imagine, for instance, that Underworlds works the kinks out and determines that they have a pretty good idea of how to get a reliable number of the rate of depression,” says Lipson. “That’s very valuable information that could be marketed to different companies, perhaps, who want to put their efforts into certain communities with higher rates of disease. And a scenario like that raises real questions.” Ghaeli says future partners, such as the City of Cambridge, will help the team decide both what information to collect and how that information should be shared.

Mario, a robot built at MIT to gather sewage samples, prepares to descend on a mission.
Mario, a robot built at MIT to gather sewage samples, prepares to descend on a mission.

Underworlds can’t answer all the questions that public health officials pose. For instance, Lipson was hoping to get data on influenza levels, but it turns out the flu virus breaks down too quickly to provide a clear signal. Still, he’s excited by the prospect of having new kinds of data. He thinks it could change how Cambridge helps residents, letting them know, for instance, if anti-obesity or anti-drug measures are working.

To have broad impact, Underworlds needs to create robots that do their work when the manhole covers are in place, without people supervising them. When the team deploys Luigi, the area around the manhole is cordoned off, and someone needs to raise and lower the robot three times, to fill the three bottles needed for a full sample. Ghaeli and Matus are leading development of a new robot  that can be put inside a manhole and left there. It will carry three bottles at a time, lower itself, fill the bottles, then raise itself back up and wait until someone comes to collect its samples and replenish bottles. Where Luigi senses only temperature and flow, its successor could also send alerts about low battery levels and perhaps be fitted with environmental sensors that measure aspects of the water, such as acidity — or pH — that currently are tracked in the lab. It would be even easier to swap out parts and could have different kinds of filters to allow different purposes.

To help advance those new uses, the team is working to create a microfluidic array for the robot, which will let them analyze tiny amounts of sewage directly using a variety of chemical reagents. That could yield a sensor that acts as a real-time bug detector, so emerging viruses or bacteria don’t have to be discovered in a lab. Underworlds hopes its robots will eventually become a plug-and-play system for answering questions dreamed up in labs and city offices.

The team’s next set of research projects will include seeing if it can use sewage to map antibiotic resistance and find its causes and to search for bio-markers for obesity, diabetes, and other chronic diseases. If the Underworlds researchers can create an inexpensive robot that any municipality could afford, the potential seems limitless. “We knew there was a reservoir of information [in sewage], but I think by carrying on the project, we realized this is even bigger than we thought initially,” says MIT’s Carlo Ratti.

They started out on a robotic fishing trip fraught with risk. It looks like they’ve landed something big.

Cynthia Graber is a writer and podcast host in Somerville. Send comments to Follow us on Twitter at @BostonGlobeMag.