I love this analogy: “Let’s say you want to get a 3D picture of the inside of your body today; you have an MRI, CAT Scan, and Ultrasound,” says Jeremy Suard, Co-Founder & CEO of Exodigo. “But it’s 3 devices, 3 pieces of software, and potentially 3 different doctors that have to analyze them.”
This is the challenge faced when seeking to comprehensively map underground utilities. As noted in ASCE and UESI literature, the use of multiple sensors is the best practice for catching-it-all. Ground penetrating radar (GPR) and especially multi-channel GPR (MCGPR), time-domain electromagnetic induction (TDMI), magnetic (MAG) methods, sonic, ultrasonic, etc. I love my MCGPR unit, but have realistic expectations, especially in various soils type and moisture profile environments.
Mixing multiple sensors means having expertise in interpreting/analyzing each, and then you need to put it all together in a map. I don’t care how much MCGPR gets billed as #D underground mapping, it still takes a lot of experience and skill to “read the tea leaves”. But what if analysis with other sensors could be automated to corroborate and/or fill in what the other missed. I’ll make an analogy to the integrated positioning systems that are on mobile scanning systems and some UAS: an IMU is processed along with the precise GNSS—one can help catch outliers in the other.
Processing Together With AI
Exodigo has done this for multi-sensor underground utility mapping, for ad hoc systems mounted on carts, or for some limited applications, even a drone. “It's better than doing all of them separately. And it removes the human interpretation factor in the field,” says Suard. “Because today, even though a lot of calculations are done in the field on your device, without broader context or corroborating data, it's a lot of human interpretation. You can lose a lot of data there. A lot of data, lots of room for decisions. So, you need to be a real pro at GPR, or whatever sensor.” With staffing shortages across AEC industries presently, let alone finding an expert in each type of sensor, such an approach can be appealing.
An example of multi-sensor processing and analysis showing the locations, quality levels, and type of detection method used. Source: Exodigo
“So, the way we do it is that the measurements in the field are very, very systematic,” says Suard. “Then all the calculations are done on the cloud computing with our AI, combining all the different signals. It can be a lot of data; like 150 GB per acre. And everything's uploaded to the cloud. The AI basically processes all of that very, very fast and efficiently, then we have quality assurance by engineers. Then we export the map you need.”
UAS can be the platform in very limited conditions, but it is still cool that is an option. “Once you've got the basics down, once you switch to an operating manner in the field that's just crisscrossed with all the sensors, then you can put the sensors on a drone,” says Suard. “The first ones to do that was actually the mining industry, by the way. We didn’t invent that, but we just put way more sensors on the drones.”
Suard explained that sensing from drones must be done very close to the ground. “So then it's an operational question. Like, if you can you fly a drone that low, why not just use a cart or trailer? It really depends on the site, conditions, environment etc., and the application. Of course, inside of cities, we will always use the cart to be closer to the ground and make sure we catch everything.”
Services, Not Hardware
Exodigo does not want to get into the hardware business; rather, they tap into the expertise of those who develop the best sensors. “We buy sensors from the best sources; we are not interested in selling hardware. Our vision in the future is to be able also to pass the tools to anyone who wants to scan. And then it's easier to do smaller projects everywhere, we will still offer the full service if someone wants it. But once we've reached a certain point where we trust the tools in the field to be very simple, no problems, no bugs, then we can pass the scanning to everyone—your students, your interns, your new crew members—because anyone can do it.”
Typically the multiple sensors are deployed on a cart, but under certain conditions, and depending on the application, some can be deployed on a UAS. Source: Exodigo
Presently, Exodigo sends their own teams and sensor arrays out to clients sites; they have teams in Europe, and both the east and west coasts of the U.S. “We buy the best sensors we can find,” adds Suard. In examples he showed me, they have a cart with GPR and other sensors, GNSS for positioning, etc., and some low-flying UAS examples.
Case Study
What really piqued my interest in this was a formal test performed, a pilot study, of underground utility mapping in a suburban setting in Santa Rosa California. The results were published in a white paper (Exodigo will send it to you). Exodigo did a crisscross pattern on a subject street with multiple sensors on a cart.
The verification step was to send conventional locate teams to cover the same site. From the white paper:
“The findings of the Utility Research “SUE” were marked using CGA Best Practices standard paint markings and flags inside the white lined work area. (All of the markings were by the subcontracted locator since the area is a private mobile home park). The locator used a combination of limited mapping, field research, surface features, electro-magnetic locators, and GPR. The markings appeared orderly and consistent to sound industry practices.”
The conclusion of the study noted that:
“The Exodigo findings were consistent with of the suspected utilities marked by the locator (which appear to have been primarily based on field interpretation and induced Electro-Magnetic Locating Equipment). In addition to all the accurate locates of the traditional locator, Exodigo cited additional lines and was able to eliminate inaccurate phantom lines. (The Locator’s mismarks were later suspected to be attributed to soil features associated with previous construction). There were no lines that Exodigo missed that were found by the traditional locator.”
In the pilot example, they found twice as many utilities as the conventional locator, avoiding potentially 5 hits from missed lines. Suard says that clients almost without exception find more lines than conventional locate methods.
A Model Moving Forward
I would expect to see mapping carts hitting the market in the near future that have multiple sensors, or package options—the hardware is the easy part. The heavy lifting is processing and analyzing these huge datasets and automating the analysis. It is likely that this may never be practical to do onboard, so services such as this example will become standard for underground mapping workflows.
Subscribe