# Running the World's First Autonomous Drone Program on a Floating Offshore Facility with Shell - Event: NestGen ’26 - Published: October 2, 2026 - Length: 29:30 - Watch: https://nestgen.org/library/watch/LP43NidoiGY/ > The transcript of a recorded NestGen session, for reading or for use with an AI tool. It comes from the video's > captions; automatic captions can mishear names and terms. ## About this session Offshore facilities operate under some of the toughest inspection and safety constraints in industry. When access is limited and response windows matter, autonomous drones can give operators a new way to inspect assets and improve visibility without adding unnecessary exposure for on-site teams. In this NestGen session, Shell will share what it means to run the world’s first autonomous drone program on a floating offshore facility, and what oil and gas teams should think through when evaluating similar deployments in remote, high-consequence environments. ## Transcript **[0:00]** Awesome. So, welcome everyone to the first session of the Western Hemisphere that is we have with Shell. So, it's going to be about running the world's first autonomous drone program on a floating offshore facility. And this is Deep. I am your host today. So, I work in FlytBase in the business team. Manage all our, you know, customers and accounts across the oil and gas sector. And today with us we have Ross Duke from Shell. So, he's an offshore robotics program manager. He resides in Houston where his job is actually to, you know, work with robots and drones out of a lab like their Shell Technology Center and then put them to work on real assets. So, he started as a deepwater robotics engineer doing subsea inspections and now he runs robotics across, you know, Shell's offshore portfolio. So, today Ross will walk us through something that nobody has **[1:01]** ever ever done before which is again running an automated drone autonomously from a floating facility in the Gulf of America. So, yeah. Ross, over to you. >> Hi there everyone and good to be here and invited to speak uh today at the conference. I'll bring up my my slide deck now so we can get started. And let me just bring that here. Okay. Okay. So, before we get started and go too far into the uh uh the presentation, I do have to put up for reasons of of corporate personal uses a cautionary note which says basically, "Please do not use any information we present here to make any um projections on Shell's performance or any investments on that, okay?" And so, I think what Deep just said there is important for us. Um we've been looking **[2:03]** at the use case uh for drone in a box offshore for a while. We know that in locations uh such as Norway, there's a lot of good progress has been made with the DJI Dock 1, for example, on fixed structures. When we look at our uh basin within the Gulf of America, we operate primarily deepwater facilities, meaning that their the connection to the seabed is no longer a fixed uh connection in the same way as in shallow water. And we can see our facilities move quite a bit. And when we think about the comparison of of deploying drone in a box offshore versus onshore, of course, the environment itself produces a number of challenges which we have to work through as we look to scale our program. So, of course, we're remote. We're roughly 130 mi offshore on the asset we went to for our first go. Our facility moves. So, whilst we operate tension leg platforms, which are connected to the seabed by long **[3:05]** um steel members, which provide a tension back to the the the seabed from the surface, we do see up to about 16 ft of travel uh during a 30-minute flight. Um so, here, unlike on land, we we do have that that to encounter. We also um have a very congested uh layout. So, the other piece we have to work around is on a onshore facility, generally there is always um a laydown space or area which is, you know, uh cut off from the rest of the facility, which we can install our drone in a box. We've got a safe place to operate and maintain that box. Uh offshore, uh every piece of deck is is premium. Um so, we had to work around how do we manage operations with cranes, with marine vessels, and with all the other risks that come with with working offshore? And that also does create the issues with with compass and GNSS um degradation at times. **[4:05]** These assets are incredibly valuable to our business. So, having a drone-in-a-box capability for instant response and accelerated recovery is incredibly valuable to us. In April 2026, we took the first uh major step on our program where we did a a simulated BV loss operation on one of our facilities in the Mars Corridor. So, the Mars Corridor is an incredibly prolific part of the Gulf of America when it comes to oil and gas uh production. Absolutely critical to US energy supply and has long been an area that's pioneered the deepwater exploration and production um domain. We took a DJI Dock 3 with a Matrice 4 TD uh with an onboard obstacle sensing system. Uh we took that offshore in a lifting basket with um additional um equipment such as a Jackery power supply, Starlink, **[5:05]** all the pieces that uh we needed to do a temporary deployment. We also had a a 5G or sorry, Starlink LTE network on the facility which we used. So, we had a SIM card um into our uh Azure router there so that we could test failover both the tablet and Starlink systems. And we had a secondary ADSB in receiver as well. The crew in command, so the crew in command was primarily offshore. So, for reasons of ensuring regulatory compliance, we made sure that we we designed our flight crew around that. And what we did is we made sure that the people onshore could use FlytBase um to you know, take pictures and control the payloads. So, nothing on the actual aircraft itself beyond uh payload control. So, um that kept us line regulatory-wise, which is good. Um it simulated the FlytBase and the kind of latency we have from onshore crew working with an offshore system uh still worked with the backhaul we had. **[6:05]** And it made it uh we also were able to check a number of other things around the automation, which was very good. Some of the big problems we had to uh come into and address were things like the moving RTK reference. So, for those of you that, you know, use RTK, you'll know that if you move the base station midway through an operation, that can off- often create error. Um we know that on open water, we don't have the same references we do onshore um to design uh uh teach a repeat mission or a map capture mission. We don't have that same capability. We know that the steel, the heat, the interference that could be caused by all the other systems working at that facility can can can cause major issues for GNSS systems and for compasses. And we do see them routinely when we do VLOS operations with you know, Matrice systems, with Flyability's ELIOS systems. So, testing the limits uh of our command and control continuity was really important here. We also know knew that dark sighting was of **[7:06]** course really important. So, again, where we build offshore facilities is deck upon deck upon deck. And we have things like cranes, we have things like class one division two areas which are called explosion risk. We've got all the uh helicopter landing pads um uh issues that are around putting anything else there. So, when we went offshore, one of the big uh tasks that we performed was to understand what is the right place to site uh drone in a box on an offshore facility. And the reason we look at that is is really because we want to get very good at managing from shore simultaneous operations. So, today when we perform VLOS operations, our pilots do a great job to work with the facilities. That could be on Well, some of our facilities like Mars, we have we were drilling operations, we've got marine operations, we've got crane operations, we've got painting, we've got maintenance. There's a lot of activities which go on every single day on these facilities. So, managing um simultaneous operations is how we deconflict for drone operations within that **[8:09]** context. And so, many of the problems that we were trying to address were not specifically, you know, drone technology. They were more operational in nature. And I think that that's where a lot of the value came from from doing this piece of work. Wha- When it comes to autonomy and using the autonomy that we have through flight base and through the you know the DJI dock in this case, we had to make sure that we could earn that right to use the automation. And so, we did a number of processes. We knew that RTK was going to be incredibly important to how the uh drone navigates across the facility. Of course, we have both, you know, visual means, we have GNSS, and we have RTK all supporting the navigation of the drone across the site. And so, we may We have to make sure that from uh shore, we could um calibrate RTK from from locally, we could calibrate RTK. We had to make sure that on takeoff and landing, we had some understanding of the both the the weather and uh ocean limits. So, um offshore, we we don't have wind and **[9:11]** rain to contest with. We also, because we're in a floating facility, have heave, pitch, and roll. Which anyone that has set up a a drone in a box typically follows the instructions of the uh vendor who built that system. And typically, we always want those boxes to be nice and nice and level. So, understanding what heave, pitch, and roll would do to takeoff and landing was of course really important. We also wanted to make sure that we understood that how we could do mapping effectively to build automated missions in the future. So, we did do a very good manual survey of the facility to build a nice 3D reconstruction of the site, which we could use then for mapping. The other piece was the shared control. So, we like I mentioned, we we were limited in what we can do you know on this piece from a regulatory perspective because the the the the flight crew has to be offshore. But on payload control, we were we were able to simulate we could take images, we could switch camera settings. So, we get we had good confidence that control from the shore was going to work **[10:14]** effectively from the beach once we left. And we also ran a number of autonomous missions through the FlytBase flight orchestration software which is all supervised from a pilot locally on deck with a hand controller. This was really good and actually through the the process we we even tested a number of the fail-safes. I'll talk about those now. So, one of the most important things for us here was we know that the drone in a box technology works from our our experiences onshore. We've been testing drone in a box systems now for for 4 years and and many of our sites we're now doing you know 1,200 flights a month with drone in a box. So, we know the technology can work. For us, it was understanding how we might fail given the different environmental challenges that we faced. And so, some of the big things we tested were things like lost link both for the aircraft to dock, from the dock to the internet, and from shore the shore-based pilot to to internet as well. And so, we did things like you know disabling Starlink mid-flight **[11:18]** to see what would happen. We did things like shutting down the e-stop during the flight as well to understand how the alternate landing would happen as well. We also went through a number of failure injections into our flight test program such as we would test the smart return to home feature to make sure it avoided the no the no go zones that we'd set up around things like the Derrick. And it would maintain the offsets we'd set up in the obstacle avoidance software. We also tested the you know if we had closed down a certain sector, so say we've got crane operations going on on the east side of our of our facility. We would want the drone to avoid the east side of the facility. So we did we we we ran through each of those modes. Again, this is really about how do we operationalize this safely and how do we deconflict appropriately when we go to fully scaled operations. And one of the things we're also really happy with is we were also able to conduct a night flight of our facility. Which you know just looks brilliant I think in the in **[12:20]** in the images. And we're actually able to test some of the things that we we hadn't really thought about beforehand either. Things like could you use a drone in a box offshore to support search and rescue should either you or or another sea sea farer have an incident where someone goes over deck at night and and we need to try and support them and find them. We also were able to test a number of items around how far could the drone travel towards the underside of our structure. That's a place where historically we've had real issues both with the things like glare coming off of the of the sea as well as C2 degradation once we went under deck and and got a little bit more obstructed from the radio frequency. And this is really really important to us cuz understanding that we that we have these proof points allows us to make a more informed decision about how we take the next steps in scaling our program. **[13:20]** And so we believe this was the first fully autonomous dock landing and operation on a floating deep water facility that's been done. >> [clears throat] >> We did a number of progressive stages of the test, which we've talked about here. We went and we did a number of the inspection types that we've seen our V-loss pilots do in our existing V-loss program. And we demonstrated that we could perform operations both at day and at night. And so really important to us to understand the capability to help us scale the next the next evolution of the program. What we'd recommend other operators doing and and really I think about before they they go and and do this and and what we've taken from it really is to make sure that you know, communications [clears throat] I think networking from my experience not just in drones but in all aspects of robotics, networking is often, you know, the most important piece to get right when we turn up at a site. Understanding positioning and how RTK can can both help and and and potentially impair you is really important, too. I really having a good view on on where to site **[14:27]** your dock. So, our biggest takeaway was we need to have our dock location as far away from the busiest crane on the facility as possible while still being outside of any kind of class one division two you know, potentially explosive environments that could be there. And also understanding how that that dock sighting then impacts the access egress routes for personnel in emergency situations. Right now we have a really good existing ways in which to deconflict other operations on our site from drone operations. So, things like inbound helicopters other people doing work we've got really good ways to deconflict those, right? So, it's really about how do we try and do those same things but with the crew not offshore. So, offshore right now we've got the the the benefit of of not really having to do too much. We we can go and we can speak to people. We can use radio. We can do all those sorts of things. So, how do we bring those same processes to bear **[15:29]** when our pilot is on shore? So, that's been a really kind of important part of this. I think the failure rehearsals were really good. We had a very structured program around testing those. And we went through and kind of ticked off the boxes. Understanding how we feel is not just important to us at Shell, but also to people like DJI and to FlytBase and then how they evolve their product um to to deliver more effective system in the field in the future. And I think a big piece we're we're still working on within Shell is, you know, once we do this, once we do this really hard thing on getting the drone in a box offshore and scaling the program, we know there's going to have to be a good way in which we serve data to people who are not necessarily drone pilots, but are within the business. And it's been good to see FlytBase put some thoughts on that at the start of this um uh webinar as well. So, like I said, I think the aircraft is for us has been very much the easy part. The hard part is the operating model that goes around the entire program. Uh so, from there um and that first flight, we gained in um the June or July **[16:33]** of this year, just before July 4th, we got the what we believe to be the FAA's first waiver for drone BVLOS drone in a box operations across all Shell owned and operated infrastructure in the US Gulf of America. That was a huge moment for us. For a long time, the FAA has not uh taken a position uh to provide um that type of service, waiver services for the uh high seas, very much due to some of the kind of legal pieces around ICAO and jurisdiction. And so, for us to get full regulatory approval was massive when it comes to be their ability to scale this program and to do truly BVLOS operations offshore. Within Shell, it's really important to us that we maintain our license to operate. And really the starting point for that is making sure that we get regulatory compliance right. And so I think that now we've got that, we're in a process we're looking to scale roughly to about 11 docks, 7 to 11 docks by the end of 2028. And we're already taking the first steps in that now. For us the the the really big **[17:36]** use case we have with this program is around hurricane response. And that really for us is the defining operational risk when when operating in the Gulf of America or even on the Gulf Coast. Um anyone that does operate a refinery or offshore plant there knows that hurricane events can be really uh detrimental to your throughput, to the your asset availability. And it can have major consequences for for HSSE and and asset integrity as well. So we see that Drone Up Box is a really key enabler for us to be able to support that better. And thank you very much to everyone that supported that that program, to Flight Base, and to everyone for hosting this gym. I'm happy to take some questions now. >> Uh thank you so much Ross for this, you know, very very insightful session. Uh again, the deployment was first of its kind and that is what everyone knows, you know, over here. So I would start with a **[18:38]** question from Santiago. So are there any really good flying conditions or are they just severely limited by the weather? >> Yeah, so this is this is a good question. The way um our kind of time period worked in Mars was um we had some pretty rough weather and a lot of helicopter activity right at the very start of our deployment. And so due to the helicopter activity we were conflicted for operations for the first couple of days. And so when when that was finished, the weather actually improved. So, we got to test in pretty good conditions. But, we have been testing in in in pretty harsh weather since. And in general, anything we can do with a VLOS pilot, and when we do best part 2,000 flights a year with VLOS pilots offshore, um anything we can do with a VLOS pilot, we can do with the drone in a box system. We found that on landing, um the landing really likes kind of sustained um uh wind **[19:41]** speeds. Um it it does take a lot more time, and it's a little bit more sensitive to um very turbulent air. So, again, looking at where you site the drone in a box system is incredibly important, because if you can provide some shielding to or or avoid areas where there's very turbulent wind conditions, that's very important uh to to the the success of the technology. >> Right. Awesome. Uh the other question the next question that we would take is offshore platforms have a strict uh hot work and ignition source rules. How did your HSEC team classify the drone? Uh was it treated as an ignition risk? Uh and what did you have to prove before they allowed autonomous flights over uh live producing uh facility? >> No, great question. And like I say, we we we we we do best part 2,000 flights a year on our facilities. So, drone operations are certainly not new to our HSS teams or operations teams on those platforms. So, we've got an established way of doing **[20:44]** it. Indeed, you you need to look at the the right uh permit uh some hot work if if that's the case. And we do um have have rules around, you know, where we can set the drone in a box, and we have the ability to set no-go zones around, you know, class one division one, class one division two areas that align with the NFPA um 496 regulation on on that. So, um yes, we do we do have those rules that are in place. Um, we have pre-existing procedures, uh, operation procedures to to handle it. And it's certainly not something that's new. It's about how do we map those same procedures to the system and keep everyone safe. >> Awesome. Uh, the next question is, uh, the facility is floating, so it's constantly moving. Uh, you've already mentioned that, you know, it moved up by 16 ft or so. So, how does the drone dock actually handle that? Like, what were your primary observations in that? Uh, does the landing pad compensate for the motion or, uh, did you have to like set some sort of tolerances for sea state? Uh, did you have to put the dock in vehicle mode? Like, what was the whole idea behind that? >> No, **[21:47]** great question. Uh, and this goes to, you know, what we were thinking about how how we might fail when we operate. And so, for for us, um, like I say, during the during one of our 30-minute flights, we've seen about 16 ft of of lateral movement of the facility. Um, which, you know, it is not actually that much when you think about it, because when we do return with the the drone, typically, maintains quite a high, uh, separation above the the dock. And we've got quite a wide field of view. So, in general, that the the drone, I think primarily through its visual system, was able to, uh, re-determine the position of the of the dock and and and and course correct there. So, that that was useful. On the second piece around compensation, um, we didn't really see the need for any compensation here. And and our testing offshore, at least on on our, um, tension like platforms, has shown that we don't need it. We typically see about less than a degree of rotation in any heave, pitch, or roll or or heave or, uh, sorry, roll or roll or pitch. And and the heave that we **[22:50]** see is is is not not significant. We have not yet got to the stage where we've got a really good characterization, I would say, on the tolerances for sea state. And that's something that we're going to have learn and test as we go. Again, we we we anticipate that as an industry here, offshore operators are going to have to kind of carry the can on this and try and understand what what is the the maximum heave, pitch, roll, and and lateral excursion we can take. >> Right. The other question that we have is are you guys planning on deploying the docks for any other use case apart from offshore? Shell has multiple onshore sites globally. So, what's your larger plan? >> Absolutely. Yep, we are we've we've just completed a a major tender here in the US for drone in a box services covering both offshore and onshore. And we're doing the same for Canada. Um so, we've got about 47 sites in the US and Canada. And we're about to go and do Europe and the Middle East next and and then potentially APAC in the future. So, **[23:53]** there there's lots of I think good use cases globally uh for drone in a box. Um we see demand from our um Shell own ventures as well as joint ventures, especially in the Middle Middle East right now. That's a big demand there. And so, yeah, we're we're very much in the case of no longer kind of testing this technology. It's it's now a proven technology at this point um especially for onshore. It's about how do we scale it the right way. >> That's amazing. Uh so, for uh like there are you know multiple hundreds of offshore facilities globally. Uh in the GOA in the Gulf of Mexico specifically, uh the reason is post-hurricane response. Like you know, getting the getting the site run faster. So, on other areas on other parts of the world, what can the primary use cases be for people who are planning to deploy this offshore? >> No, it's a good it's a good good point. And so, like obviously for the Gulf of Mexico, you know, hurricane response is something that we've uh spoken about um as **[24:53]** the defining operational risk. It's the one thing that you know, it's different to on many other the operating regions of the world. Uh but but in general, there there's just a paradigm shift. We see drone in a box not as like a technology deployment, but as infrastructure. And this is about, you know, not a single use case. There's no single use case for drone in a box. This is about making your infrastructure more capable. Um Today, you know, we we we we we've done that historically with things like CCTV, with uh sensing technologies, so that we don't have to person go out in the field and and trend values in a gauge, for example. Um And so, we do see this as a further extension of our uh infrastructure capability. And to to that end, we we we see a broad variety of use cases, and many of which uh we didn't we didn't even know that. There's a good anecdote from this trial at Wheatland Offshore in April, whereby [clears throat] when when when it we posted it internally on our on our on our pages that we we use for, you know, socializing this type of type of work. Uh but but three different people came to us with use cases we'd never thought **[25:56]** of before we tried it. And so, we feel like once we bring the infrastructure capability, we we we we will see an increasing demand. And everywhere we where we have globally deployed drone in a box, that has been generally the case. Where once it's installed and and operational, it very much becomes the case of the use cases start to snowball. So, so we think there's a very good story there, um without being one. But certainly, the hurricane response is the one that makes everyone in the Gulf of America stand up and say, "Yeah, we should invest in this infrastructure." >> Yeah. Yeah, no no, that makes 100% sense. Like, you know, the ROI is pretty much very clear >> [laughter] >> of the whole program. If even if you're able to save like 4-5 hours uh in that case. Uh we have a question from Sergio. Uh what are the mitigations if the conditions are not favorable for the drone to land? >> No, great question. Um so, there's both prevention and mitigation, right? Uh so, Sergio from Cyberhawk. So, um the mitigations for us really are kind of left on the side of the bowtie. We're just trying to build a good characterization of what the sea states are so that we don't **[26:59]** fly and we don't fly in a rising tide. We have incredibly good information offshore about what the sea state is doing over the next 6 to 12 hours as well as what the weather conditions are doing generally. Um so we do have the ability to fall back to our current landing site and take over control the aircraft through FlytBase and do it do a manual land now. So we do have that option as well. That's what we'll probably default to if we can't land in the dark, we'll probably land at another safe location which will be designated prior to the flight and as part of the the operational documentation we put in place for the asset. >> Awesome. Uh we have another question uh which is what are the differences in data you generate from repeated inspections versus manned yearly drone inspections? >> I don't think there is a a a difference really. Uh this is a case about how how do how how how much inspection can you do uh a risk basis. So the risk basis **[28:00]** considers the you know the cost of getting the data and then the complexity of getting the data against the value of it. And so one of the things that we've been thinking about quite a bit is really around um thermal data. Right now we don't do a lot with thermal data because we can't we can't trend it in the same way because to go and get it we have to get a drone crew offshore. We have we have to fly fly for us. That means they have to get uh you know uh uh a drone crew offshore. There's we're paying for the helicopter flight there. We're paying for the helicopter flight back. We're bed space when they're there. But if for 1 hour of of a pilot's time a day we can get thermal data on our fuel gas compression system or our vents or our flare, we can start to deliver those characterizations and trend things over a longer time without the same resource cost. And so I think that's a really exciting space because there's probably nuggets within that that we can unlock better insights. But we just from a data gathering perspective, we we can't do today. >> Right. I I think we just have 30 seconds left and there's a very interesting question from Jared, which is do you believe you **[29:01]** would see the same success with a different brand system if you couldn't use DJI? >> We we build what we want to build confidence in a broad range. You never want single source supply on anything within your business, hopefully. Um so we we have been testing other systems. But you know, we we build that trust logically, right? So once we see it work, we we build more trust. Thank you for all the questions. Um thank you for hosting. I need to leave and I appreciate the opportunity to present. Thank you so much. >> Thank you, Ross. Thank you for your time.