Seafloor Systems Saves 4 Hours of Labor per Robot Build Using AWS IoT Greengrass

Hydrographic robot developer Seafloor Systems needs to build and manage its fleet of uncrewed survey vessels (USVs) efficiently to effectively scale. The startup creates bespoke autonomous robotic boats for its global customer base. Remotely operated, these boats collect data to chart bodies of water. Seafloor Systems wanted a globally scalable set of tools and resources to streamline the development and production of its robots.

Seafloor Systems found those tools and resources on Amazon Web Services (AWS). Using AWS Internet of Things (IoT) solutions and AWS RoboMaker, which provides a fully managed cloud infrastructure for robotic developers to simulate, test, and securely deploy robotic applications at scale, the startup can build and update its robots more simply, quickly, and cost effectively to better meet customer needs.

Seafloor Systems serves hydrographers at research institutions and companies globally. In addition to its USVs, it offers instruments, software, and support for hydrographic projects of all sizes. The startup began using AWS in January 2020. “We’re a small team, so we’re always balancing development, deployment, and building up our R&D,” explains Marcos Barrera, lead robotics and artificial intelligence research engineer at Seafloor Systems. “You can’t build an infrastructure and have people in disparate locations work well together without a strong backbone and scalability. That’s one reason I was drawn to AWS.”

In 2020, Barrera began exploring AWS IoT Greengrass, an IoT open-source edge runtime and cloud service, to deploy updates to robots in the field. AWS IoT Greengrass would provide the agility Seafloor Systems needed to improve its applications quickly and securely—enabling it to better respond to customer needs and add more functionality without having to start from scratch. However, Barrera also needed a way to perform simulations from home during the COVID-19 pandemic. After searching for a solution, he began to explore application development on AWS RoboMaker.

By running simulations on AWS RoboMaker, Seafloor Systems could quickly test a variety of use cases on its robots without the complexity. This solution would eliminate the need to test software on physical robots in remote locations, enabling the company to reduce costs and increase the speed of its delivery. Because AWS RoboMaker and AWS IoT Greengrass operate on cloud infrastructure, Seafloor Systems could use the flexibility of the cloud to streamline and scale the development of its robot fleets.

AWS RoboMaker is also compatible with Robot Operating System, an open-source collection of frameworks for robot software development. This was an important consideration for Barrera; with this compatibility, Seafloor Systems could collaborate with a community of roboticists. Robot Operating System compatibility would also enable greater modularity, allowing the startup to implement and test different functionalities without replacing its existing work.

Barrera initially used the first version of AWS IoT Greengrass to manage software bundling. It wasn’t until joining the private beta in November 2020 for AWS IoT Greengrass Version 2, which adds ease-of-use functionality in response to customer requests, that Barrera saw how the service could change how Seafloor Systems builds its robots on a large scale. AWS IoT Greengrass Version 2 builds on the ability of AWS IoT Greengrass to package a software stack by adding modularity, letting users set their stacks up as a network of dependent components—a capability that complemented the startup’s bespoke designs. When building a new robot, it can package all system components and reuse as much as possible, increasing efficiency. When Seafloor Systems needs to exchange a thruster system or sensor package, for example, it can do so remotely instead of having to start from scratch. “It’s less work and really surgical,” says Barrera. “And no one on my team has to fully understand the nuances of all the interconnections. We can just plug in different services for the robot, and then all the necessary drivers, connections, and algorithms are there.”

In December 2020, Seafloor Systems deployed a proof of concept for deploying and managing software at the edge in its robots, using AWS RoboMaker and AWS IoT Greengrass. A key benefit of the solution is its ease of use, especially because Seafloor Systems maintains unified product development across physical locations. The solution saves at least 4 hours of manual labor in building the software stack of each robot. Given Seafloor Systems’ current fleet of 2,000 active HyDrones, its smallest and best-selling USV, that savings adds up quickly. “As our technology stack has grown, we wouldn’t be able to manufacture at the scale at which we sell without using AWS IoT Greengrass,” says Barrera. Using AWS IoT Greengrass also enables Seafloor Systems to deploy applications to robots in remote locations in times ranging from minutes to 1 hour, compared to the days or weeks it would previously take. The company saves approximately $1,500 per robot deployment by using AWS.

Seafloor Systems plans to use AWS RoboMaker for much of its robotic development and simulation. The company is exploring simulation pipelines for maritime robotics, using AWS services to test field use cases without the complexity, cost, and difficulty of performing identical live tests on the open water—something Barrera sees becoming increasingly crucial as the field evolves. “Using AWS RoboMaker for simulation enables us to use the latest and greatest research as it becomes available,” says Barrera. “That’s going to be a differentiator for us.” Using AWS RoboMaker could also pose new opportunities for Seafloor Systems’ work with partner universities and institutions. Students and researchers who don’t have access to USVs could instead learn by running simulations using the dimensions and software stacks of Seafloor Systems’ USVs.

On AWS, Seafloor Systems is prepared for the future of IoT in maritime robotics. Barrera imagines an IoT network in which the boats collect data and share it to a federated dataset in the cloud that everyone can access to learn about a location. Then, a global machine learning model could perform analytics at the edge using AWS IoT Greengrass.

Using AWS services, Seafloor Systems can not only build and deploy its robots more quickly but also facilitate innovation so that its robots can work for more use cases. For example, a New Zealand research institution used a HyDrone and computer vision to identify invasive aquatic plants in freshwater areas that couldn’t be seen with the naked eye. “Hydrographic scientists and businesses will always need our boats,” says Barrera. “But there’s this whole other world of pure robotics that benefits from being able to put sensors out in the world to collect data. There’s going to be so much advancement in various fields just because we were able to give them robots to do these things.”