Deep-sea engineers are plagued by a specific type of failure known as Nereus. The $14 million robot, which was constructed at Woods Hole and is regarded as one of the most capable underwater vehicles ever created, vanished in the Kermadec Trench in 2014 at a depth of about 9,990 meters. Down six and a half miles. It was probably crushed by the pressure. Almost anything you put into it will be affected by 16,000 pounds per square inch.
You can still sense the impact of that loss in the way MIT engineers describe their work when you speak with them today. It’s not overly dramatic. It’s just wary. The kind of caution you acquire when a major project in a field comes to an abrupt end.
The current work appears surprisingly modest, primarily through the MIT–WHOI joint program. motorcycle-sized robots. Strangely, hulls resemble hoagie sandwiches. Cost: less than $200,000 per unit, which is more of a sketch than a luxury in the deep ocean. The most noticeable example is the Orpheus class, which was created in partnership with NASA’s Jet Propulsion Laboratory. It is small, untethered, and intended to function in the hadal zone, which is the band of trenches between 6,000 and 11,000 meters where people have hardly ever lived.
The similarities to MIT’s flight labs are difficult to ignore. The same line is frequently drawn by academics like Nicholas Roy, who teaches autonomous vehicle design: “A robot on Mars, a robot on the seafloor—both operate somewhere no one can rescue them.” Perception, planning, and the type of fault-tolerant control that enables a machine to recover when something unavoidably goes wrong are all necessary for both. Roy has repeatedly stated that neither industry nor research have found a solution to this issue. That admission reveals an engineer’s unease about how much is still unknown.
Three broad categories describe what MIT teams are searching for. First, life: weird, pressure-adapted organisms that might not look like anything on the surface and that scientists think could be similar to what might be found on Enceladus or Europa. According to Tim Shank of WHOI, pressures at hadal depths are similar to what would be expected at the bottom of the hypothetical ocean on Europa. In this way, the hadal zone is a stage for rehearsal. Second, geology: unmapped and jagged terrain that has not been captured in useful detail by any satellite. The third, and perhaps most subtly significant, behavior is the movement of hydrothermal vent plumes and the carbon cycle between the deep and the surface.
As the field develops, it seems like no single robot will be able to solve any of these problems. Swarms are the more recent goal. Several vehicles, communicating within a limited acoustic bandwidth, creating fragmented 3D maps. This is why the autonomy research is so important. At 9,000 meters, bandwidth is cruel. There must be a robot down there that is actually its own scientist.
Flying quadrotors through obstacle courses as part of a new autonomy capstone, MIT students are working on what may appear to be a different problem in the high bay at the Kresa Center. Really, it isn’t. The questions remain the same. How can you trust an unreachable machine? How can software created by a group of people who are only able to test in bits and pieces be trusted?
Deep-ocean technology investors appear to think the industry is finally changing course. Smaller budgets, more intelligent software, and less expensive hulls. Whether any of this will scale before another well-known car is lost is still up in the air. However, the engineers do not sound impatient when they describe this work. If anything, they sound a little patient, which might be the ideal temperament for the hadal zone.
