"The sea has never been friendly to man . . ." wrote Samuel Coleridge in The Rhyme of the Ancient Mariner, and this has not changed since Coleridge's time.

In the submerged environment our ocean offers cold, pressure, darkness, corrosion, detritus, and nothing to breathe. It provides frustration in abundance. Further, whether human or machine, the underwater worker toils in isolation. This isolation becomes progressively more pronounced as projects move deeper.

In the contest between ocean and facility construction, offshore oil typically chooses any of five modern tools to reach and perform submerged work: the ambient pressure diver, atmospheric diving suit (ADS), remotely-operated vehicle (ROV), manned submersible or submarine, and autonomous underwater vehicle (AUV). Ambient pressure diving has the longest history, while AUV history is so short as to remain mostly promise.

The goal with each of these five options is the same: do something useful at the work site that makes progress toward fulfilling the ultimate intentions of the project.

A Work-Flow Model
Borrowing a model from Feedback Control Theory, we could illustrate the process with the model depicted in the figure. In the accompanying Underwater Work-Flow Model illustration, the box labeled "actuator" is the point were progress occurs. But we cannot "do something" until we know the work site status, and it is the "sensor systems" that start the train of signals telling us where things are.

In general, sensors convert one form of energy to another (e.g., light to electrical current in a video camera). The converted signal travels through some transmission medium to a display interface, which processes the signal and provides some sort of representation for operator interpretation. That interpretation, compared to our intent, produces an action command that should, if properly executed, move the work site condition toward a desired result. A command interface sends a signal back through the transmission medium to the actuator. The actuator, following the received command signal, then influences conditions at the work site. We intend that the actuator's response moves the work site toward the desired outcome, or "intent." The sensors detect the progress made toward the intent and transmit that information back again to the display for comparison with our intent ("interpretation") and further action commands.

This little model is a way to think about the various modes we choose for underwater work. The idea is that we can use the model to identify aspects that will require radical, rather than evolutionary, advances along the road. Let us begin with the oldest method.

Ambient Pressure Diver
The human being is the best sensor package available. It is a myth that we have only five senses (sight, hearing, touch, taste, and smell). We have an array of senses that are not included in the common five. Proprioceptors tell us, among other things, where our limbs and joints are. We can detect which way is down. We sense acceleration, turning, tiny movements, hot, cold, pressure, and vibration-a plethora of things not included in the simple five senses. To date, our best efforts at replicating the complex human sensory package in a machine have fallen far short.

In ambient pressure diving, the transmission medium between the diver and his senses is eliminated, along with any display interface. While low-to-zero underwater visibility can generate insurmountable problems, other senses remain extremely accurate. Further, the direct nerve transmission to the diver's brain eliminates several opportunities for inadequate or corrupted information.

With all of this information so effectively presented, the interpretation phase is rapid, natural, and automatic. A command interface is also unnecessary. Limbs and hands respond without cognitive dissonance. Then, the senses continue providing immediate and unimpeded feedback, keeping the cycle going.

In typical commercial umbilical diving, there is a second loop that uses the diver as an actuator and a sensor package. The diver converts what his senses tell him to the more limited medium of speech. Significant information is lost in this conversion. Then, the speech travels through a noisy helmet environment, converts to an electronic signal, and travels up an umbilical to the surface. The umbilical wire, if properly maintained, offers little transmission distortion in comparison to the noisy helmet environment and the microphone response. On the surface, a diver radio (the display interface) converts the electronic signal back into audio.

At that point, a supervisor may have cause to alter the intent or relay instructions to other team members who can aid the diver in completing the task. For any instructions issued to the diver, the command interface is the same diver radio. The signal transmits back down the same reliable umbilical transmission medium to speakers installed next to the diver's ears.

Ideally, the diver and support crew are skilled and knowledgeable, requiring only limited exchange of status information to coordinate construction operations. They do not burden this second loop with detailed action instructions.

What is there not to like about the diver as an intervention tool? Except for some possible glitches in the link between topside personnel and the working diver, this seems to be an excellent approach. Indeed, most negatives associated with ambient pressure human diving are not performance-based as illustrated in our model.

Instead, the negatives are health risks, liability, and physiological limitations. The plain truth is that diving is hazardous to the diver's health and life. This adds cost, and decompression requires including long non-productive time periods. Further, work sites are moving beyond depths where ambient pressure human divers can be used.

Remotely-Operated Vehicles (ROVs)
A work-class ROV is a robot on a tether. It is "telepresence," so to speak, where the operator resides far above on a ship or platform, and the actuator is a machine performing the requested tasks as a stand-in for the absent human. ROVs reach the deepest working depths, and there is every reason to believe that steady evolution will provide technical solutions that will allow them to work in even the deepest ocean trenches. ROVs abolish depth limitations imposed by human physiology.

So, what is there not to like about the ROV as an intervention tool? When we look at the sensor system, we find it sorely inferior to a human at the site. We are stuck with a limited sensor set. No machine tells its operator where all of its limbs are. Nor is it easy for a machine to portray where it is looking, which way is down, what kind of tug it feels on its tether, or a whole host of senses that humans instinctively enjoy.

Further, the ROV does not deliver its limited set of sensor data in an instinctive or coordinated manner. That is, a diver is not even consciously thinking about the dive umbilical; yet, any unusual change in tension or direction on the umbilical innately pushes umbilical awareness to the forefront. Likewise, positional proprioceptors coordinate efficiently with what a diver sees to provide immediate and smoothly interpreted orientation information. And we have yet to discuss the comparison between hand dexterity and manipulators.

So this much more limited and poorly coordinated stream of sensor data goes traveling up a long tether to the surface. A sound tether does not introduce too many transmission limitations, but at great depths it is becoming an expensive and high maintenance item- not to mention being an impediment to some actions like traveling underneath or into structures. The display interfaces, be they video display screens, dials, beepers, or lights, are just excessively two-dimensional. But, then, that is why we like skilled and experienced ROV operators who have learned how to interpret various displays.

Once our ROV operator interprets the sensor displays, he sends control signals back down to the vehicle's manipulator arms. All the while, of course, he may have to consciously work out proper ROV orientation and formulate control commands that actually work-something he could do intuitively, without a thought, if he, instead of the ROV, were actually at the work site. Tasks that are simple for a diver may become tedious and complex to accomplish through an ROV.

Responding to received control signals, the ROV acts. Results follow the ROV actions, and the sensor system sends back results to the display interfaces for continuous operator interpretation.

What's not to like in the ROV system? Inferior sensor information and clumsy manipulation, that is what. That little tether has caused a few problems as well, being so costly and in the way.

Machines can compensate for sensory and dexterity shortcomings. After all, even in the more sedate land-based environment, we do not excavate or haul heavy loads by manpower. Instead, we use backhoes and forklifts.

ROVs do bring a great compensation to the work environment: power. A person's best performance will consistently be under 0.5 horsepower (370 watts) compared to between 150Hp and 300Hp (110 to 220 kilowatts) for a work-class ROV. A human's maximum force, usually around 200-300 pounds (890-1300 Newtons), is also a fraction of the force available from an ROV. All this power is at hand partially because that annoying tether frees an ROV from carrying an onboard power source.

If we were to dream of the perfect ROV, what would it look like? Perhaps the experience for the operator would be a kind of virtual reality. Maybe ensconced in some box on the surface, he would see through cameras that turn on the machine when his own head turns. If he moves his arm, the ROV would move its arm-complete with force feedback so he would know when the ROV arm meets resistance. If the ROV rotated, the operator would rotate. Ideally, even if the ROV vibrated, the operator would vibrate. Talk about a dream world. This would require a technological revolution, not an evolution.

Atmospheric Diving Suits
Okay, let's put the operator back at the work site. We can do that with an atmospheric diving suit (ADS). An ADS encloses a man in normal, one-atmosphere pressure while allowing him to operate underwater. No compression means no decompression schedules and no hyperbaric-induced health effects. Depth limits also extend beyond human physiological limits, but suit joints restrict the ADS to shallower depths than ROVs. Presumably, evolutionary technological advances will allow the ADS to creep into deeper depths, but not rapidly enough to overtake maximum ROV operating depths.

Like an ROV, an ADS is on a tether to provide launch and recovery, communications, and power. Thrusters, powered through the tether, allow the otherwise too-heavy-to-swim system mid-water operational capability.

With our human operator back on the work site, the human sensor package is back. The dexterity of an ADS does attenuate the human sense effectiveness. However, now the operator knows exactly where his limbs are. His innate sense of direction and relative positions effortlessly convey direction of vision and which way is up. If there is vibration or turbulence, the operator has no difficulty detecting it. Further, all these sensory inputs are again instinctive and coordinated. ADS suit dexterity, while inferior to direct human hands, is reportedly superior to ROV manipulator arm dexterity.

By placing the operator back on the bottom, we have also eliminated transmission of primary operator data through a medium and onto dials and display screens. Sensor data interpretation is back to biologically natural rendering. Though attenuated from an ambient pressure diver, the sensor data is of better quality than that available to an ROV operator, not least of which is much improved depth perception.

The ADS pilot is now equipped with an interpretation of the work site condition. Just like the ambient diver, the pilot's responses are innate, although somewhat clumsier and more frustrating than a diver's. Orientation, limb positions, and force feedback are natural and interpreted in instinctive fashion. Sensory feedback is also back to being mostly biologically natural.

Another nice feature is that ADS systems require much less deck space and weigh less than an ROV spread (or saturation system, for that matter). Some are even helicopter-transportable.

What is there not to like in the ADS system? Manipulators are a bit awkward, so it helps if designers pre-plan for ADS use (also true for ROV intervention). The tether is still around, and the ADS does not have the power of an ROV to compensate for clumsiness and an inability to squeeze into places that a diver could go.

We are also back to putting a human in harm's way, though we have gained the advantage of not subjecting the pilot to the long-term hyperbaric health effects. However, the risk is still greater than letting an ROV go into the deeps. ADS joint design, too, just cannot keep up with the depths to which an ROV can travel.

The perfect ADS suit would be swimmable. The pilot could also easily detach it safely from the tether, and it would be barely bigger than the human inside. The ADS would also approach the same joint flexibility that a person enjoys. The manipulators would become hands... we can dream, can't we?

Subs as Intervention Tools
Want to keep that wonderful human at the work site? Drop the tether? Increase available power? Increase space for a few desirable things (like redundant life support)? How about putting a whole team of people, instead of one lonely guy, at the site? Then a submersible or submarine is your machine.

Submersibles are an old tool. The 1960s and 1970s saw many operational submersible applications. They dominated the deep ocean interventions until ROVs supplanted them, and very few were built in the 1990s. With new configurations, both large and small, we are now seeing some revival of subs.

For the purposes of this article, a submersible is a small vehicle that relies on a mother ship for transport from ports. A submarine is a larger, but similar, vessel that does not require a mother ship, but instead transits from port to the general work area on its own. We could view an ADS suit as a really tiny submersible, or we could stick ADS-like articulated arms on a small submersible and call it a really big suit (such as Nuytco's DeepWorker). Unfortunately the larger size makes the manipulators awkward because the sub's bulk prevents getting close enough for their limited length. Instead, typical submersibles with manipulation capability have the same mechanical arms that are mounted on ROVs.

In our submersible, the operator remains at the work site. The smallest submersibles provide the operator with just a bit more space than an ADS. The largest provide space for several crew members, and a submarine might even provide space for crew to live-including sleeping, eating, and toiletries. We retain all the advantages of no hyperbaric exposure. Submersibles also travel to the same depths as ROVs. Significantly, also, we have dropped that pesky tether.

In a sub, we find that the sensor systems are a hybrid of human and ROV-like machine sensors. Because of its size, a submersible pushes the human back away from the work site. Fortunately, the transmission distances are not great, and human senses supplement the information provided by electromechanical sensors. We hang onto some positional, sound, and vibration information. With advanced observation port design we also get good visual depth perception, but sacrifice a lot by getting pushed so far back from the actual work by the greater submersible bulk.

With the ADS arms replaced by mechanical manipulators, we have also lost those nice proprioceptor senses and have to rely on what we see for arm position information. Thus, we retain some of the human sensor package, but have to rely more heavily on an ROV-like set of sensors and visual cues instead of touch and limb position senses.

Sensor transmission is either direct (sight through a porthole) or travels short distances down conductors. This is generally reliable. Interpretation is a mixture of direct natural senses and artificial display system. The artificial systems dominate more where visibility degrades. We can place the sensor data, transmission, and interpretation at somewhere between those of an ADS and an ROV. The exact placement depends on the submersible's size and the water clarity.

Commands to actuators are very similar to ROV action. Sensed feedback is considerably better than in an ROV, but remains, again, between an ADS and ROV and influenced by submersible size and water clarity.

What is there not to like about subs? Submersibles require very large support vessels, and while they can drop the tether, they do so at a great sacrifice to power. Without power, how does one compensate for being so much less dexterous and more blind? Submarines, because of their size and the subsequent room for a power plant, retain the power. This size, though, limits their access to detailed work sites.

What dreams should we have for subs? The perfect small submersible would deliver horsepower adequate to function like a true piece of heavy equipment. Presently, the smaller submersibles have much less power than the skid steer loaders you might see pushing dirt around on residential construction sites. It would also be nice for submersibles to take less deck space and a smaller mother ship.

Our dream submarine would not only have independence from a mother ship and plenty of power, but would have a way to reach into the small spaces without its size becoming a risk to the job site. Perhaps marrying submarines and ROVs would produce an ideal approach by providing submarine-launched ROVs that would be unaffected by surface weather conditions while getting the operator close enough to occasionally peer out at his actuator.