Automating Everything - Self-Repair – The Ghost in the Machine - Part 4

Many basic repairs of large appliances and other electronic equipment can be accomplished by incorporating remote cameras, manipulators, adjustable tools and fixed tracks in the object itself. When a problem arises, normal software diagnostics can be combined with these internal mechanisms. Each device will run its routine examination₁, looking for obvious, predefined problems before a human technician on the other end surveys the situation₂. Manipulator hands and tools₃ can pull parts loose₄ (whether directed by software or a remote technician), cameras can give the tech a 3d view and the results of programmed diagnostics and whatever issues the owner reports can put all of this in perspective. The system will even be able to show side-by-side₅ or superimposed images₆ to compare the factory-floor state of the appliance to its just-installed condition to its present status. The solid image of its present configuration can be overlaid with a false color tracery of its correct blueprints, or with translucent, false-color images showing how it looked in the past, thereby highlighting differences and hence, problems. As the software develops a larger library of reference images and the cost of video recording drops, the cameras could even start periodically recording performance of large moving parts and use deviations to assist in internal diagnostics as well as in helping technicians get to the root of the problem. For example, clear signs of excessive vibration can be used to signal a problem that the appliance’s diagnostic software can bring up as a warning, much as a washer or dryer can use other sensors to detect large scale vibration in existing diagnostics.

Depending on the nature of the service plan an appliance has, these updates may even happen automatically, with software and service technicians checking for issues without requiring active authorization to look at the internal performance of the product. If they see an issue emerging, the owner may be notified with an email, text or automated call₇, along with a proposed solution. To re-level a dryer, for example, or to replace a faulty part before it goes out completely. The mini-robots₈ moving around on their internal tracks₉ will be able to examine a machine’s operations in real time from many vantage points (aided, in some models, by additional fixed internal sensors), without ever having to bother the owner unless there is a need. Their tracks will be able hold the sliding trolley of the moving robot securely in that groove₁₀, much like the track for a sliding van door, but will be laid out to allow the robot actuators to move into any needed position as flanking rubber wheels (driven by an electric motor) propel the robot on its platform as needed.

Many basic repairs of large appliances and other electronic equipment can be accomplished by incorporating remote cameras, manipulators, adjustable tools and fixed tracks in the object itself. When a problem arises, normal software diagnostics can be combined with these internal mechanisms. Each device will run its routine examination₁, looking for obvious, predefined problems before a human technician on the other end surveys the situation₂. Manipulator hands and tools₃ can pull parts loose₄ (whether directed by software or a remote technician), cameras can give the tech a 3d view and the results of programmed diagnostics and whatever issues the owner reports can put all of this in perspective. The system will even be able to show side-by-side₅ or superimposed images₆ to compare the factory-floor state of the appliance to its just-installed condition to its present status. The solid image of its present configuration can be overlaid with a false color tracery of its correct blueprints, or with translucent, false-color images showing how it looked in the past, thereby highlighting differences and hence, problems. As the software develops a larger library of reference images and the cost of video recording drops, the cameras could even start periodically recording performance of large moving parts and use deviations to assist in internal diagnostics as well as in helping technicians get to the root of the problem. For example, clear signs of excessive vibration can be used to signal a problem that the appliance’s diagnostic software can bring up as a warning, much as a washer or dryer can use other sensors to detect large scale vibration in existing diagnostics.

The end-effectors of these machines, which are basically very small industrial robots₁₁, can include more general gripping hands or use any number of powered tools such as electric screwdrivers or ratchets. A single robot would have a general purpose grip₁₂ while unfolding other tools₁₃ as necessary and then folding them away like a Swiss army knife. Alternatively, if any major tool proved too large or awkward to integrate with others, it would be handled by a mini-robot on its own platform₁₄, and be moved into position on the track (while moving other robots out of the way) as needed. The system would not need every possible tool, but only the ones that could be practically employed inside the machine being repaired. So if an electric ratchet₁₅ were included, for example, it would only need as many heads as there were sizes of bolts to be loosened or tightened. If a larger array of ratchet heads were needed, a covered storage alcove₁₆ could conceivably hold them and the actuator could select the head needed and replace it in the alcove when it was no longer needed. In addition to mechanical precision, magnetic heads might be useful in this circumstance, assuming there were nothing in the vicinity which might be disturbed by small magnetic fields (such as a microcomputer’s memory).

Unlike many industrial robots, which may reach out a considerable distance from a fixed position, or a CNC, which has free range over the area in which it is cutting away material or otherwise operating, these tracks will move the internal mini-robots into the preset, ideal position to perform a task, and will engage their preprogrammed actions once locked into place using a simple brake₁₇ for smaller systems generating less force and vibration that could knock them out of position. More powerful systems (especially external repair systems in repair garages and other specialized facilities) can use an optional electromagnet₁₈ to extend to and adhere to a ferrous metal housing (such as the track itself) if it will not disrupt any associated electronics (as it magnetizes the metal it is anchoring to) or they can literally lock themselves in place by dropping a smooth shaft or bolt₁₉ into a pre-cut, matching hole₂₀ (in this instance, one cut into the track itself). With the track preventing the robot from being lifted away and the shaft halting any horizontal motion, the robot will be effectively immobilized in any preset position with a prepared hole for the braking bolt. If the robot is positioned between these prepared spots, however, it will have to depend on braking its wheels and using friction or magnetism to hold it in place. If a technician is overriding these functions, they will have the option of changing locations to somewhere other than the standard, prescribed position, but with those caveats.

 

Simply moving, for example, an electric screwdriver opposite of a screw that needs to be tightened will begin by moving the actuator down an internal track until it is in the best position to access that screw. If automated, the robot will likely be following a pre-programmed set of actions. Once it moves into position and locks in place, the basic actuators will move the screwdriver in place over the screw, align the screwdriver tip with the slot of the screw, push them firmly together, and then steadily tighten the screw in place₂₁. A simple bit of upkeep which might require the owner or a technician to open up the appliance completely can be taken care of without bothering anyone at all, save for whatever human oversight is remotely authorizing the action. If the internal robot is actually required to follow a track up the wall of an appliance or other machine, then the entire track will be flanked by metal strips₂₂ perforated by a continuous series of rectangular holes₂₃. The flanking, hard-rubber wheels₂₄ driving the robot down its track will, in this case, be studded with triangular “teeth”₂₅ able to grip these perforations and thus drive the robot vertically as needed. Whenever a track and its flanking metal strips need to go up a wall, they will arc upward, allowing a smooth transition between surfaces.

 

A fully functional if small arm₂₆ such as the MEC500, the uArm or preferably smaller robots may not be necessary for most tasks, but if small, bare-bones, specialized actuators are insufficient for more involved tasks in more complicated machines, a more flexible arm₂₇ can be kept in reserve for more complex challenges. Even a full robotic hand could simply be folded up in a tight “fist” and only unfolded as needed, for machines complex and valuable enough for it to be practical to have that level of adaptability in their internal repair system. A scaled-down version of existing hands and/or arms such as the 3D printed prosthetic hands from YouBionic could be adapted to this use and would mesh well with the haptic gloves used by overseers.

 

If the issue is something basic, the human technical support associated with the software may be able to mail the owner a part which they will put into a small panel₁. The manipulators₂ will pick that item up₃ and shuttle it down their tracks₄ to the correct area of the machine, only to replace the old part₅ and send it back to the panel₆ for removal₇. If it is something the owner should handle (say checking the fuse box, or replacing a water filter) the system will notify them accordingly₈, and even provide graphics showing how to handle common issues₉ (which can be displayed on any available screens, sent to a mobile device or email, or transmitted as a link, as convenient). If the system failure is something more complex, a human technician₁₀ (or remotely overseen robotic one₁₁) can be dispatched. But in that case, the initial diagnosis will already be complete, and the demand on the service company for live in-home service will be that much less, reducing the wait time in getting a repair technician’s personal assistance.

This entire augmented diagnostic/repair subsystem will be on its own separate power conduit, so when the owner does not want this capacity turned on it will be shut down by a manual switch that cuts its power, which flows on a separate circuit inside the machine being monitored and repaired. Because this internal system requires manual activation in order to function, it can not be hacked while shut down. When the owner wants anything more than standard internal software diagnostics (such as appliance error codes), they can flip that switch and ask for help.

If a repair proves beyond the capacity of internal self-repair robots, humanoid robots under remote oversight can be dispatched₁, for example if an appliance needs to be physically disassembled to replace a large part or if an error in the original installation needs to be corrected. By supplementing with humanoid robots, highly skilled technicians are able to visit far more customers in one day, and can access their peers and mentors at their remote office when faced with an unusual product, repair or situation. Further, two robots can be periodically be deployed during training or a probationary period in which a new employee’s skill is being assessed, allowing both faster service (at least when an extra set of hands would be helpful instead of just getting in the way) and more in-the-field training.

When dealing with larger pieces of equipment, a similar, external version of the same system is possible. For example, a government agency with many vehicles and a more limited number of mechanics and electricians might want to be able to repair that equipment more rapidly. A simple (if more robust) set of tracks could be set down₁, and a mix of automated robots₂ and human-directed machines₃ could take charge of those repairs, checking parts and connections₄, testing performance₅, authorizing the order of replacements₆ not on site, etc. Automated robots moving in set patterns around a temporarily fixed object to complete specific rote tasks with tremendous efficiency is the default way robots are used on assembly lines. Applied to maintenance, many tasks can be completed with minimal oversight, especially on machines that have not suffered damage that radically alters their basic shape (such as a jeep partially crumpled from a wreck or an explosion). Software aside, the purely physical actions of clamping on to a tire, removing bolts, pulling off the tire, putting on a replacement and bolting it into place are well within the strictly physical capabilities of robots of a design which likely put those bolts and tire on in the first place. There are, of course, still issues with novel object recognition and manipulation for most robots and any fully automated motions will have to be programmed in the first place. But the make and model of a fixed array of vehicles in an organization’s fleet – be it a military base, local government, or auto dealership – means the mechanics will know what they have just rolled into the system and will be able to pick that option off a menu, assuming the specific vehicle does not already have its own file. Given that the repair units will have access to the basic schematics (or at least those external and engine schematics they actually use in their repairs, if not every modular element which might be a trade secret or classified technology), they will be able to identify what is in front of them that much more readily. When certain specifics elude them, such as grasping a particularly odd or heavily damaged part, human overseers₇ can again take over with haptic gloves₈ to handle that step of the operation. More importantly, with any computer network processing this human-aided repair system, the software can record the human manipulations used to handle the problem₉, and add them to any learning database they may have₁₀, until they can master the basic maneuver used, and eventually learn to apply it in similar situations – or, in the case of more sophisticated programs, in novel situations. Learning pendants and lead-by-the-nose techniques are already used to modify the routines of industrial robots.

Where industrial robots on fixed tracks fall short, humanoid robots could step in₁₁ under remote control to lend further assistance. Again, if a machine had to be disassembled by hand, for example, or needed custom repairs which also had to performed by hand, the robots could provide the hands and the technicians, in their haptic gloves, could provide the skill. Because the remote operators can work in shifts, these robotic repairs can potentially go on 24-hours a day, enabling a basic set up to provide a tremendous amount of service to areas with limited on-site personnel.

If secure, high-speed connections can be set up, even parts which might require highly specialized skills and technology and very high security clearance to repair might be reparable to a sufficiently thorough version of this system. For example, some “black box” pieces of technology whose internal workings are beyond the skills of most personnel, on-site or otherwise, and whose critical functions may actually be a trade secret, might still be diagnosed and mended by the organization which produced it in the first place. A unique sensor, for example, might be a complete mystery to technicians who normally just replace it if it fails. But with 3D printing, software controlled CNC microwelding and other resources, including remotely controlled humanoid and industrial robots, the supplying company might be able to return such a device to service, at least until a replacement is available. Providing such service remotely could be included as part of the contract in some sales. Ultimately, a government or other organization might end up leasing a technology design rather than buying the product in discrete units, with highly adaptable CNCs, 3D printers, RepRaps and other robots constructing and repairing whatever units are needed from the basic parts and materials on hand.

With all versions of remote oversight, control of a particular model can be limited by a number of safeguards, though if highly secure access limited to only one or two sites is required, both the controlled and controlling units can be outfitted in advance with a constantly changing encryption key required for communication and control. Hence, decrypting a part of the key used only moments before would be meaningless as it would never be used again.

Part 1

Part 3

Part 5