Since self-replicating machines could be considered a form of life, borrowing traits from living things is useful in developing functional self-replicating machines. One such trait of living things, in my opinion, has been given short shrift: the ability to influence internals and surroundings to better suit the needs of the organism.
In biology, it is often the organism that is best able to modify its environment that is able to survive and propagate the best. This ability is found in all living things, and is necessary for life. The first cells were successful when they gained a cell membrane that effectively walled them off from the exterior world, and allowed them to evolve their delicate internals with less interference from outside. Today, organisms control their own internals in a myriad of ways: temperature, pressure, pH, nutrients find their way automatically to where they are needed, waste is automatically removed, food is processed, the list goes on. Life’s complex processes are generally feeble and need a very specific environment to function properly. This means that organisms that influence the world around them to suit themselves can survive where others fail.
Translating this important trait to electromechanical self replicating machines is a challenge because of the unknown complexity of designing a self-replicating machine in the first place. It would be useless to describe the trait in all but the broadest generalities. When considering an imaginary ultimate machine, it would be defined as complete control over three dimensional space and all materials and energy within it. The ultimate machine is obviously not build-able, but it does give us a specific goal and trait to maximize in a self-replicating machine: versatility and control over itself and its immediate environment.
Rather than revel in the complexity of such an idea, it is easier to think of it simply as the ability to move any amount of matter (M) or energy (E) to point X, Y, Z in three dimensional space. Control of such a system becomes radically simplified, and would be easily controlled by modern computational methods. In fact, it starts to mirror modern computing, which revolves around moving and processing arbitrary data. For the reason that this system appears analogous to computer RAM (Random Access Memory), I call this automation system RASA, which stands for Random Access Spatial Automation.
Following the information technology lead, we can divide space into a three dimensional grid, and then we need to maximize the ability to move arbitrary lumps of stuff around the grid. We will need:
Arbitrary matter to move around: Like computing science, it makes sense to divide it up into pieces of standard size.
Something to move the matter around: Since the matter will be of uniform size, the machines that can move the matter around can also be uniform. Standardization is important.
An extensible medium to move through: Some kind of framework that allows the cargo machine to move freely in 3D space.
By dividing the self-replicating machines into these three categories, we can now turn this into an exercise of architecture for robotics. To begin, it is important to identify general desirable characteristics for each of the three categories:
The bare minimum physical construction needed to allow uniform machines to travel throughout on all three axes. To allow for easy addressing, a three dimensional grid of support struts and tracks is needed to allow a simple machine to travel throughout.
Since this framework will define the physical extents of the self-replicating machine, it will likely be the most massive part of the machine. Optimizing the design will require it to be of simple construction, of uniform standard size, and as inexpensive as possible. Ideally, it should be built of materials easily harvested from the local environment.
Since this is a self-replicating system, the grid must also be easily constructed by machine.
Must be able to easily travel throughout the grid in all three dimensions, and be able to carry an arbitrary cargo to any point in space within the grid.
Cargo haulers must have power, actuators, control systems, sensors and related systems to carry out their tasks of moving throughout the grid.
The cargo bin must be able to handle matter in the most arbitrary way possible.
Arbitrary Matter Specialization:
So far, we have been discussing the standardized parts of a self-replicating machine. Here is where we get into the specializations that make the system function and do what is needed for itself and for us. When it comes to specialization, again it can be useful to draw inspiration from living things. Certain parallels can be made with multicellular organisms. These can be considered as specialized cells that all do their thing for the benefit of the whole.
As this is a system that will be created by people at first, this can be seen as mirroring an economic system. As the system develops, many functions will at first be done by people. Much machinery will be built and serviced by humans, and tasks will be slowly replaced by innovative robots doing the same work. The profit motive has been incredibly effective in development of new and better machines, and self-replicating machines should not be left outside this process.
Functionality of what we can put inside the cargo haulers can be divided into four categories:
Cargo Level 0: Inert Cargo: Simply put, this is inactive stuff we can move anywhere it is wanted. Examples could include solids, liquids, gas, manufactured goods, waste, etc.
Cargo Level 1: Spatial Grid Machinery: This category would include any machinery involved with building, repairing and maintaining the spatial grid. Developing this machinery would give this system its first stage of being a self-replicating machine.
Cargo Level 2: Cargo Hauling Machinery: Any machinery that builds, repairs, or maintains the cargo haulers themselves. Construction of these machines would be the second stage of achieving self-replication of this system.
Cargo Level 3: Interactive Machinery: This is the category of machines that can interact with each other or the outside world. Examples could be many types of machines that could combine to form an ad-hoc assembly line, or machines that mine ores, or machines that provide goods or services to people, or machines that maintain and repair other machines at this level. Building these necessary machines would “close the loop” and make this a fully self-replicating machine.
A visualization of the basics of this system can be seen above in Figure 1. The coloured grid represents the spatial grid, and the grey cube represents the cargo hauling machine and the cargo or machine hidden within. The red posts are the structure of the grid, holding everything up. The cargo hauler can ride upon the green rails using retractable wheels. Also, retractable wheels can hang the box off the blue rails, allowing movement in the other direction, allowing for full two dimensional movement within the grid. Allowing the retractable wheels to slide up and down vertically on the box gives it the ability to climb up and down levels, making this a system able to access any random point within.
Now obviously this system has its physical limitations and overhead, but within the system, we have an infinitely configurable system of machinery. The obvious parallel of this architecture can be seen in the structure of Internet TCP/IP communications. In the same way we have wires, data packets, and the data carried within the packets, we have a grid framework, cargo haulers, and the cargo and diverse machinery within.
As I see it, the first obvious use of this system is storage and warehousing. This system can easily be set up as an automated warehouse. As far as the first step towards self-replication, I anticipate a machine inside the cargo bay of the box able to construct new segments of the grid by itself. These uses could be profitable for some companies, and I foresee the profit motive quickly driving many innovations once it is released upon the world.