The nexus of AI and robotics may soon be one of the most important technological arena’s in the world; as it offers the most promise for solving many of society’s current shortcomings. Many may not realize that the field is over six decades old. The world’s first AI research group was established at MIT in 1959. The Computer Science and AI Lab (CSAIL) is now the largest laboratory at MIT.

This allocation of resources at the world’s most renowned technical institution is no coincidence; as AI and robotics is now the flagship arena for cutting edge R&D.

When it comes to “smart” robots, the primary means of machine learning is “reinforcement learning” (RL). RL is a way of helping a computer learn through experience, enabling it to make a series of decisions that yield desired outcomes. This works even when the computer is not equipped with any prior knowledge of how its actions will affect its immediate environment.

The key here is that RL can be married with other approaches to machine learning. For example, it could make use of on-line educational systems so as to ascertain the effectiveness of alternative teaching methods. At the end of the day, the aim is to procure the right kind of data necessary to develop systems that make good decisions. At MIT, the primary focus is developing RL algorithms (and concomitant statistical techniques) to give computers the ability to develop good suggestions; and to do so without needing to make use of massive amounts of data (which can be difficult and very time-consuming to collect).

That said, we should not leave the entire process to the computers. Crucial is “human-in-the-loop” reinforcement-learning, which can accelerate the process and enhance the capacity to discern the most prudent course. Human-computer collaboration allows algorithms to “reason” about their own limited performance; and reach out to humans for help whenever the occasion warrants.

One of the primary purposes of robots is the construction of various objects–from the ultra-small (micro-tech)…through the modest sized objects (voxels)…to the large (bridges, spacecraft, etc.) Usually, the pieces are built at separate facilities, shipped via massive cargo planes / ships, and taken to the location where a final assembly will take place. This old-style manufacturing process can be very time-consuming and overly expensive. This is where assembler robots can be extremely beneficial.

When it comes to larger structures, robotic arms can pull themselves across a structure by effectively viewing themselves as part of the structure (rather than as separate objects that they are helping to build). This is called “relative robotics”.

Another perk of the newest generation of assembler robots is they don’t require expensive navigation systems to find their way around the work-space. “Relative robots” move in relation to the pieces they are building. As moderate-sized structures (i.e. voxels) are assembled, the robots can adjust their positioning accordingly–modifying their navigation-space relative to the newest version of the structure. This makes it quicker (read: less expensive) to repair what the robots are working on. (The smallest robots can move around like inch-worms, clamping onto the voxels by opening and closing their V-shaped bodies, counting their steps as they go.) This makes it easier to go back and fix manufacturing errors as they proceed. Thus the precision is effectively built into the structure, not the robots. Each robot just needs to know where its next step is.

Conventionally, manufacturing robots are separated into two categories: those made of expensive, custom parts (designed to complete very specific tasks; like the factory assemblage of cars) and those made of inexpensive, mass-produced parts that don’t perform as well. The new kind of assembler robots move beyond these limited categories. They are simpler than traditional factory bots; and are much more modular and capable…while still boasting the capacity to build massive structures like airplanes or bridges.

When it comes to space stations and extra-terrestrial habitats, the robots would essentially live on the structure, undertaking continuous maintenance and repairs. In such circumstances, the robots would live on the structure. Software allows groups of the assembler robots to coordinate their tasks. NASA and Airbus SE, both sponsor R&D in this exciting new area. Meanwhile, those at MIT and other labs are striving to perfect RL technology. Stay tuned for further developments.