Soft computing

Computation is often distributed. When we refer to computing, we often summon the image of the computer, where a central microprocessor does most of the computing while the rest of the computer obeys its orders. This is also often compared to the way our brain controls our body. Although this is useful metaphor, as is often the case, the reality is more complex. 

The reality is that to some degree, computing is often shared among various parts of your computer, or various organs of you body. A most classical example is if you touch a hot surface. Your body will act to remove the limb in danger before pain is registered by you brain. How does this happen ? Simply put, some computation happens between your hand nerve endings and your brain. This reflex motion reduces the reaction time to issue simple commands and move the body out of harm's way [1]. Some animals such as the octopus can also have very decentralized central nervous systems [2,3].

Computation is physical. Computation is sometimes seen as some ethereal process that happens in the heart of your computer. However, computation involves the motion of electrons in our computers, or ions in our brains. Although happening fast and at small scales, it is a very physical process. The first computers were mechanical pieces of machinery that would compute using gears and levers [4]. 

In classical robotics, a central computer, the controller is in charge of the robot control. The rest of the robot obeys its orders, and the body of the robot is made as rigid as possible so that there is as little difference between the controller "imagined" position of the robot and the robot's real world position. In soft robotics, because the robot's body is soft, a different dynamic may emerge. Indeed, the body's softness is also an opportunity to give the robot new abilities. This idea has been named "morphological computation" [5,7]. Whether this adaptation can really be called computation is still being argued about [6], however it is clear that soft adaptable bodies have the potential to adapt to their environment and to simplify robot control requirements.

Soft robots need computation. We already discussed in "Why soft robots" the applications covered by soft robotics. Whereas traditional robots are mostly made to operate in well defined environments such as factories, soft robots will most likely be tasked with operating in the real world. However, the real world is both cluttered and changing, and it makes it hard for robots to operate satisfactorily. The more robots operate in the real world, the more they need sensors, and the computation power that follows.