Evolutionary locomotion design

Researchers from Boston Dynamics claim to have built 'the most advanced quadruped robot on earth'. While usually I would not be very interested in robots that basically emulate legs, this one is particularly striking, almost walking in a natural, biological way. It can walk at 5 mph, climb 35 degree slopes, and keeps itself balanced even when forcefully kicked or when traversing ice. It is actuated hydraulically, and has a system to reuse energy that would normally be lost in each step.

Generally, this is a nice move towards intelligent motor behaviour, but still it is a kind of backwards emulatory approach to technology design. Reverse engineering systems we find in nature just won't give technology its own intelligence, its own character and its own elegance. Technology needs to develop this from within itself, by self-reflection and gradual adaptation during its entire existence.

The silliness of this approach is almost caricaturally illustrated by a movie of another quadruped robot, Scott II, developed by McGill University:



It is so myopic to develop technology being inspired by things like the jolly tred of your neighbour's puppy teckel. Is this what people need? It seems that many robotics engineers are just too isolated, and a common framework is lacking to develop technology in a more holistic, deeply human way.

I'd say that some virtual technologies can conceptually score way higher on the mind-blowingness scale. Researchers in Oxford, Britain, developed a system to make virtual characters dynamically respond to their environment. No more inverse kinematics, no more mocap; these are clumsy, static, externally imposed and simply unnatural approaches towards the design of intelligently behaving systems. What these guys did was using genetic algorithms to have a virtual character learn it's own way of behaving, depending on the interaction of its own specific body with its own specific environment. 'Dynamic Motion Synthesis' is what they call this process of simulating a motor nervous system and physical body.

The nice thing is that this latter approach does not simply result in blindly evolving systems; the human designer can still specify his desires to some extent, like an end pose of a movement, or the goal of a movement. This is very elegant I think, and can be an inspirational approach to the design of physical products as well. You give the entity room to develop itself, but also provide some overall vision, an over-encompassing framework. So you balance a centripetal and a centrifugal force, which is very human I think. Natural evolution by itself is no holistic solution; it results in loose ends everywhere and a lot of uselessness. It is tedious, messy, and slow. A detached, artificial approach to technology design is opposite to evolution, and does not work either, as it is myopic and does not respect individuals. It dictatorially tries to frame in the world, inviting people to shrink and abstract their lifeworld. These two opposite approaches towards how entities in the world are created results in seeming dichotomies, and a lot of tensions in the world as a whole. We need a new approach that transcends this, stemming from the perspective that all is one and one is all. We are animals, we are technology, and we are the air that surrounds us.

My view is that future technology design needs an entirely different approach, merging these two opposites into one that transcends these. We need to give the artefacts we create an initial potential for positively changing our world, but also the openness and ability to develop themselves from interacting with their world, hence developing individually appropriate behaviour. But we need to constantly monitor and steer all individual development processes, and guide them like a friendly father, informing them, empathizing with them, and gently but convincingly and directly steer them as well. Like we should raise our own children.

My future is one where the biological merges with the technological, and everything feels socially connected as in a global family of highly interactive and morphologically highly diverse entities.

More info at:
Boston Dynamics
Natural Motion