Semantic robot control and biomimicry

Recently, there have been a couple of amazing examples of robots attaining some amazing low-level movement and navigation autonomy.  Here are a couple of examples.  First, a quad-copter that can perform some startling complex high speed navigation stunts:

Second, a beetle-like walker that can easily navigate rough survaces:

Progress on robots has been pretty steady for years, so what is interesting about these particular examples?

The ability to direct a robot’s higher level movements with reliable outcomes.

That is, a shift from directing robots with syntax (manipulating control surfaces, motor speeds, servo positions) to directing with semantics (go to the living room).  Robots like those shown above will soon be able to take commands at a high level and carry them out reliably and safely.

Broad access to robot servants

We have reached a point in sophistication and cost where small, autonomous robots can be controlled with higher level commands.  To learn to fly a quad-copter via remote control of throttle and pitch in a way that can navigate a window with 3 inches of clearance takes hours of practice and crashed copters. As shown above, now we can merely command that it be done.  The cost of these devices has reached a few hundred dollars each and will continue to fall quickly.

Swarms of robots (“many hands make light work”)

These developments mean that small robots can be directed to “got into the living room” or “fly across the park without hitting anything and find little Bobby.”

The possibility of cheap, small autonomous robots performing tasks such as searching for hurricane survivors, mapping oil spills, collecting anti-personnel mines, carrying out surveillance, assisting you in the garage or mowing your lawn is now very real.

With the ability to direct them at a high level, the opportunity to carry out large tasks with many small specialized robots is in reach. With our current knowledge of collective behaviors, swarms of robots could be called on to perform large scale work requiring collaboration and coordination.

Bio-mimicry and fear

There is something unsettling about both of these videos.  They both demonstrate movements that are life-like, and often associated with animals we’d rather not have too big or too close.  Both robots execute robust recovery and control actions much like beetles or house flies. I postulate that this bio-like attribute is core to the technical success demonstrated here.  But it is also clear that this attribute may be more a than a little distasteful to people.

Will we adapt to the creepiness and have a swarm of crawlers clean the house? Or is this technology DOA because it violates some survival instinct?

Another concern is the new, cheap capabilities to do harm that come with any technological advance. Will ethics and countermeasures evolve fast enough to head off a major mis-appropriation of robot swarms?