You are probably wondering how we could possibly program robots to collaborate when most of us have such a difficult time doing it effectively ourselves! Well MIT researchers have recently discovered that it is quite a challenge indeed. (“Helping Robots Collaborate”, Kurzweil Accelerating Intelligence News, February 14, 2014)
Currently researchers are working on applying a combination of robotic control programs to enable groups of robots to collaborate. The most recent tests of this complex system have included a simulation of a warehousing application where teams of robots are required to retrieve arbitrary objects from indeterminate locations, collaborating as needed to transport heavy loads. It sounds simple enough right? Well, as with any collaboration communication has proven to be one of the primary issues. There are far too many variables involved to program a detailed set of communication conditions – similar to communications in any work environment. And similar to human work environments and modes of communication the greatest success in this venture for robotic collaboration has come when the robots are given the tools and freedom to ‘decide’ how best to communicate and accomplish what they need – like a self organizing team. Each robot has a series of coloured lights to use for communication when their direct relay systems are slow or out of order. Originally, the programmers were attempting to create a specific light response for any situation/communication need that may arise, which of course proved impossible given the infinite number of variables and qualifiers involved in a group collaboration. What has provided some success is programming the robots to identify the coloured lights as a method of communication and allow the artificial intelligence algorithm determine whether use of the lights is necessary and what the colours mean.
The Kingbridge Insight this week is an extrapolation of the lessons this robotic experiment has to offer while attempting to program collaboration behaviour.
Traditionally, organizations work in a hierarchy, where actions and behaviours are determined by a superior officer. Even though many organizations have taken some steps towards creating ‘bottom up’ environments the underlying structure for the most part remains the same. In this experiment however, the most successful collaborations resulted from self organization and the absence of command and control.
What conclusion would you draw from this about the conditions required for successful collaborations?