Mimicking quorum sensing to synchronize dancing Nao robots
So far, there have been research groups and companies who demonstrated how their robots can function in sync. However, influence from various parameters and events in real world applications do affect robots. A group of researchers from Massachusetts Institute of Technology (MIT) in Cambridge, MA, developed an approach able to partially handle these problems by mimicking biological phenomenon of quorum sensing.
One way to coordinate the robots is to designate a leader (or a network system) which coordinates them by broadcasting details of its movement and position over a network that the other robots all follow. Currently used network dynamics are not as predictable as desired, and they suffer from delays. While small delays of up to half a second are more common, some directives may be delayed by several seconds – a time long enough to completely blot the desired routine.
Alternative to that approach is to program each unit to perform the desired routine and coordinate their behavior by synchronizing their internal clocks at the start of the whole routine or the part when that robot gets a role. The advantage of this approach is that the chances of the clocks becoming desynchronized are usually very small, however, there is no way to regain coordination once there is some unexpected interruption of the routine.
MIT researchers Patrick Bechon and Jean-Jacques Slotine were inspired by the ability of bacteria and social insects to sense not only the presence of other members of their species and to synchronize their behavior in a process named quorum sensing. In biology, quorum sensing functions thanks to signaling molecules individual units emit into the environment while at the same time measuring the local concentration of these molecules.
Bechon and Slotine applied this method onto robots by enabling each robot to have access to every other robot’s position. The robots coordinate by accessing a global variable such as the average position or average clock time, each with their own right to influence the variables since that information contributes to the average and most synchronized motion.
In order to test their approach, the researchers used a group of 8 Aldebaran NAO robots, each with an internal clock which attempts to synchronize with a global average time maintained by a central server. Unlike other systems where a server acts as master with the robots as slaves that simply follow its signal, this system enables the robot to continue with their routine even if the connection to the server is lost.
As you can see in the video above, this approach enables the robot to continue with their routine even after falling over. The central server acts as the environment does for ants, and it enables the robots to sense and interact with the current actions of the group they are belonging.
When it comes to dancing robots, the researchers stated that it is possible to enable the robots to synchronize with music by determining the frequencies in order to configure the tempo they need to follow (something similar to what My Keepon robotic toy performs), thus synchronizing via music instead of the quorum signal. It can also be a good way for robots to improvise dancing.
The system could eventually enable robots to synchronize with all other robots sharing the same speed, size and functions. Aside dancing, synchrony enables large scale co-operation, and it could be applied in processes of manufacturing and construction.
For more information, you can read the paper: “Synchronization and quorum sensing in a swarm of humanoid robots” [540KB PDF].