Crowdsourcing Swarm Control
Swarm robotics is all about multiple embodiment. Specifically controlling multiple bodies as effectively a single unit. A single drive uniting them. Each is free to move individually, but each has a role to play in the final form.
Through this method intelligent particle swarms can be created, or squadrons of planes that think with one mind.
Swarm robotics is still in an early stage however, and we still don't know the best ways to control even simple swarms of robotic entities. The data is even thinner on the ground when it is a human doing the controlling as opposed to the swarm controlling itself. We simply have no idea what the most effective interface method might be.
Enter Aaron Becker, a postdoctoral researcher at Rice Universitys Multi-Robot Systems Laboratory. His idea for how to get a handle on the concept is to test it as many times as possible with a league of human volunteers.
An online game has thus been created by himself and his team, to allow any individual who wishes to, to attempt to control a vast swarm of simplistic, virtual swarm bots. These bots all have individual starting locations and orientations. Through a variety of interface methods your task is to get them to move as desired to achieve a goal. For example, one of the goals is to steer an injected mass of swarm bots around a patient's body so as to deliver a critical radioactive dose only to specific body structures whilst minimising the effect on other systems.
The data from these games will help us better understand how to use multi-robot systems with massive populations to perform coordinated, complex tasks, said lab director James McLurkin, assistant professor of computer science at Rice.
The simulated robots in the games are based directly on the r-one, a relatively cheap but sophisticated swarm robot system that director McLurkin began designing in 2009. This means that the team can directly map successful strategies from the virtual environment to the physical one.
So long as a single obstacle remains in the control field the robots operate in, the robots whether virtual or physical, can be moved to assume any shape with the same simple commands: A rotate 90 degrees, and move forwards/backwards command. As such, swarm control becomes as much about intelligently recognising and responding to changes in the environment around the swarm, as it does in controlling the swarm itself.
With precision, the swarm of robots, all of which originally faced the same way, and all have identical controls, can be moved to any formation and any number of orientations by all following identical commands and interacting with this environment. This is the core of the swarm control the lab is trying to crack.