A swarm of small, wiggling robots that communicate by flashing lights at each other can make collective decisions. This is similar to the process bees use to reach consensus on where to build their nest.
“We believe that in the near future there are going to be simple robots that will do jobs that we don’t want to do, and it will be very important that they make decisions in a decentralised, autonomous manner,” says Carmen Miguel at the University of Barcelona in Spain. She and her colleagues set out to test how imitating bees might help with that.
When bees go house-hunting, scout bees communicate their preferred location to others through a “waggle dance”. The more a bee is advocating for one location, the longer and more vigorously it waggles – eventually other bees join them, and they reach a consensus when a majority are waggling together. Researchers previously translated this behaviour into a mathematical model, and Miguel and her colleagues used it to program decision-making rules into small robots called kilobots.
Each kilobot is a circular disc about 3 centimetres in diameter with three spindly legs. They were each equipped with an infrared-light emitter and receiver, and a coloured LED light. Within a swarm, kilobots could move around, turn clockwise or anticlockwise and use infrared signals to exchange up to 9 bytes of information. Over more than 70 tests, the researchers placed between 10 and 35 kilobots into a walled-off circular arena and tasked some of them with advertising for a particular state – the equivalent of an opinion – by turning their LED a red, green or blue light.
Ezequiel Ferrero at the University of Barcelona says that across all the experiments, kilobots reached consensus within about half an hour, even when they weren’t very crowded together and didn’t have many immediate neighbours to communicate with. He says that getting the right combination of how long they spend transmitting their message and how much they walk around allowed them to make a collective decision in the end.
Keeping the robots simple is key for assessing where they might be useful in real-world situations, says Andreagiovanni Reina at the Free University of Brussels in Belgium. For instance, tiny robots used inside the body for medical purposes couldn’t carry sophisticated electronics, and robots used to respond to emergencies cannot be expensive because they might get damaged often, he says.
In his own work with kilobots, Reina found that making small changes to their behaviour, such as letting them be more indecisive, can have unexpectedly large effects on how they reach consensus. He says future kilobot experiments may uncover mechanisms of collective decision-making beyond just imitating bees.