Researchers from the University of Bristol and the University of the West of England (UWE) have developed a new generation of swarming robots called “Teraflop Swarm” that can adapt to new environments and independently learn new behaviors through a process of artificial evolution. This can pave the way for developing robotic systems for environmental monitoring, disaster recovery, infrastructure maintenance, logistics and agriculture. Results are published in Advanced Intelligent Systems.
The custom-made swarm of robots can have their own rules assigned to them, giving rise to certain swarm behaviors. (Image Credit: University of Bristol)
Artificial evolution has always been run on an external computer that connects to the swarm, where the best strategies are communicated to the robots. This can be limiting because it requires a laboratory setting and an external infrastructure.
Researchers used a custom-made swarm of robots that includes high-processing power fixed within the swarm. They were then able to find appropriate rules to give rise to chosen swarm behaviors. This could allow robotic swarms to continually adapt in the wild on their own, carrying out any tasks and conditions that have been assigned to them. The advanced controllers can also be queried, clarified and improved by making them understandable to humans.
"Human-understandable controllers allow us to analyze and verify automatic designs to ensure safety for deployment in real-world applications." Simon Jones, lead author from the University of Bristol's Robotics Lab said: The team used advances in high-performance mobile computing to develop the swarm of robots, seeking inspiration from actual swarms found in nature. The teams “Teraflop Swarm” is able to run the computationally intensive automatic design process within the swarm, allowing it to function without the use of other resources. It can reach a high level of performance in just 15 minutes, which is a lot quicker than other embodied evolution methods designed in the past and without depending on external infrastructures to operate.
“This is the first step towards robot swarms that automatically discover suitable swarm strategies in the wild." Dr. Hauert, Senior Lecturer in Robotics in the Department of Engineering Mathematics and Bristol Robotics Laboratory (BRL), said: "The next step will be to get these robot swarms out of the lab and demonstrate our proposed approach in real-world applications."
Removing the swarm’s dependency from an external infrastructure and making it possible to perform analysis, understand and explain the controllers will allow researchers to develop robots with automatic design of swarm controllers in the real world. This would be useful in different scenarios like disaster recovery and environmental monitoring.
Starting from scratch, a robotic swarm could have a suitable strategy communicated to them in situ, and have the strategy altered whenever the environment or assigned task changes.
"In many modern AI systems, especially those that employ Deep Learning, it is almost impossible to understand why the system made a particular decision. This lack of transparency can be a real problem if the system makes a bad decision and causes harm. An important advantage of the system described in this paper is that it is transparent: its decision-making process is understandable by humans." said Professor Alan Winfield, BRL and Science Communication Unit, UWE.