Graph showing how often the dumping of new trash into the urban environment is disposed of. Source: Alfeo et al.Hoping to improve the efficiency and autonomy of urban waste management, researchers from MIT Media Lab, Université Libre de Bruxelles and the University of Pisa have been investigating the use of swarm robotic systems.

"With autonomous vehicles, swarms of drones for deliveries and teams of robots organizing warehouses, the city of the future will be a cybernetic ecosystem consisting of machines and humans," Antonio Luca Alfeo, one of the researchers who carried out the study said. "This is a fascinating scenario, as well as a huge technological challenge. The management of such systems results in a complexity that grows exponentially as more machines are deployed, especially if we want them to actually 'live with us' and react autonomously to the changing needs of their surrounding environment."

As such, the research team explored applying techniques inspired by nature to robot swarm management.

"We propose a trash disposal system with swarms of robots whose self-organization is based on the behavior of social insects, also known as stigmergy-based foraging," Alfeo explained. "The proposed system deals with the disposal of trash, from trash bins to a few central deposits."

The robots created by Alfeo and his team are not instructed or controlled by a remote service. Instead, they make all of their decisions in the moment, driven by real-time observations and perception.

"Among these perceptions, there are also 'virtual pheromones' released by other robots on RFID tags in the urban environment," Alfeo said. "These mark the most convenient path from one or more non-empty trash bins and the nearest deposit. Most importantly, thanks to its collective behavior, the swarm is able to autonomously self-organize in order to target the areas with the greatest amount of garbage, always providing an effective response."

Covering different areas of expertise and disciplines, Alfeo's team members came from a variety of backgrounds including data science, city science and bioinspired robotics.

"Team members specialized in city science provided the model of the robot used in the study and dealt with the modeling of urban space in GAMA, a multi-agent simulation platform," Alfeo said. "Those skilled in the behavioral design of swarm of robots realized and tested the logic of the swarm in a properly augmented model of the urban environment. Finally, the analysis of the implications of each design choice was supervised by highly-proficient data scientists."

The swarm system for waste management was evaluated by Alfeo and his team and they determined that it outperformed other techniques currently in use.

"We showed that a swarm of self-organized robots could lead to great improvements in the context of waste management, without any external information source or prior knowledge about the trash disposal demand," Alfeo said. "Moreover, we provided insight on the design of such a system with the aim of balancing its two main properties, that is, the exploratory capability and the responsiveness of the swarm. The first can reduce the occurrence of full trash bins, while the latter helps to reduce the amount of trash in the urban environment."

As Alfeo and his colleagues continue their work, they also see the potential in using their swarm robot systems for other applications.

"The proposed approach is not specific to waste management and it could be tested in a number of different applications, such as autonomous vehicles," Alfeo said. "Since in this case safety is a critical factor, however, the design of such a system should include further specific solutions, such as block-chain technology."

The research is pre-published on arXiv.