muteSWARM

The aim of the muteSWARM project is to propose novel control techniques in aerial swarm robotics that are not based on radio communication (i.e. muted). The key strategy will consist for a robot, as animals do in the wild, to sense the close neighbors and its environment by means of its perception and not by direct communication and even without labeling of the drones detected. We propose to investigate hearing here in particular, which is little explored in robotics. Multi-rotor robots are noisy and the sound UAVs produce is relevant information for their localization by neighbors, without needing to design and embed extra sound transmitter devices. The acoustic localization of an aerial robot will be explored, particularly when using the natural sound produced by the robot itself. Here, it would be a premiere maintaining a swarm of UAVs in formation without radio communication, only based on the acoustic signature of neighbors to determine the inter-UAV distances. In addition and to go even further in bio-inspired flying robotics, it is also planned to make the swarm navigation more robust by applying multimodal sensing: not only acoustic localization will be studied, but also a combination of acoustic sensing and low-resolution visual alternatives. In this context, AI-enhanced swarm formation control algorithms will be developed. The project is intended to be frugal, seeking to design a low-tech and low-cost UAV swarm, with lower-resolution perception, low resource requirements, and low energy consumption. It is also relevant to point out that the hoped-for gain in terms of parsimony, scalability, and resiliency, could apply to other robots or swarms of robots, not only to UAVs. Besides, algorithms will be made as generic as feasible so that they can suit any system. The targeted experimental validation scenario will be to maintain a flight formation in a mixed indoor and outdoor environment, where the full potential of the multimodal proposal will be exploited (in the framework of the French national research infrastructure Robotex 2.0 and Equipex+ TIRREX). The acoustic mode will offset the blinding effects when UAVs fly from inside to outside, while conversely, the visual mode could be helpful for the reverberation effects when the robots return indoors.