Bio-inspired UAV swarm flocking guided by information-theoretic principles

Autonomous UAV swarms emerge as a disruptive technology with impact on many areas of our life, from monitoring key ecosystems and hazardous environments to enabling advanced data insights.

However, creating large, decentralized UAV swarms with synchronized flocking behavior and autonomous collision avoidance is a challenge.

This project aims to develop a holistic framework for supporting the design of UAV swarms based on generic information-theoretic principles - empowerment and relevant information - complemented with the bio-inspired influential neighbourhood principle.

The key objectives are to design locally optimized control mechanisms guiding the collective motion and to provide theoretical measures for real-time characterization of swarm cohesion.

The influential neighborhoods theory provides an efficient strategy for maintaining group cohesion in the coordinated motion of animal groups.

It allows agents to limit their attention to a small subset of their neighbors, reducing information processing and cognitive load and offers explicit and concise models that can be implemented directly.

Empowerment models agents behaviour with the principle of keeping ones options open, assuming that, maximizing the future potential outcome of ones actions is a survivability driver in nature.

Relevant information provides a measure, quantifying the minimal amount of information an agent needs to process in order to achieve a certain level of utility.

Identifying influential neighbourhoods dynamically will serve as a base for maximizing agents empowerment which will guide the UAV swarm self-organized collective motion.

The integration of relevant information, influential neighbourhood identification and empowerment maximization into a novel principled framework for optimizing agents decision-making will provide enablers for the future adaptive and scalable UAV swarms, maintaining swarm cohesion while manoeuvring in uncertain and complex dynamic environments.