Welcome to the FlockAI project website!

The aim of the FlockAI project is to deliver a framework capable of enabling Machine Learning and its applications to drone technology for handling time-critical missions (e.g., search and rescue missions). Specifically, FlockAI will advance the current research plain by developing innovative AI-enabled self-adaptive algorithms to ease energy consumption and improve data delivery timeliness in drone swarms. To achieve these goals, the FlockAI project will explore the use of various power-efficient machine learning models for dynamically adjusting, in place, the data sensing and routing of data over drone swarms while maintaining mission requirements. The methods delivered by the project will be placed in a modular and reusable framework for drone swarm operation.

Project Objectives

• The FlockAI benchmark suite. An open benchmarking framework that enables drone technology researchers and data scientists to rapidly deploy, test and evaluate AI algorithms.

• Self-adaptive sensing for drones. Design a novel algorithmic process that dynamically adjusts the sensing intensity of multi-variate data streams collected by drones through low-cost estimation models to capture the runtime evolution and variability of the data.

• Decentralized data routing for drone swarms. Design a novel routing mechanism that can be integrated with the communication interfaces of drones to support multi-hop routing over geo-distributed drone swarms.

Current Work

The FlockAI research team is currently working on designing and developing a framework that will enable researchers to rapidly test advanced AI models for drone technology. Users of our benchmarking framework will be able to rapidly provision -reproducable- emulated testbeds for AI experimentation without the need of knowing complex internals of drone technology or emulation frameworks.

Stay tuned for the first release of our benchmarking suite that will run on top of the open and popular webots emulator!

Key Features

• Python SDK to design drone testbeds with pre-built drone templates
• Deploy ML models and configure on-board or remote inference
• Define and test “what-if” scenarios
• Monitor system utilization, network overhead and energy consumption
• Export experiment data in json, csv, xml and pandas

Video of the Beta Release of the FlockAI testing suite for the Webots Simulator!

The Team

Principle Investigator

Dr. Demetris Trihinas (Geo-Distributed Data Processing)

Researchers

Dr. Michalis Agathocleous (Deep Learning) Mr. Karlen Avogian (Cloud DevOps)

Consulting Research Team

Prof. Athena Stassopoulou (Swarm Intelligence) Dr. Ioannis Katakis (Data Mining/Machine Learning) Dr. Ioannis Kyriakides (Signal Processing)

Publications

• Towards Energy Consumption and Carbon Footprint Testing for AI-driven IoT Services. Demetris Trihinas, Lauritz Thamsen, Jossekin Beilharz and Moysis Symeonides. In 10th IEEE International Conference on Cloud Engineering (IC2E 2022), Sept, 2022.
• FlockAI - A Framework for Rapidly Testing ML-Driven Drone Applications. Demetris Trihinas, Michalis Agathocleous and Karlen Avogian. In 2022 IEEE 42nd International Conference on Distributed Computing Systems (ICDCS), July 2022.
• FlockAI: A Testing Suite for ML-Driven Drone Applications. Demetris Trihinas, Michalis Agathocleous, Karlen Avogian and Ioannis Katakis. Future Internet 2021, 13, 317. https://doi.org/10.3390/fi13120317
• Composable Energy Modeling for ML-Driven Drone Applications. Demetris Trihinas, Michalis Agathocleous and Karlen Avogian. In 9th IEEE International Conference on Cloud Engineering (IC2E 2021), Oct, 2021.

Contact

For more information about the project and our current work, please contact the project principle investigator, Dr. Demetris Trihinas at trihinas.d {at} unic.ac.cy

Acknowledgement

This project is run by the Artificial Intelligence Laboratory (ailab) and is co-funded by the University of Nicosia Seed Grant Scheme (2020-2022).