Wireless Communications & AI Research
Tyndall’s Wireless Communications & AI research programme addresses the fundamental research challenges in future wireless communication networks and edge AI:
- Hyper-flexible wireless networks to enable intelligent and ultra- fast ultra-reliable low-latency connectivity between the human, physical and digital worlds.
- Future AI algorithms and automatic creation of digital twins to enable efficient and robust on-line learning with limited data, reduce manual configuration, and create more universal problem solvers.
- Co-designing hardware enablers for wireless networks and AI at the extreme edge including programmable radio frontends and novel brain inspired AI ASICs.
The resulting innovations will create 10-100x improvements in key dimensions to enable new AI and real-time applications and make them accessible to humans and machines anywhere. We are creating the building blocks for next generation IoT communications, Wi-Fi, 5G, 6G and beyond – pushing the boundaries in science, helping industry partners to innovate, and creating high value start-ups with global impact.
Our mission is to push the boundaries in science at the intersection of Future Networks and AI through co-design of algorithms and hardware to enable reliable and instant access to all information, computing resources, and AI, for humans and machines anywhere. We work closely with industry to translate our research findings into tangible benefits for societal and commercial impact.
We are contributing to several AI policy white-papers and technology innovation roadmaps at EU level, through the participation to European Technology Platforms such as EPoSS (European Platform on Smart Systems Integration ) and AIOTI (Alliance for IoT and Edge Computing Innovation) and are defining the future for AI in networks through EU projects such as 6G-XCEL.
Applications include digital twins to train AI without disruption, holographic communication, medical device communications, Industry 4.0, high-resolution sensing, and many others.
Research Challenge
AI and wireless networks have become essential technologies in our lives both at home and at work, allowing us to be always connected and giving us access to increasingly sophisticated services. Both technologies are intrinsically linked as AI relies on networks to both access and exchange data, and to deliver results to human and machine users. To achieve this anywhere, wireless networks such as 6G are required that are also increasingly relying on AI to automatically optimize their operation. To enable the continued evolution and growth of these critical technologies, our researchers are addressing the following key research challenges.
Hyperflexible Wireless Networks: To enable the next 10-100x leap in performance, future networks will have to become more specialized. However, specialization has been expensive in the past and the industry has adopted a 10-year innovation cycle between network generations (3G to 4G to 5G) to achieve economies of scale. Tyndall are creating Hyper-flexible AI defined networks that are fully programmable from the RF hardware, Physical and MAC layer, to the communication protocols, and able to adapt to the most extreme requirements in a cost-effective way. This way, we enable the next generation of services including holographic communication, remote control of robots and drones, medical devices, connected implants, as well as additional capabilities such as sensing and localization.
Fundamental advances in AI techniques: AI moved from labs to the real-world and is currently applied to a broad range of sectors and industries worldwide. However, the high energy consumption, the need for manual fine-tuning, poor robustness, and trial and error during the learning process, are limiting the application of AI, particularly for resource constrained applications such as those at the extreme edge. We are creating efficient learning at the edge with limited data, novel AI techniques combining neural networks with genetic evolution of algorithms combining best of biological and traditional computation, and digital twins for enabling online learning without disruptions. This will create the key enablers for a wide range of new AI applications including fully autonomous cars and robots, material discovery and manufacturing, telemedicine, and the next generation of wireless networks.
Co-designing hardware enablers for wireless and AI at the extreme edge: The advances in wireless communications and AI rely on co-design of hardware enablers. In the area of wireless communications we are creating programmable radio frontend hardware as the key enabler for AI defined networks. Examples include programmable filters, antennas, antenna arrays and programmable RF surfaces. This is enabled by using novel reconfigurable materials, their engineering to yield attractive switching characteristics and ultimately their use in reconfigurable RF-to-sub-THz -wave front ends. In the area of AI we are creating custom AI ASICs based on neuromorphic computing and hardware accelerators, to improve energy efficiency by orders of magnitude and accelerate novel AI techniques.
The featured projects below give an overview of some of the research in the area of Wireless communications and AI at Tyndall.
Recent Publications
S. Bulja, R. Kopf, A. Tate, M. Cappuzzo, D. Kozlov, H. Claussen, D. Wiegner, W. Templ, D.M. Syahkal, “High frequency resistive switching behavior of amorphous TiO2 and NiO“, Scientific Reports 12 (1), 1-16, (2022).
V. Kirillov, D. Kozlov, H. Claussen and S. Bulja, “In-Vessel Resonant Communications“, IEEE Access, (2024).
B. Galkin, L. Ho, K. Lyons, G. Celik and H. Claussen, “Experimental Evaluation of Air-to-Ground VHF Band Communication for UAV Relays“, IEEE International Conference on Communications Workshops (ICC Workshops), Rome, Italy, pp. 1428-1432, (2023).
Project: Hyperpath
In the Enterprise Ireland Commercialization Fund project HyperPath, we enable combining multiple unreliable wireless links into an ultra-reliable low latency connection. This is achieved through a patent-pending P2P multi-connectivity architecture to minimize latency and cloud costs, novel multi-path replication and aggregation protocols, that is easily deployable as a user space application. This functionality is a key enabler for beyond visual line of sight connectivity for drones (UAVs/UGVs), Industry 4.0, connectivity for future medical devices and other applications that require reliable connectivity anywhere. We presented HyperPath at Mobile World Congress in Barcelona in 2024 and are currently trialing with several unmanned Aerial and Ground Vehicle companies across Europe, with the intention of creating a spinout company in 2024.
“Connectivity is critical for the operation of our fleet. If connectivity drops our customers’ business is negatively impacted, this must not happen. HyperPath uniquely solves this critically important problem for us” — Mike Potts, Founder & CEO, Street Drone.
Find out more: www.hyperpath.ie .
Project: DTIF GUARD - Drug Interdiction Using Smart Drones
The Disruptive Technologies Innovation Fund project GUARD will create a new smart autonomous drone system with the aim of helping the Irish Navy cost-effectively monitor Ireland’s large coastline to prevent drug smuggling into Ireland and the EU. The GUARD solution will be a leap in capability, compared to the state-of-the-art, able to operate fully autonomously in harsh weather conditions with 800 km range, support vertical take-off and landing from both ships and shore, automatic creation of flight plan and permissions for operation in civil airspace, automatic creation of a digital twin of survey area & AI based video analysis, gigabit mm-wave communications & reliable low latency multi-connectivity, and VR control enabling intuitive ‘human in the loop’ high value decision making. The capability will be demonstrated in collaboration with the Irish Navy and position Ireland as a leader in autonomous drone technology, a market that will grow by 15.5% annually to $45.8bn by 2025.
Project: AI for CRFID
Passive Radio Frequency Identification technology (RFID) has been a key wireless communication technology enabling IoT. Recent advances have paved the way for battery-less, chipless Radio Frequency Identification (CRFID), which eliminates the need for an integrated circuit (IC) component on the tag. This project introduced a novel design strategy for concentric rings-based polarization-insensitive CRFID sensing tags, enhancing data encoding capacity by 88.2% compared to conventional designs. Implementing Radar Cross Section (RCS) nulls for data encoding enables accurate sensing by the innermost ring, with an additional outermost ring aiding robust detection. Artificial Intelligence (AI) is integrated on the reader side, employing machine learning (ML) and deep learning (DL) techniques for decoding RCS EM signatures. ML/DL models generalize well, with 1D-CNN DL models outperforming conventional ML models in ID and sensing value detection. Additionally, a 3-bit depolarizing CRFID tag is developed and enabled for surface and shape robust detection using AI, achieving low normalised RMSE values (0.48%) for tag ID detection. These contributions significantly advance AI-enabled CRFID systems for robust IoT applications.
