In short: A consortium including Attocore, the University of Bristol, the Satellite Applications Catapult, Parallel Wireless, and Cellnex built a hybrid satellite backhaul system that combines LEO and GEO links with terrestrial connections for Open RAN networks. The project also developed ML-based handover failure prediction for drones and proved that Emergency Services Network SIMs can make calls on private 5G networks when the home public network is unavailable.
Key Takeaways
- Hybrid LEO/GEO satellite backhaul makes Open RAN viable where there is no terrestrial connectivity — the system dynamically selects the best available backhaul path, trading off latency, bandwidth, and cost in real time
- ML-based handover prediction reduces connection drops for drones — the model anticipates failures before they occur, enabling more reliable beyond-visual-line-of-sight UAV operations
- Emergency services interoperability on private networks — ESN SIMs were validated making calls on the private network, proving private deployments can serve as fallback coverage for blue-light communications
In a nutshell
The Backhaul Bottleneck
Every mobile network needs backhaul — the connection from base station to core network — and in urban areas, fibre or point-to-point microwave links make this straightforward. In truly remote locations, however — mountain rescue posts, offshore platforms, disaster response sites, agricultural estates far from any exchange — backhaul determines whether a network can exist at all.
Traditional geostationary (GEO) satellite links carry around 600 milliseconds of round-trip latency: acceptable for voice but problematic for low-latency 5G applications. LEO constellations such as Starlink and OneWeb cut latency to tens of milliseconds but introduce moving satellites, frequent handovers, and availability dependent on constellation density. The O-RANOS project, funded under DSIT's Future RAN competition, built a system that uses both — and we believe this hybrid approach is fundamentally the right architecture for remote connectivity, even though it introduces significant orchestration complexity.
How O-RANOS Works
The hybrid architecture dynamically selects between LEO, GEO, and terrestrial connections. Low-latency traffic — voice, video, robotic control — routes preferentially over the LEO link; bulk transfers and delay-tolerant traffic shift to GEO, which offers higher sustained bandwidth; and if a terrestrial connection exists, even a constrained rural microwave link, the system incorporates it. This selection happens transparently: the base station and connected devices do not need to know which path their traffic takes.
Attocore provided the compact 5G core orchestrating backhaul selection. The University of Bristol contributed radio and network research. The Satellite Applications Catapult brought satellite systems expertise. Parallel Wireless supplied the Open RAN stack, and Cellnex provided infrastructure perspective and site integration.
Predicting Handover Failures for Drones
Drones move between cells far more quickly than ground users, and handover failures cause connectivity dropouts — for search-and-rescue, pipeline inspection, or beyond-visual-line-of-sight operations, a dropped connection can mean a lost mission or a safety incident.
O-RANOS developed a machine-learning model that predicts handover failures before they occur, using trajectory, speed, altitude, cell geometry, and historical performance data. The network can then adjust handover parameters, prepare fallback cells, or instruct the drone to modify its flight path. This directly supports the UK's ambition for routine BVLOS drone operations, a priority for both the CAA and DSIT. The trade-off, notably, is that ML-based prediction requires substantial training data from real flight operations, and the model's accuracy will inevitably vary between environments until that data is accumulated — but the approach is sound, and the alternative of reactive handover management is clearly worse.
Emergency Services on Private Networks
The most immediately consequential output, and the one we find most compelling: O-RANOS demonstrated that Emergency Services Network SIMs can originate and receive calls on a private 5G network when the home public network is unavailable. Police, fire, and ambulance personnel frequently operate in locations with poor or no commercial coverage — and a private network deployed for another purpose, whether a farm, factory, event, or transport corridor, could therefore provide fallback coverage for emergency communications. The calls worked, the authentication worked, and voice quality was adequate. This is not a theoretical capability; it was tested and validated.
Connecting the Unconnectable
Projects like O-RANOS and Aerix's ONE WORD (the GBP 10 million Open Networks Ecosystem Competition effort, the largest single award in the programme) share a common thread: extending the reach of open 5G into environments conventional deployments cannot serve. ONE WORD focused on integration and deployment challenges in real-world settings; O-RANOS tackled the backhaul and mobility challenges that arise when those networks must operate in the most remote locations.
We have seen from our own deployments that the UK has more geography that is hard to connect than geography that is easy. Hybrid satellite backhaul, intelligent handover management, and emergency services interoperability are consequently the building blocks that make universal coverage a practical proposition rather than a policy aspiration.
If you need connectivity in a location where conventional backhaul is not available, get in touch. Read more about our rural and remote deployments on our rural connectivity page.