Cities are getting busier, more crowded, and increasingly strained by the consequences of e-commerce growth. According to recent estimates, urban freight traffic already accounts for around 25% of traffic emissions in major European cities and occupies more than 30% of available road capacity during peak hours. Without targeted intervention, these figures risk climbing even further. The question is: how do you organize last-mile logistics more intelligently without further choking the city?
The answer that researchers and practitioners are giving with growing confidence: digital control towers, fed by real-time data and AI. The European Horizon project GREEN-LOG developed and tested exactly such a system across five European cities: Athens, Barcelona, Leuven, Ghent, and Oxford.
What is a digital control tower, exactly?
The name sounds grand, but the concept is manageable. A last-mile Digital Control Tower (DCT) is a platform that brings together all relevant data on delivery operations — from vehicle locations to parcel flows, from traffic delays to policy restrictions — into a single shared system. This allows both logistics service providers and local authorities to monitor, anticipate, and intervene in real time.
The GREEN-LOG platform uses a layered architecture: a processing layer that manages the underlying logic, a service layer with modules for route optimization, demand forecasting, and monitoring, and an interface layer accessible to drivers, LSPs, and municipal staff. What makes the design particularly practical is its modularity — cities don’t need to implement everything at once, but can connect piece by piece based on their own level of digital maturity.
From policy goal to practical delivery route
One of the most concrete innovations in the system is the direct link between municipal policy and operational route planning. Through the DCT, local authorities can define geolocated zones — think school areas, historic city centers, or low-emission zones — with associated time windows and permitted vehicle types. Those rules are automatically translated into routing constraints that delivery drivers see directly on their phones.
In Athens, this led to a shared consolidation point just outside the city center, where multiple logistics providers could bundle their trips before entering the congested historic core. In Leuven, local shops and recipients earned “green credits” for choosing more sustainable delivery options — a smart nudge mechanism tied directly to the policy framework. In Barcelona, the system was extended to include multimodal routing: combinations of trains, autonomous delivery robots, and cargo bikes for zones that traditional vans simply cannot reach.
The measured results from the Athens pilot are striking: a 35% drop in kilometers driven, 26% less travel time, and 12% fewer stops. That’s not just good for air quality — it directly reduces costs for logistics service providers too.
AI as the engine under the hood
What makes the whole thing genuinely intelligent is the predictive layer. The GREEN-LOG system uses a hybrid deep learning model (a combination of Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks) to forecast demand patterns by postcode. CNNs capture spatial patterns in delivery data, while LSTMs learn from time-dependent trends such as day-of-week effects or seasonal peaks.
This is no theoretical exercise: the Oxford case study, with delivery company Pedal & Post, shows how historical parcel data per postcode is turned into forecasts that give drivers direct route suggestions the following day. The system draws on anonymized delivery data covering a two-year period, a solid foundation for reliable predictions.
What’s holding back scale-up?
The technology is ready. The willingness to share data is less so. The authors are candid about the biggest bottleneck: competing logistics service providers are reluctant to share operational data, fearing competitive disadvantage. Without that data exchange, however, the DCT performs at only half its potential: forecasts become less accurate, route optimization less effective.
This is not a technical problem, but a governance one. Cities that are serious about smart urban logistics need to invest in trust structures, shared standards, and, where necessary, regulation that compels or rewards collaboration. The technology will follow.
This blog is based on chapters from the book Urban Logistics Transformation (Springer, 2026), covering the GREEN-LOG project (EU Horizon, grant no. 101069892).
Ntemou, A. et al. (2026). Last-Mile Digital Control Tower. In: Melkonyan-Gottschalk, A., Li, Y. (eds) Urban Logistics Transformation. Greening of Industry Networks Studies, vol 15. Springer, Cham. https://doi.org/10.1007/978-3-032-08387-6_3
Georgakis, P., Hocking, M., Jia, G. (2026). Deep Learning Demand Prediction for Sustainable Logistics as a Service (LaaS). In: Melkonyan-Gottschalk, A., Li, Y. (eds) Urban Logistics Transformation. Greening of Industry Networks Studies, vol 15. Springer, Cham. https://doi.org/10.1007/978-3-032-08387-6_4