Ravenna terminal backs AI logistics digital twin

Altmann Sapir Intermodal Autoterminal is preparing a €25.4 million artificial-intelligence and digital-twin programme at Ravenna to improve the coordination of automotive cargo moving between vessels, trains, trucks, and terminal storage.

The SINAPSI project — Sistema Cognitivo Integrato per la Logistica Intermodale Automotive — has secured €1.09 million of support from the Emilia-Romagna region. Ravenna will provide the operating environment for the pilot, while the system is being designed so that the underlying approach can later be replicated at other logistics nodes.

The programme addresses one of the more persistent problems in intermodal freight: individual processes can be efficient while the complete movement remains poorly coordinated. A vessel may arrive late, a train path can move, truck collections can bunch together, and yard occupancy can change faster than a static plan created several hours earlier.

SINAPSI is intended to respond to those changes by combining operational data with predictive models and optimisation algorithms. The project includes analysis of logistics processes, development of the AI models, creation of a digital twin of the terminal, and integration of the components into a unified software platform.

The digital twin will allow the terminal to reproduce operational scenarios virtually before altering the physical process. Operators could model the effect of a late ship on yard capacity, test alternative vehicle-storage locations, examine how a rail departure should be sequenced, or identify where truck collections are likely to create congestion.

Automotive terminals are particularly suited to this type of optimisation because finished vehicles occupy large areas and must remain individually traceable. Unlike containers, cars cannot be stacked several units high, while unnecessary repositioning consumes labour, equipment time, yard capacity, and energy.

Moving vehicles repeatedly can also increase the risk of damage. A planning system that reduces avoidable internal moves can therefore improve cost and productivity without requiring the terminal to handle more cargo through a faster crane or an additional piece of heavy equipment.

The same principle applies between transport modes. Faster vessel unloading is of limited benefit if the additional vehicles simply accumulate in storage because the next train is unavailable, while a rail departure cannot run efficiently if the vehicles allocated to it are dispersed across several parts of the yard.

A digital twin can model those relationships as one system. The intention is not merely to optimise berth, yard, road, or rail performance independently, but to find a sequence that uses the available capacity across all four without solving one bottleneck by creating another.

The programme is expected to move towards a functional alpha prototype validated in a representative operating environment. That stage will be important because logistics software often performs well against clean historical datasets but encounters more difficulty when live operations introduce incomplete information, equipment downtime, late arrivals, and human intervention.

Data quality will consequently matter as much as the AI model. The software can only make a useful recommendation if it has reliable information on ship arrivals, train availability, truck collections, yard position, cargo status, and available handling resources.

Integrating those feeds may prove harder than developing the optimisation logic itself. Ports, carriers, rail operators, trucking companies, and terminal systems often use different platforms and data standards, and the information needed for one decision may belong to several organisations.

If SINAPSI can combine those sources effectively, the benefit should extend beyond shorter vehicle dwell. Fewer unnecessary moves can reduce equipment utilisation and energy use, while better sequencing can improve the use of trains or road capacity already booked.

Simulation can also help distinguish between physical and operational capacity. A terminal may believe it needs more land or infrastructure when the existing layout is being constrained by poorly sequenced flows. Modelling different operating patterns provides a way to test that assumption before committing capital to physical expansion.

None of that removes the need for conventional investment where throughput genuinely exceeds the limits of the site. A digital twin cannot manufacture an additional rail siding or parking area, but it can make clearer whether those assets are required and how they should interact with the rest of the terminal.

The Ravenna project remains a development programme rather than a completed commercial system. The regional funding confirms that substantial work is moving forward, but performance will ultimately have to be demonstrated through measurable outcomes such as dwell, vehicle moves, equipment utilisation, and transfer reliability.

The replicability objective raises the commercial stakes. A system built solely around one terminal may improve one operation; a model that can be adapted to other nodes could transfer operating knowledge across a wider network without reproducing every physical layout.

That is where the €25.4 million investment will have to justify itself. The value is not in creating an elaborate virtual representation of Ravenna, but in whether the model helps the real terminal move vehicles through ship, rail, road, and storage with less wasted time, less unnecessary handling, and more usable capacity.