CROSSMARE aims to transform European maritime surveillance and strengthen the security of EU sea borders through an innovative, multi-domain interoperability framework. The project combines advanced technologies, artificial intelligence, and cross-border collaboration to deliver a unified and more effective approach to maritime situational awareness.
To achieve this, a consortium of 17 big industries, SMEs, RTOs, and end-users:
- utilize cutting-edge technologies in the context of maritime surveillance: High Altitude Platform Stations; Remotely Piloted Aircraft Systems; High-Res SAR Sat imagery; Unmanned Aerial Vehicles; Autonomous Underwater & Unmanned Surface Vehicles; novel Flash LiDAR sensors; as well as legacy surveillance systems
- apply innovative AI algorithms; Large Language and Visual Models; Advanced multimodal data fusion
- extend the capabilities of existing C2 and HMIs, also supporting intelligent mission management of manned and unmanned systems
- support the interoperability of novel and legacy surveillance systems, improving cross-border and crossagency collaboration. To this end, CROSSMARE aims to demonstrate interoperable data exchange with CISE and proposes an extension to the current CISE data and service model, enhancing existing and enabling new functionalities.
The project will validate the proposed technologies through a series of validation activities:
- continuous integration testing,
- 2 Small Scale Tests (Italy, Lithuania)
- a Table Top Exercise (Romania)
- a Full Scale Demonstration (Spain) involving real assets, end-users and relevant stakeholders.
A Security and Privacy-by-design framework will ensure compliance with fundamental rights, privacy and personal data protection regulations.
Eviden leads WP5-WP6 in CROSSMARE. Eviden main effort is concentrated in the implementation of an Agentic AI solution specialized on the detection of anomalies of the ship data. This Agent, based on the combination of VLMs and LLMs, will gather and fuse the signals provided by the sensors available, and will provide, as an output, recommendations and risk-based support information that can be used to enhance the situational awareness of the final user.
The agent will act, in the input, as a multi-modal fusion service. Then, this information will be analyzed and combined with pre-stored expert knowledge related to the maritime domain and with other external sources, to provide the risk of the ships analysed.
The agent will be designed to improve the detection capabilities and performance, reduce false detections and assess emerging threat incidents.
The agent will identify anomalies and unusual patterns from different type of data structures (e.g., suspicious vessel activities, underwater threats, unusual patterns etc.) Additionally, it include explainability to help the user understand and audit the chain of reasoning followed by the Agent.
