BOSCH employs smart sensors to collect data that enable Environmental Sensitive Traffic Management following a close loop approach. This use case offers the opportunity to prove the benefits of smart sensing by combining the merits of smart sensing with a platform for secure and trustworthy data management. A smart sensor interprets, manages, processes, combines and fuses different single sensor signals and therefore leads to more accurate, interpretable data to choose and evaluate specific measures to reduce noise and air pollution from traffic.
Τhe challenge is quickly scaled by the number of devices deployed in the wild and the size of the surveillance area. In single sensor environments, the aim of a robust and accurate vessel tracker is to resolve measurement-to-object association ambiguities, especially in cluttered multi-object scenarios. The deployment of several devices turns it in a multi-sensor, multi-object tracking problem. The increased number of measurements in multisensory systems aggravates the measurement-to-object association problem which arises if multiple closely spaced objects or clutter measurements are present. In this case, the challenge is to combine the results at a regional level providing macro analytics on aggregated data from the edge devices.
Vessels serve as distributed mobile sensors, hence, Machine Learning will happen in-situ (on-board) of the vessels (close to the data production). This will enable efficient data operations across the maritime data life cycle and will lead to better situational awareness, increased safety, and reduction of illegal activity.
This use case will validate how crowdsourced nautical sensor data can be processed and qualified for improving navigational safety at sea using a decentralized data acquisition and Federated Learning approach. The Federated Learning approach will prioritize what information is urgent to share with the centralized server and what can wait until connected to a low-cost communication channel. All vessels with standardized equipment can become a Federated Learning crowdsourcing participant for improving safety at sea. This approach enables different data feedback loops providing different levels of contributions to nautical charts ranging from instant high-priority warnings against hazards to generic verification of existing models and validations of improvements to models.