The Swedish Transport Administration Research and Innovation fund for Maritime research funded the project "Reference data and algorithms to support research and development of smart ships". The project goes by the working name, and is communicated as, Reeds. It responds to a synthesis of a number of different needs identified in previous projects and studies. The background to the project is that in recent years the focus has been on developing algorithms to interpret and act on the physical environment around different types of craft. In order to be able to develop and evaluate these algorithms, it has become clear that open datasets and a fair benchmarking platform are required that allow various developers in industries and researchers to evaluate algorithms. In the road vehicle sector, Kitti, as of 2013, is the largest dataset used as a reference dataset. The dataset in this project contains sensor data from several data collection occasions within a maritime context, from high-precision sensors such as cameras, radar, lidar, and IMU. For marine applications, there has been no similar dataset with anywhere near the same amount of data and time synchronisation between sensors. The reference data and reference algorithms were available periodically during the project through an online service where researchers and developers could upload their algorithms to use the dataset.
In addition to the dataset itself, Reeds adds additional strengths compared to other reference datasets:
- New approach to comparing algorithms fairly, where new algorithms are always compared on a centralised hardware in a cloud service and re-evaluated when new data is added, i.e. an unbiased algorithm evaluation service.
- Method that combines NTP and PTP time protocols for synchronisation between the sensors with microsecond accuracy
- More types and more modern sensors that can be used at a higher level of abstraction and can thus be applied in more areas.
- Sensor fusion of both onboard and land-side sensors
- Identify areas of application for navigation and surveillance on land based on the algorithms developed during the project and the use of new sensor types not established in shipping.
-
The project built up a maritime reference data set that enables the creation of a digital description for the ship's surrounding environment and developed reference algorithms to demonstrate new navigation and monitoring methodology in the area of "enhanced navigation".
"Enhanced navigation" is defined under the project as the use of new technology based on developments in digitisation and autonomous functions, where new navigation methods use sensors both on board and ashore to increase maritime safety and robustness. The project has built a web-based user interface referred to in the report as "Crowsnest" that handles these new sensors and visualises this data in a familiar interface similar to an overlay in ECDIS that is openly available for the public to build on. Which was used for the evaluation and concept development of new user interfaces based on feedback from pilots and VTS operators.
By providing reference datasets and reference algorithms with demonstrations, researchers and companies now have the opportunity to develop algorithms for the intelligent and autonomous ships of the future.