Sensor platform

The platform for data collection is a 13 m retired pilot boat. The basic sensor platform consists of a GNSS system with three antennas and real-time kinematic positioning (RTK) capabilities and a fibre optic gyro IMU system, four single-channel high-performance vision sensors, two three-channel high-performance vision sensors, a 360° documentation camera system, two high-performance laser range scanners, a conventional marine 360° spinning radar sensor, a weather and barometric pressure sensor, automatic identification system (AIS), and various engine and vehicle state sensors. The GNSS sensor and the IMU jointly provides ground truth positioning.

IMU1KVH P-1775
GNSS1 (3 ant.)Anavs MSRTK
1-channel vision4Flir ORX-10G-71S7M-C
EO 16mm f/4 1″ HPr lens (43.2°)
3-channel vision2Flir ORX-10G-71S7C-C
EO 16mm f/4 1″ HPr lens (43.2°)
360° documentation camera2 (8 cam.)Axis F44
lidar2Velodyne Alpha prime
360° radar1Simrad Halo 20+
weather sensor1Airmar 220WXH

Sensor positions

An overview of the sensor platform. Each dot in the image below correspond to a sensor, and the colored triangles correspond to areas covered by the high-performance vision sensors. All sensors are calibrated into the same coordinate frame.

The pink dots correspond to the high-performance vision sensors, where the pink areas correspond to monochrome field of view (43.2°) and the orange areas to three-channel (RGB) field of view (43.2°). Note that good stereo coverage is given towards the front. Furthermore, the red dots corresponds to the GNSS antennas, the orange to the IMU, the teal to the 360° radar, the green to the 360° documentation camera system, and the yellow to the lidar sensors.

On-board logging equipment

The logging equipment consists of two data center servers, each equipped with: Two AMD EPYC 7352 CPUs (four in total), 128 GB memory, four 16 TB Seagate Exos SATA disk drives, one 15.36 Micron 9300 NVMe U.2 SSD (3.5 GB/s write), one 2 TB Samsung 980 PRO NVMe M.2 SSD (5.1 GB/s write), one Nvidia GeForce RTX 2080, and one Nvidia Quadro RTX 4000. Each server is connected to three Flir cameras and one Velodyne lidar via dedicated 10Gb Ethernet links, and the rest of the sensors are connected via a switched network. The IMU is connected via RS-485 through a dedicated computational unit. All computational nodes run Linux kernel 5.10 with PREEMPT_RT activated. The system is connected to the Internet using a 4G link. All computational nodes are time synchronized using PTP through a common global clock synced through GPS, except for the Axis system which is synchronised using local NTP.