Assisting ship navigation with an accurate and robust solution.
The maritime industry relies heavily on global navigation satellite system (GNSS) technologies to assist vessel navigation. While GNSS is well-established, it must be combined with other positional sensors (such as lidar, cameras radar or sonar) to achieve the near 100% availability that is required for safe autonomous shipping.
Most GNSS systems cannot provide positional accuracy to less than 10m when they are close to large objects, due to signal blocking and multipath errors (the latter occur when satellite navigation signals are reflected off nearby structures such as bridges, buildings and other ships). Jamming or spoofing of GNSS signals also present a risk, through interference from external signals or by delivering intentionally incorrect information about a ship’s location or time zone on an identical frequency band. These signals can lead navigation systems to make inappropriate decisions and leave them vulnerable to theft or hijacking.
To overcome the issues surrounding GNSS navigation, typically, onboard sensors are used such as cameras and lidar, both of which – in certain operating windows – provide detailed information about the vessel’s environment. However, these sensors also have limitations and create vulnerabilities. Cameras and lidar have limited detection ranges and can struggle with the adverse weather conditions often present in marine environments. Additionally, glare from light reflected from the water’s surface, or when the sun is low in the sky, can also be problematic.
Radar-based localisation, as an alternative to GNSS systems, is both accurate and robust and can overcome the shortcomings of GNSS and other sensor solutions. Navtech Radar’s W-band radar (76–77 GHz), can be integrated into existing localisation pipelines to assist with vessel positioning.
Radar localisation works by building a radar map, which allows it to specify the vessel’s position with centimetre-level accuracy, without a global positioning. Typical maritime applications for radar localisation range from small, unmanned surface vessels on inland waterways to large cargo ships navigating ports.
Radar localisation can permit the autonomous navigation of boats in environments where GNSS cannot operate, in any weather conditions and at any time of day (or night). This improves vessel productivity, since it allows work 24 hrs a day without worry about unreliable GNSS positioning or sensor failures. What’s more, when vessels take the most efficient routes they produce fewer emissions, helping operators to achieve their sustainability goals.
Other benefits include:
• GNSS-free: robust localisation that lets any vessel know where it is, in any environment, based solely on radar data.
• No infrastructure-required: zero dependence on external systems makes deployment simple, fast and cost-effective.
• Designed to be flexible and integrate into existing autonomy stacks and sensor solutions: this includes GNSS, inertial data, laser, camera and sonar SLAM solutions.
BAE and Navtech Radar collaborated to trial localisation, using data collected from BAE’s data collection vessel along a stretch of the River Thames. System performance was benchmarked against an onboard GNSS with corrections, and particular attention was paid to areas of GNSS denial, such as under bridges and during sharp turns.
BAE Systems chose a 25.8-mile stretch of the Thames as the trial location; this contained 18 bridges and two sharp turns. One complete traversal of the route was conducted in order to build a map of the environment, then a second was completed to localise against it.
The radar solution uses around 10MB/km of space for map building, thus less than 3.5 GB of space would be needed to map the entire River Thames. The results of the trial showed that the solution provided successful localisation, with an average of 10.3 cm accuracy throughout the journey. The sensor localised the complete length of the journey, including throughout areas that are typically challenging to sensor-based localisation techniques.
Even when off map, with distances up to 50 m, the sensor could still localise itself in the radar map as well as in feature sparse locations where landmarks were up to 300 m away. Figure 1 shows the high-resolution data provided using Navtech radar.
Figure 1: Raw radar data collected on the test vessel, overlaid on a satellite image of the same location.
Figure 2: Radar point clouds (shown by blue and red dots) extracted by the solution to localise vessel position within the environment.
Landmarks were then extracted from the radar data (see Figure 1, above) and used in odometry, scan to scan motion and localising back to the map. This is shown in Figure 2.
Navtech also compared the radar yaw estimates with GNSS headings, then compared this across the localisation run. Results are shown in Figures 3 and 4. Five areas along the route were found to have spike in error. GNSS estimates can be seen to jump significantly at the exit of a bridge and when the boat makes a sharp turn. In contrast, the radar continued to localise successfully, providing accurate positioning in all of these locations.
Figure 3: Radar odometry integrated yaw vs GNSS heading
Figure 4: Locations of instances of GNSS estimation jumps
This trial demonstrated the capability of radar-based localisation, which gave an average accuracy of 10.3cm throughout the journey. When directly compared to GNSS, radar provided superior localisation under bridges and during sharp turns.
Navtech Radar is now looking for partners within the marine industry, willing to collaborate on proof of concept trials using our radar sensor for localisation, perception, anti-collision and obstacle detection.