Just like self-driving cars, we can expect autonomous ships to become more common in the coming years. However, two safety challenges which need to be solved along the way, and which are different for ships compared to cars, are navigation without GPS and remote control. In his research, Mårten Lager has developed user interfaces for remote control of ships, as well as positioning using water depth, magnetism and landmarks. On 12 February at 2.15 pm he will present his thesis "Digital Cognitive Companions for Marine Vessels – On the Path Towards Autonomous Ships".
- The thesis for download at portal.reseach.lu.se (PDF)
- Abstract and registration for the online PhD Defence 12 February at 2.15 pm
“As I am an industrial doctoral student from Saab Kockums, my research is about ships”, begins Mårten Lager, an industrial doctoral student from Saab Kockums on the WASP* research programme.
Navigating like the ancient Greeks
For ships to be more autonomous, it is important for them to be able to manage even if satellite-based GPS signals and other communications fall away or are restricted. To solve the navigation problem, Mårten has developed software which, using artificial intelligence, combines information from several different sensors, such as measuring water depth, directions towards landmarks and the strength of magnetic fields.
“I then compared the results with known maps such as nautical charts and magnetic maps and thus worked out where the ship was located. In essence, this is not new; even the ancient Greeks navigated using landmarks and water depth. The magnetic fields are probably a little newer, however. What is different, perhaps, is that this technology is aided by modern computing power, can simultaneously calculate thousands of hypotheses and uses probability calculations to ascertain the most likely position.
“For positioning, I have used a particle filter that keeps track of a thousand parallel positional guesses, or particles. The algorithm constantly checks the probability that the position of the particles is reasonable by comparing sensor data with known maps. I compare water depth, bearings against landmarks and magnetic intensity against nautical charts and magnetic maps. I tested the algorithm in the Västervik archipelago on one of Kockum's autonomous ships with good results.”
Remote control with limited bandwidth
The second problem addressed by the thesis is how to remotely control ships that are in locations where bandwidth is limited.
“In terms of teleoperation, I have focused on small vessels - and especially small vessels with low data transfer speeds, which is normal when you are out at sea”, Mårten explains.
To solve the problem of low-speed data transfer, Mårten had an operator remotely control ships in a virtual world that was a copy of the real world, much like in a computer game. This meant there was no need to transfer any information about the fixed environment such as islands, beacons or lighthouses, as they were already in place in the operator's virtual world.
“I then had the ship use image recognition to detect moving objects in the water and send real images that then appeared in the right place in the operator's virtual world.”
The user study showed that the technology works but also that the VR equipment offers several advantages, such as that the operator gains a good perception of the world around them and can, for example, predict collisions effectively.
What led you to do your doctorate?
“I enjoyed my job at Saab Kockums, where I worked extensively on the IT system of the future A26 submarine, but then the opportunity arose to learn more about AI, autonomous systems and other important technologies via WASP. Such technologies will shape our future in virtually all areas, so of course I couldn’t miss out on that opportunity.”
Wants greater focus on augmented reality
Mårten also explains his surprise that technologies such as Augmented Reality * (AR) and Virtual Reality (VR) are used relatively rarely in research in Sweden.
“These technologies fascinate me. Augmented Reality and Virtual Reality offer the opportunity to provide the user with any kind of interface, which means machines can support people in a fantastic way.
Both techniques were hyped up five years ago but feel relatively forgotten now. The area deserves more focus, both in terms of studying it as a technology and as an aid for studying other areas”, explains Lager.
What is happening with you and your results now?
“Saab Kockums is working actively on autonomous ships, two important elements of which are teleoperation and GPS-free navigation. Of course, the results are also shared with other parts of the WASP programme.
I myself am passionate about the technologies surrounding autonomous systems. I'm back at Saab Kockums full time and will try to use as much of this new expertise as possible. Next I will supervise a thesis in computer science at LTH looking at the annotation of images at sea for machine learning”, concludes the doctoral student.
* Wallenberg AI, Autonomous Systems and Software Program (WASP)
Text Jonas Wisbrant