16 April 2019
* This article originally published with The International Associates of Ports and Harbors
Over at least the next two or three decades, it is highly likely that ships with varying degrees of autonomy will be navigating alongside manned vessels. This raises new challenges for the maritime industry. Innovative solutions will be needed in the areas of collision avoidance, ship and shore communications, human factors, mariner and pilot training, cyber security, insurance, and liability.
In terms of autonomous navigation and collision avoidance, the introduction of new software and hardware on board vessels is bound to cause legal issues. Will, for instance, software suppliers risk becoming liable for navigation under regulations on prevention of collision at sea (COLREGS) based on algorithms developed by them?
The answers to this and other questions are unlikely to be addressed quickly given the relatively slow pace of the International Maritime Organization (IMO). In contrast, national legislators can move much faster as international conventions allow for national exemptions.
"One of the ... requirements for autonomous vessels is that they should navigate safely and avoid collisions"
Various nations are likely to follow the Norway example of paving the way for local deployment of autonomous vessels before they make their way into international waters. As the commercial case for small autonomous vessels is also strongest for coastal and inland waters vessels, disruption is likely to start in more restricted waterways such as ports and inland waterways.
One of the basic requirements for autonomous vessels is that they should navigate safely and avoid collisions with any other ships/obstacles or with a land mass. Furthermore, to operate harmoniously with other ships – either manned or unmanned – an autonomous vessel should behave in a manner like that of other ships in the vicinity. Since all manned craft are required to adhere to the coastguard COLREGs defined by the IMO, COLREGs-compliant behaviour will be an integral element of any autonomous vessel navigation system.Research by companies such as Rolls-Royce and BMT have demonstrated the application of new algorithms that allow existing COLREGs to be applied safely in a crewless environment.
This has been achieved by introducing artificial intelligence (AI)-based navigation systems that are able to enact the rules to avoid collision effectively, even when approaching manned vessels that interpret rules differently. Most of these software developments incorporate some form of multi-objective optimisation framework, using a technique called particle swarm optimisation (PSO). PSO has been adopted in wider computational science using the position and velocities of proximate objects to simulate social behaviour, such as movement of bird flocks or fish schools.
Much of this development by BMT has been funded through the Innovate UK Shared Waterspace Autonomous Navigation System (SWANS) programme. Working with a range of industrial partners, including UK pilots’ representatives and an autonomous vessel designer, BMT has adapted the REMBRANDT navigation simulator to incorporate COLREGS-compliant AI decision making for autonomous vessel hydrodynamic simulation. BMT is also involved in a parallel Innovate UK programme, Synthetic Imagery training for Machine Vision in Extreme Environments
(SIMVEE). This second programme is funding new machine vision developments to supplement the REMBRANDT COLREGS-compliant AI systems developed in SWANS. Both programmes together provide enhanced situational awareness and risk evaluation by fusing collision avoidance algorithms with vessel specific manoeuvrability and seakeeping simulation.
REMBRANDT is a Windows-based, time domain navigation, and seakeeping simulation software application. It is fully scalable, designed to run in full mission and single/multiple desktop modes, and has a variety of user controls needed to simulate conventional vessels and vessels with varying levels of autonomy.
It is installed in pilots’ and ship operators’ offices, training centres, and onboard ships. It is also used by autonomous vessel designers and operators.
REMBRANDT incorporates a wide range of user configurable controls to allow for a variety of environmental conditions. These include day and night screens, reduced visibility, multidirectional swells, and complex current flows. Visual topography, including bathymetric representation is automatically generated from the ENC chart. The system is AIS-enabled, allowing real-time displays of live traffic as part of an autonomous and manned vessel simulation.
The system includes a large database of fully calibrated hydrodynamic ship models of conventional and autonomous vessels.
Physical phenomena such as squat, bank effects, and ship-ship interaction are included to extend applications to collision avoidance in inland waterways. This new generation of software can be used to explore future scenarios that comprise a growing mix of manned, part autonomous, and fully autonomous vessels in and around port areas. This allows a better understanding of risks associated with
For example, the REMBRANDT software is currently being used as part of wider UK studies looking at the effect on the loss of individual vessel ship and shore communications and area-wide GPS.
The future challenge will be for AI navigation algorithms to detect and quantify a deliberate GPS spoofing and compensate accordingly. This will require real-time or near real-time processing of the available data. Other ongoing work involves simulating rudder and engine failure of autonomous vessels and identifying what actions are required from the autonomous vessel, while drift predictions serve to highlight effects on wider port operations and traffic.
Future scenario testing described above is currently ongoing within the Innovate UK SWANS and SIMVEE programmes and will be reported in detail over the coming months.