COLREGs-based Evasive Decision Making in Maritime Vehicles

Unmanned surface vehicles (USVs) are routinely being deployed in applications such as remote sensing, surveillance, coast patrolling and providing navigation and communication support to unmanned underwater vehicles (UUVs). In many instances, they are remotely operated to perform a specific mission in open or confined waters. The intelligence of these vehicles primarily reside in the navigation, guidance and control (NGC) systems design. Ideally, the vehicle needs to operate without any human intervention. This means that the vessel's onboard control system must be self reliant and able to maintain and supervise each onboard component. Having said that, even with the most advanced NGC design, the craft cannot be fully autonomous without the presence of an obstacle detection and avoidance (ODA) system. Studies have shown that, in manned vessels, more than 60% of casualties at sea are caused by collisions. In addition, it has been found that human error is a major contributing factor to those incidents. This could be due to an ever-decreasing number of crew members adding more responsibilities per person. For uninhabited surface craft, this cannot be overlooked as their collision with other manned ships could endanger human lives. Hence a human operator is always required to maintain a constant lookout for any potential obstacles. The aim of this project is to design and develop an ODA system primarily for an uninhabited marine craft. The evasive manoeuvres will be based on marine 'rules of the road' or collision regulations (COLREGs) on prevention of collision at sea defined by the International Marine Organisation. The ODA strategies that will be developed herein will be directly addressing one of the shortcomings of the current generation of unmanned surface vehicles, however, it could also be employed in manned vessels and other land based vehicles. The majority of the existing motion planning strategies either ignore parameters such as ship dynamics, environmental conditions and COLREGs or treat them on an ad-hoc basis. It is envisaged that this project will bridge this gap by considering these factors and thus automating this vital navigation component. In order to achieve this, multi-objective optimisation will be employed which will satisfy the criteria as specified above to determine a feasible path. A vision-based obstacle detection algorithm together with a laser range finder will also be investigated for close-range encounters and integrated with the path planning module.

This project covers a variety of research aspects. The most obvious is the formulation of path planning algorithms for unmanned marine vessels based on standardised rules. Hence any industry which specialises in unmanned marine craft will be a direct beneficiary. In fact, all the ship building companies could benefit from this research due to health and safety legislations for the onboard personnel. Hence by developing an ODA module, an attempt is being made to improve the security and safety of the onboard maritime personnel. This would provide the industry with cost cutting measures such as eliminating the use of a support boat and hence a reduction in manning. An unmanned vehicle with the ability to avoid obstacles could operate (especially in hazardous areas) for extended periods of time without any direct human supervision. This feature is useful for large data sampling operations, in particular for researchers working in the field of marine science and environment. In addition, by incorporating the marine `rules of the road' i.e. COLREGs, the algorithm can be implemented in manned vessels which would eliminate the human error and hence reduce the risk of collisions. Strategies will have to be devised for a vision-based obstacle detection system. These will be borrowed from the computer science field and are likely to be software-based. An automatic control system will also be implemented with the visual system providing the feedback to close the loop. In this case, researchers working in simple vision-based tracking systems as well as visual feedback systems will benefit from the algorithms developed in this project. Many land-based robots utilise camera-based sensing mechanisms and this proposal is likely to add another dimension to the existing research. Some automobile manufacturers are now providing collision detection and automatic breaking system as standard on their vehicles. This research would help those industries to develop additional safety features for their cars to be offered to their customers. The International Marine Organisation which has defined the COLREGs rules may also want to necessitate the use of automatic ODA systems as standard on all marine craft. Note that an automatic collision avoidance system is now installed as an integral part on all commercial aircraft and therefore similar legislation is required for maritime vessels as well. It is noted that the UK does not have any significant research programmes in autonomous ODA systems for ship's safety. Hence this proposal will surely add to UK competitiveness in maritime research worldwide. Although the proposed research will be directly addressing the shortcoming of the USVs used in civilian applications, nonetheless, the defence sector may inevitably get involved as more and more unmanned systems are being deployed in the battlefield worldwide to avoid human casualties. In the UK, there are two main research programmes initiated by the UK Ministry of Defence (MoD); MIMIR EVI and SWIMS. However, none of them is fully autonomous primarily due to the lack of an ODA module. Thus the UK MoD may find the results of this research of particular interest to them. The project will also provide excellent training opportunities for the research assistant who will gain valuable research and development experience improving his/her employability in academia as well as in the industry. Any research results will be disseminated in public domain in the usual way through publishing in relevant journals and presentations at international conferences including specialised marine research meetings. This will also help in forming global networks and generating new research avenues. Regular progress updates of the project will be notified through a website. A list of researchers mainly working in the UK on ODA systems for marine vehicles will be compiled and their research will be linked to the project website.