A review on improving the autonomy of unmanned surface vehicles through intelligent collision avoidance manoeuvres

Abstract

In recent years unmanned vehicles have grown in popularity, with an ever increasing number of applications in industry, the military and research within air, ground and marine domains. In particular, the challenges posed by unmanned marine vehicles in order to increase the level of autonomy include automatic obstacle avoidance and conformance with the Rules of the Road when navigating in the presence of other maritime traffic. The USV Master Plan which has been established for the US Navy outlines a list of objectives for improving autonomy in order to increase mission diversity and reduce the amount of supervisory intervention. This paper addresses the specific development needs based on notable research carried out to date, primarily with regard to navigation, guidance, control and motion planning. The integration of the International Regulations for Avoiding Collisions at Sea within the obstacle avoidance protocols seeks to prevent maritime accidents attributed to human error. The addition of these critical safety measures may be key to a future growth in demand for USVs, as they serve to pave the way for establishing legal policies for unmanned vessels.

Introduction

Following decades of research and development focused on the autonomy of aerial and underwater vehicles, comes resurging interest in Autonomous Surface Vehicles (ASVs). This is timely with the Defence Advanced Research Projects Agency’s (DARPA’s) announcement that it requires $3 billion in fiscal 2012 for projects involving ASV development for submarine tracking (Doyle, 2011). Extensive research has been carried out to date on Unmanned Underwater Vehicles (UUVs) due to implications for oil and gas exploration, deep sea pipeline monitoring and mine detection. Furthermore, research in Unmanned Aerial Vehicles (UAVs) has also largely overshadowed that of Unmanned Surface Vehicles (USVs) as they have become a superior tool in military strategies, e.g. the deployment of Predator and Reaper drones in Iraq and Afghanistan by the United States Air Force. UAVs are now heavily relied upon for surveillance, intelligence, search and rescue, reconnaissance and strike missions. Significant attention has also been given to Unmanned Ground Vehicles (UGVs) which encompass driverless cars, military tools for surveillance or bomb disposal and mechanical mule type vehicles for transporting heavy payloads. BigDog, for instance, is a well-known quadruped robot designed by Boston Dynamics to carry over 150 kg payload in difficult terrain and has since been adapted for autonomous navigation (Wooden et al., 2010). Technical similarities can be drawn between UGVs and USVs particularly in terms of the number of degrees of freedom and the need to safely operate in the presence of ambient traffic. However, motion control of underactuated ships in the presence of harsh environmental disturbances and an open navigational space poses far greater challenges.

It should be noted that semi-autonomy is highly typical of unmanned vehicles and in the past has often been favoured over full autonomy due to the diverse nature of missions. Further development of these uninhabited systems is required to extend their capabilities to include more complex and optimal mission planning in order to become less reliant on human interactions and subject to human error. The latest developments in the fields of artificial intelligence, advanced smart sensors, wireless networks and optimisation techniques now present greater opportunities than ever before for USVs and maritime technology on the whole (Corfield and Young, 2006).

The aim of this paper is to review and highlight the design aspects of the USV Navigation, Guidance and Control (NGC) system with respect to the International Regulations for Avoiding Collisions at Sea (COLREGs). The COLREGs describe potential collision scenarios such as crossing, head-on and overtaking, to mention a few (Commandant, 1999) and suggests possible manoeuvres to avoid a collision. Although the rules provide a set of guidelines for safe manoeuvring at sea, they were written for human navigators, who, based on their experience alter the course of the ship accordingly. This subjective nature of COLREGs is one of the major causes of ship collisions. It is estimated that human error contributes to between 89% and 96% of marine collisions, both active and latent, e.g. amateur manoeuvring (Rothblum, 2000). The RMS Titanic which infamously sank due to collision with an iceberg (Brown, 2000) as well as the Exxon Valdez oil spill disaster which struck a reef (Leacock, 2005) and the MV Doña Paz ferry disaster upon colliding with the MT Vector (Strauch, 2004) are amongst the most devastating peacetime maritime catastrophes of all time. Each of these collisions has been attributed to some form of human error, which could have been prevented. Poor judgement and failure to respond promptly are issues which can be resolved with an intelligent Obstacle Detection and Avoidance (ODA) system, substantially minimising risk.

The modern COLREGs were set out in 1972 by the International Maritime Organisation (IMO) as a set of guidelines for vessel encounters at sea, i.e. Rules of the Road. It is expected that all vessel operators comply with these regulations, which outline procedures for determining right of way and correct avoidance manoeuvres. Thus, if marine vessels can operate intelligently in accordance with these guidelines, many traffic-related accidents at sea caused by human error could be avoided. It should be stressed that motion planning for marine vehicles has been investigated in detail, however little attention is paid to COLREGs compliance. This lack of interest can be attributed to a number of factors. The foremost is the non-existence of any laws or regulations for the operation of USVs. Hence, until now the industry has not demanded USV development in preference to manned vehicles, primarily due to deficiencies in the decision-making abilities of an autonomous system.

In the last few years, noteworthy reviews have been carried out outlining USV motion planning and obstacle avoidance methodologies. One such paper (Statheros et al., 2008) describes various soft computing techniques for obstacle avoidance, mentioning only limited heuristic search methods. It does not address the USV control systems or COLREGs themselves and how to implement them in any detail. A recent review of close-range collision avoidance (Tam et al., 2009) gives a chronological account of approaches taken to the guidance problem and discusses related studies which tackle path planning with regard to collision avoidance. The authors did not discuss unmanned vehicles, but rather focussed on increasing the autonomy of manned craft to avoid human error during navigation when executing COLREGs. Another paper (Benjamin et al., 2006) describing behaviour-based control using Interval Programming discusses COLREGs protocol selection and action averaging accompanied by sea trials. However, only the four main COLREGs manoeuvre-based rules are investigated and the vessels used in the trials maintain wireless communication in order to determine the position of the other without relying on sensor information. It is concluded therefore, that existing literature has not sufficiently addressed problems establishing behavioural patterns based on obstacle classification (i.e. static, dynamic, geographic or other vessels) and complex encounter situations regarding COLREGs. It is also necessary to develop fail-safe methods for reactive avoidance, should the USV encounter unforeseen situations. This paper presents the recent developments in a wide range of fundamental topics relating to USVs and how the synthesis of these developments along with robust real-time motion planning can provide a comprehensive solution for a COLREGs compliant USV.

The following sections discuss the present state of USV development and highlight deficiencies and issues yet to be satisfactorily addressed. Through consideration of existing USV prototypes, NGC aspects and advanced motion planning techniques for the assimilation of COLREGs, this literature identifies key avenues to be explored for the accomplishment of an intelligent, autonomous USV.