Supporting intelligent and trustworthy maritime path planning decisions

https://doi.org/10.1016/j.ijhcs.2010.05.002Get rights and content

Abstract

The risk of maritime collisions and groundings has dramatically increased in the past five years despite technological advancements such as GPS-based navigation tools and electronic charts, which may add to, instead of reduce, workload. We propose that an automated path planning tool for littoral navigation can reduce workload and improve the overall system efficiency, particularly under time pressure. To this end, a maritime automated path planner (MAPP) was developed, incorporating information requirements developed from a cognitive task analysis, with special emphasis on designing for trust. Human-in-the-loop experimental results showed that MAPP was successful in reducing the time required to generate an optimized path, as well as reducing path lengths. The results also showed that while users gave the tool high acceptance ratings, they rated the MAPP as average for trust, which we propose is the appropriate level of trust for such a system.

Introduction

After a significant worldwide decline in serious navigation-related commercial maritime accidents from 1987 to 2002, the past five years have seen a significant spike in these accidents to levels not seen in more than 20 years (Richardsen, 2008). This recent trend is also reflected in the United States maritime operations with a recent similar spike in the US Navy accidents. Furthermore, in the past 25 years, the National Transportation Safety Board (NTSB) has investigated more than 50 collisions (with other ships and infrastructure such as bridges), and running aground incidents1. Collisions and groundings now account for 60% of the most costly maritime accidents, and in the current climate, a ship is twice as likely to be involved in a serious grounding and collision as compared to only five years ago (Richardsen, 2008). In all cases, human error is cited as a central cause, but other more latent causes that have been identified include a lack of situation awareness, an undersupply of crew, and high workload in navigation settings.

In coastal and high density traffic settings, when unexpected events occur that require immediate route re-planning, such as erratic movements of other maritime traffic, resultant plotting and charting can take several minutes, even with electronic displays. Navigation in congested and littoral regions causes significant navigator stress (Grabowski and Sanborn, 2003), as course replans and small adjustments occur frequently, increasing the navigator workload. Increases in mental workload, shown to be intricately linked with losses of situation awareness (Endsley, 1993), can lead to increased chances of allisions or collision (Grabowski and Sanborn, 2003).

Navigation is an inherently complex cognitive task since it typically involves multiple variables, many of which are uncertain (such as currents and other ships’ movements) that must be optimized to some objective function, often under time pressure (Hutchins, 1995). Moreover, navigation in coastal and especially harbor areas is especially demanding and in military settings can require up to ten different people: the navigator, assistant to the navigator, navigation plotter, navigation bearing recorder/timer, starboard and port pelorus (a compass attached to a sighting telescope) operators, restricted maneuvering helmsman, quartermaster of the watch, restricted maneuvering helmsman in after-steering and fathometer (depth) operator (Hutchins, 1995). Planning courses under time pressure, while not typically an issue for open ocean vessels, is particularly problematic for military littoral warships and fast patrol boats.

For ships equipped with the most modern technology (typically large commercial vessels), a merchant ship navigator can plot a course on an electronic map with zoom capability, which can be configured to show different layers of information such as weather and depths. In addition, some ships have radar systems that automatically identify and track other vessels in the water, such as the Automatic Identification System (AIS), which can transmit positions and speeds to an electronic display, if a ship has that capability. However, there is currently a lack of sufficient integration between the systems (Lee and Sanquist, 1996, Perrow, 1984), creating more demand on operators to process and integrate the data presented to them (Lee and Sanquist, 2000, Urbanski et al., 2008). Moreover, such electronic aids have been shown to be useful in low stress settings, but problematic in high stress scenarios (Grabowski and Sanborn, 2003). This problem is not just a maritime one, as the aviation industry has struggled with similar issues of increased workload with increased automation (Billings, 1997).

Not all maritime organizations use these electronic tools, and many ships, including most US military ships, still rely on the traditional paper chart method for navigation. The tools used in plotting ships’ path can include an alidade, which is a device that sights a landmark to measure the spatial relationship between the home ship and that landmark, the hoey, which is a one-arm protractor used in translating the angular relationship between the home ship and a landmark into a map bearing, parallel rulers, parallel motion protractors, compasses, distance scales and dividers for measuring distances (Hutchins, 1995). These devices all have degrees of error in accuracy and training and experience play a significant role in path quality and time to plot a path.

Time to plot a path can be a significant stressor in high workload navigation environments such as dense coastal settings. Personnel who plot courses on paper charts experience high mental workload when faced with the need to rapidly replan and chart in the face of new information, such as the presence of unexpected radar contacts or rapidly advancing weather. In some military operations, some ship captains will bring their vessels to a halt while attempting to replan a new course because of an unexpected event, which has clear negative mission implications, particularly in terms of time pressure.

We propose that both in paper and in electronic chart systems, what is needed to reduce workload in time-pressured navigation tasks is a decision support tool that integrates various sources of critical navigation information via an automated path planner and a user-centered visualization. Leveraging an intelligent path planning tool could greatly increase the accuracy and speed of planning a path, as well as reduce workload and error, and possibly manning requirements. While current electronic displays provide descriptive representations of the navigation environment and some limited predictions (e.g., where contacts are likely going), no tool currently in operational settings has effectively leveraged some form of intelligent decision support to aid humans in this demanding task.

Little research has investigated the use of automated path planning in maritime navigation. Rothgeb (2008) demonstrated that a fuzzy logic neural net could be used to identify high risk areas of transit given known contacts, as well as generate a recommended course based on safe areas. However, this research was focused on contact management, and not on the more holistic problem of path planning given additional variables such as weather and operator experience. In another related effort, Smierzchalski and Michalewicz (1998) developed an automatic path planner that accounts for surrounding contacts and their future positions, as well as physical characteristics of the ship such as weight, center of gravity and size of control surfaces. Their proposed algorithm, EP/N++, a variant of the evolutionary planner/navigator (EP/N) algorithm for mobile robots (Xiao et al., 1997), randomly generates acceptable paths for getting a ship from one point to another as a function of least cost. This randomized approach causes the solutions to be near-optimal at best, with the optimal solution traded for algorithm speed. This research is somewhat limited, as the proposed algorithm only takes into account up to three contacts in the vessel’s area of observation, and it does not address uncertainties for future contact positions. In addition, while the algorithms were tested in limited scenarios, no human-in-the-loop trials were ever conducted with any functional decision support tool based on the automated path planner.

Although automated path planning research in maritime navigation is limited, there is extensive research in the field of robotic path planning, which can provide useful insights to maritime navigation. Path planning in navigation is a large area of research in the computer science field (Winston., 1992), with significant research conducted in robotic path planning (e.g., (LaValle, 2006, Russell and Norvig, 2003, Thrun et al., 2005). As will be discussed in more depth in the next section, given this previous research, we elected to use the A algorithm for our automated path planner, which is an informed search method that can quickly find an optimal path to a destination, given our relatively constrained state space.

While an automated path-planner algorithm that is accurate and fast is critical for the maritime navigation problem, equally as important is the development of an operator decision support tool that maintains high operator performance, which also reduces mental workload. Users need to understand the limitations of such automated planning tools in order to know when they are correct (Layton et al., 1994). This issue raises another critical design consideration in the development of a maritime path planning tool, which is trust. New technologies in complex systems such as automated path planners in maritime settings face the challenge of gaining an acceptable level of trust from the operator before the system is accepted. When an operator has too little or too much trust in a system, the system has the potential to be dangerous. Distrust may lead to system disuse and over-trust may lead to inappropriate reliance on a system (Parasuraman and Riley, 1997).

Trust in intelligent decision support tools is affected by the reliability of the automation (Lee and Moray, 1992, Parasuraman and Riley, 1997). Research has shown that when automation reliability is in doubt, users’ trust in the automation significantly drops, causing more reliance in manual methods (Ruff et al., 2002), which then negates the usefulness of the automation. Moreover, the perceived reliability that a user attributes to automation is often related to how the information from the automation is conveyed to the user (Parasuraman and Riley, 1997). Increase in system uncertainty has also been shown to be a source of distrust for operators of automated systems (Uggirala et al., 2004). Uncertainty can stem from the environment, but it can also come from automated sensing and computation; so when designing an automated path planner, a designer needs to consider the impact of uncertainty from both these sources.

In summary, there is a clear need for more effective navigation decision support in maritime settings, particularly in coastal settings. While a few researchers have examined how different intelligent algorithms could be applied to limited aspects of this problem, no previous research has looked at the intersection of human–algorithm performance for the global maritime path planning task. Moreover, given the importance of user trust and acceptance in successful transitions of such technologies, any path planning decision support tool should be explicitly designed to mitigate uncertainty and enable user understanding of automation-generated solutions. To this end, in the next section we discuss the development of a maritime path planning decision support tool, primarily targeting coastal and high density traffic settings, that allows operators the ability to leverage automation to quickly generate multiple path options that account for contacts, weather and depth restrictions, as well as allowing them the ability to adjust their level of risk.

Section snippets

Decision support design

Since coastal navigation (in and around land, harbors and shipping channels) represents the most demanding phase of maritime navigation, it is important to determine the baseline processes and tasks involved in safe navigation in order to develop a comprehensive and functional decision support tool. In order to determine the display requirements for such a decision support tool, a cognitive task analysis (CTA) was conducted (Schraagen et al., 2000), which yielded information about the

Methods

Regardless of whether users prefer the more manual or automated modes in maritime path planning settings, we propose that MAPP should significantly reduce the workload of operators in terms of path quality and speed of path planning for maritime applications. MAPP automates the tedious work of actual plotting paths, while ensuring that routes do not violate depth, visibility, and obstacle constraints, which can be tailored by users. The ability for users to explicitly set constraints within the

Results and discussion

Using a 2×2 repeated measures ANOVA (α=.05), a statistical difference was found between MAPP and the paper chart with respect to time to plan a path (F(1,14)=92.47, p<.0001) (Fig. 3). There was no significant difference between the civilian and military solutions with respect to time, and there was no interaction effect. In general, time spent using MAPP to replan a path was just over 1.5 min, with the average time used for the paper chart for the same scenario at nearly 4 min.

With respect to

Conclusion

Despite recent technological advancements in maritime navigation settings such as GPS-based navigation tools and electronic charts, the risk of maritime collisions and groundings is increasing. Navigation is a cognitively challenging task due to the multivariate nature of the problem and high uncertainty in the environment, which is especially true when navigating in coastal areas in time-pressured settings. We propose that the inclusion of automated path planners in these systems can

Acknowledgements

This work was sponsored by Rite Solutions Inc., Assett Inc., Mikel Inc., and the Office of Naval Research. We would also like to thank Northeast Maritime Institute, the MIT NROTC detachment, the crew of the USS New Hampshire, and the anonymous reviewers whose comments significantly improved the paper.

References (38)

  • S.A. Guerlain et al.

    Interactive critiquing as a form of decision support: an empirical evaluation

    Human Factors

    (1999)
  • P. Hart et al.

    A formal basis for the heuristic determination of minimum cost paths

    IEEE Transactions of Systems Science and Cybernetics

    (1968)
  • E. Hutchins

    Cognition in the Wild

    (1995)
  • C. Kelly et al.

    Guidelines for Trust in Future ATM Systems: Measures (No. HRS/HSP-005-GUI-02)

    (2003)
  • S.M. LaValle

    Planning Algorithms

    (2006)
  • C. Layton et al.

    Design of a cooperative problem-solving system for en-route flight planning: an empirical evaluation

    Human Factors

    (1994)
  • J.D. Lee et al.

    Trust, control strategies and allocation of function in human–machine systems

    Ergonomics

    (1992)
  • J.D. Lee et al.

    Maritime automation

  • J.D. Lee et al.

    Augmenting the operator function model with cognitive operations: assessing the cognitive demands of technological innovation in ship navigation

    IEEE Transactions on Systems, Man, and Cybernetics—Part A

    (2000)
  • Cited by (0)

    View full text