Experimental Evaluations on Ship Autonomous ... - IEEE Xplore

374

IEEE JOURNAL OF OCEANIC ENGINEERING, VOL. 40, NO. 2, APRIL 2015

Experimental Evaluations on Ship Autonomous Navigation and Collision Avoidance by Intelligent Guidance Lokukaluge P. Perera, Victor Ferrari, Fernando P. Santos, Miguel A. Hinostroza, and Carlos Guedes Soares

Abstract—Experimental evaluations on autonomous navigation and collision avoidance of ship maneuvers by intelligent guidance are presented in this paper. These ship maneuvers are conducted on an experimental setup that consists of a navigation and control platform and a vessel model, in which the mathematical formulation presented is actually implemented. The mathematical formulation of the experimental setup is presented under three main sections: vessel traffic monitoring and information system, collision avoidance system, and vessel control system. The physical system of the experimental setup is presented under two main sections: vessel model and navigation and control platform. The vessel model consists of a scaled ship that has been used in this study. The navigation and control platform has been used to control the vessel model and that has been further divided under two sections: hardware structure and software architecture. Therefore, the physical system has been used to conduct ship maneuvers in autonomous navigation and collision avoidance experiments. Finally, several collision avoidance situations with two vessels are considered in this study. The vessel model is considered as the vessel (i.e., own vessel) that makes collision avoidance decisions/actions and the second vessel (i.e., target vessel) that does not take any collision avoidance actions is simulated. Finally, successful experimental results on several collision avoidance situations with two vessels are also presented in this study. Index Terms—Collision avoidance system, decision support system, intelligent guidance, ship collision avoidance, ship collision detection. Manuscript received June 24, 2013; revised December 01, 2013; accepted January 31, 2014. Date of publication April 08, 2014; date of current version April 10, 2015. The work of L. P. Perera was supported by the Doctoral Fellowship of the Portuguese Foundation for Science and Technology under Contract SFRH/BD/46270/2008. This work contributes to the project of “Methodology for ships maneuverability tests with self-propelled models,” which is supported by the Portuguese Foundation for Science and Technology under Contract PTDC/TRA/74332/2006. This work was presented in part at the 32nd International Conference on Ocean, Offshore and Arctic Engineering, Nantes, France, June 9–14, 2013. Associate Editor: K. Takagi. L. Prasad Perera was with the Centre for Marine Technology and Engineering (CENTEC), Instituto Superior Técnico, University of Lisbon, Lisbon 1049001, Portugal. He is now with Wärtsilä Finland Oy, Turku FIN-20811, Finland (e-mail: [email protected]). V. Ferrari is with the Centre for Marine Technology and Engineering (CENTEC), Instituto Superior Técnico, University of Lisbon, Lisbon 1049-001, Portugal and also with the Maritime Research Institute Netherlands (MARIN), Wageningen 7608 PM, The Netherlands (e-mail: [email protected]). F. P. Santos, M. A. Hinostroza, and C. Guedes Soares are with the Centre for Marine Technology and Engineering (CENTEC), Instituto Superior Técnico, University of Lisbon, Lisbon 1049-001, Portugal (e-mail: [email protected]; [email protected]; [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/JOE.2014.2304793

I. INTRODUCTION A. Maritime Safety

C

ONGESTED sea routes and various offshore operations enforce ships to make close encounter maneuvers, which may lead to some high-risk collision and near-collision situations. However, these issues can be countered by introducing safety training and safe ship handling procedures in the shipping industry. Even though these trainings and procedures associated with navigator`s experience could play an important role in safe ship navigation, they could also have some limitations due to human and economical constrains. Furthermore, even a welltrained and experienced navigator can make wrong navigation judgments, which can result in ship collisions with human casualties and environmental disasters. For example, even the decision-making process of an experienced navigator could be affected by unexpected situations with instrumentation and communication failures and losing vessel maneuverability conditions under various speed and environmental conditions. Therefore, as initiated by e-navigation [1], appropriate navigation aids should be facilitated to achieve the required safety levels in the shipping industry. The concept of e-navigation is introduced by the International Maritime Organization (IMO) and the International Association of Lighthouse Authorities (IALA) [2] for integrating present navigation technologies and introducing intelligent decision support capabilities to limit human subjective factors in the shipping industry. Furthermore, various studies to create next-generation command, communication, and control platforms that enhance wireless monitoring and control functions, including advanced decision support facilities to operate ships remotely under semi or fully autonomous conditions, have also been proposed [3]. Therefore, the main contribution in this study is also to support the concept of e-navigation by providing experimental results on ship autonomous navigation and collision avoidance based on intelligent guidance as further described in this paper. B. Ship Interactions In general, ship collision avoidance can be categorized under two types of environmental conditions [4]: the coastal phase and the oceanic phase. The coastal phase consists of collision avoidance among vessels in confined waters, and the oceanic phase consists of collision avoidance among vessels in open sea areas. The coastal phase comprises several navigation aids such

0364-9059 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

PERERA et al.: EXPERIMENTAL EVALUATIONS ON SHIP AUTONOMOUS NAVIGATION AND COLLISION AVOIDANCE

as traffic channels for highly dense maritime traffic regions and navigation guidance from ashore-based maritime traffic control stations [5], [6]. However, these navigation aids in the ocean phase have limited facilities; therefore, onboard decision support systems based on intelligent guidance should be developed as proposed in this study. Furthermore, collision avoidance in the ocean phase can be divided into two categories: long-range and close-proximity collision avoidance. Long-range collision avoidance is facilitated by the current law of the sea, the Convention on the International Regulations for Preventing Collisions at Sea (COLREGs) [7], formulated by the IMO. The reported data on ship collisions show that 56% of major maritime collisions involve violation of the COLREGs rules and regulations [8]. However, collision avoidance in close-proximity conditions is not facilitated by such rules and regulations, but the navigator's knowledge and experience can play an important role in those situations. Therefore, the maintenance of safe distance among vessels and other obstacles plays the most important role in improving ship safety in close-proximity conditions. This safe distance keeping among vessels, especially in overtakes and head-on situations, is also emphasized by the COLREGs [7], [9]. However, in close-proximity conditions, vessel-to-vessel interaction forces and moments are highly activated, and that could also be affected by vessels' orientations. These forces and moments could also result in involuntary course and speed changes, and that could eventually lead to various ship collision and near-miss situations. Therefore, these conditions have been extensively studied in the recent literature, as further discussed in this section. A simulation model of two ships passing under vessel-to-vessel interaction forces and moments on constant parallel course is presented in [10]. However, these vessel-to-vessel interactions may complicate ship navigation in a narrow channel [11] under vessel traffic [10], [12] and in shallow-water conditions [13], where bank effects, and weather and environmental conditions can also be influential. Therefore, ship behavior under these interaction forces and moments should be further considered to avoid collision and near-collision situations under close-proximity conditions in ship navigation. The vessel-to-vessel interaction forces in surge and sway may cause vessels to either attract or repulse from each other, and the yaw moments may cause them to either rotate toward or away from each other. However, these hydrodynamic forces and moments could also be affected by several factors: size, lateral and longitudinal separation distance, speeds, wetted hull shapes of the vessels, water depth and transverse distances from the channel banks when the vessels are close to shore, and weather conditions [14], [15]. These vessel interaction forces and moments have been calculated by several numerical methods in the recent literature. Methods for calculation of sway force and yaw moments for two vessels moving under close proximity in deep-water and shallow-water conditions are presented by Tuck and Newman [16] and Yeung [17], respectively. Other numerical methods for predicting such forces and moments for two vessels moving under close proximity are also presented by Huang and Chen [18], Sutulo and Guedes Soares [19], Xiang

375

and Faltinsen [20], King [21], Varyani and Krishnankutty [22] Xu et al. [23], Sutulo et al. [24], and Zhou et al. [25]. A theoretical method to predict the sinkage and trim conditions of two moving vessels under parallel meeting and overtaking conditions is presented by Gourlay [26]. Therefore, the possible actions against these vessel interaction forces and moments should be executed by the navigators as early as possible to avoid close encounter situations in ship navigation. C. Ship Collision Avoidance Decision-making processes in ship collision avoidance are presented in various studies [27]–[29]. Furthermore, several studies have been dedicated to the subject of collision avoidance maneuvers based on the following concepts: a clustered group of ships under close-proximity conditions [30]; state, parameter, and action optimization conditions [31]–[34]; safe navigational trajectories/routes selections [35]–[40]; case-based reasoning [41]; intelligent anticollision algorithms [42]; artificial force fields [43], [44]; fuzzy-logic-based systems [45]–[48]; IF–THEN-logic-based systems [49]; neurofuzzy networks [50]; and line of sight counteractions [51]. However, one should note that none of the above literature has presented proper collision avoidance experimental results, which in turn are the main contribution of this study. Therefore, this study proposes intelligent guidance for ship collision avoidance in e-navigation environment, which has been initiated in [52]–[56]. Furthermore, the proposed approach has been evaluated under an experimental setup, and the results on several collision avoidance situations of two vessels by means of autonomous maneuvers are also presented in this study. The experimental setup consists of a navigation and control platform, and a vessel model that is presented in a mathematical formulation as well as in an actual implementation. The navigation and control platform consists of controlling the ship model in autonomous and manual modes. The vessel model is used to create several collision avoidance situations and that is supported by an intelligent-guidance-based collision avoidance system. The proposed collision avoidance system capabilities of making multiple parallel collision avoidance decisions regarding several vessel collision situations are also illustrated. However, those decisions are executed as sequential actions to avoid complex collision situations in ship navigation in long-distance as well as in close-proximity conditions, which is discussed further. II. MATHEMATICAL FORMULATION A proposed mathematical formulation for ship navigation (i.e., autonomous navigation and collision avoidance) is presented in Fig. 1. It consists of three main systems: the vessel traffic monitoring and information system (VTMIS), the collision avoidance system (CAS), and the vessel control system (VCS). A. Vessel Traffic Monitoring and Information System The VTMIS facilitates by providing ship traffic information (i.e., ships' position, course, speed, acceleration, and trajectory

376


An extended Kalman filter (EKF)-based vessel state estimation (i.e., position, velocity, and acceleration) and navigational trajectory prediction process has been implemented on the VSETP module. This process is executed under the information given by the VDT module. The vessel traffic information (i.e., ship position, course, speed, etc.) transfers among ships and shore-based maritime authorities and that could be managed by the IVC module through a wireless network. An extensive study on the VTMIS considered in this paper is presented in [6].

B. Collision Avoidance System

Fig. 1. Mathematical formulation for ship navigation.

conditions) which can be used for autonomous navigation purposes as well as for collision avoidance among ships. Besides a scan sensor (i.e., radar/laser sensor), there are three main modules: vessel detection and tracking (VDT), vessel state estimation and trajectory prediction (VSETP), and intervessel communication (IVC). A scan sensor is used for detecting vessel positions. An artificial neural network (ANN)-based multivessel detection and tracking process has been implemented on the VDT module. It detects and tracks ships navigating in the scan sensor vicinity.

The CAS generates collision avoidance decisions/actions in a sequential format that can be executed in ship navigation. It is expected to have this system installed onboard a vessel that is called as the “own vessel” for the autonomous navigation and collision avoidance experiments. As presented in Fig. 1, the CAS consists of four modules: own-vessel communication (OVC), parallel decision making (PDM), sequential action formation (SAF), and collision risk assessment (CRA). The OVC module facilitates the communication of navigation information among ships and VTMISs. Such data are used by the PDM module to make collision avoidance decisions. The PDM module consists of a fuzzy-logic-based decision-making process that generates parallel collision avoidance decisions with respect to each ship that is under collision course with the own vessel. Furthermore, that creates course and speed change decisions for the own vessel, upon which decisions transfer to the SAF module to create proper collision avoidance actions. The rules, regulations, and expert navigational knowledge proposed by the COLREGs have been considered in the implementation of the PDM module. An extensive discussion on this module is presented in [54] and [56]. The CRA module evaluates the collision risk and the expected time until collision of each ship with respect to the own vessel based on navigation information from the OVC module. The evaluated collision risk information is transferred to the SAF module to arrange collision avoidance actions appropriately. An extensive discussion of the CRA module is presented in [57] and [58]. The SAF module “converts” the parallel collision avoidance decisions that are initially generated by the PDM module into sequential actions, considering the time until collision for each collision situation estimated by the CRA module. An extensive discussion on the SAF module is presented in [55] and [56]. Finally, the sequential collision avoidance actions that are organized by the SAF module are shared with the VCS. These actions can be categorized into two sets, course and speed controls, that will be implemented on the own vessel. The course and speed control collision avoidance actions with respect to each collision situation are executed under two subsystems: the steering control subsystem (SCS) and the speed control subsystem (SPS). The SCS and the SPS control ship course and speed conditions, respectively. An overview of the collision avoidance decision/action execution process under the PDM and SAF modules is discussed further in Sections II-C and II-D.


377

Fig. 3. SAF module.

D. Sequential Action Formulation Module

Fig. 2. PDM module.

C. Parallel Decision Making Module The PDM module consists of three main units (see Fig. 2): fuzzification, fuzzy rules, and defuzzification. The inputs of the OVC module, namely, range, bearing, course, and speed of other vessels (i.e., target vessels) for which there are collision courses with the own vessel at their respective instants, are fuzzified in the fuzzification unit. Accordingly, the following input fuzzy membership functions (FMFs) are considered: range FMF, speed ratio FMF, bearing FMF, and relative course FMF. Afterwards, the fuzzified results are transferred into the fuzzy rules unit for further analysis. Mamdani type IF–THEN rules are developed and inference via Min–Max norm is considered in the fuzzy rules unit. As mentioned before, the IF–THEN fuzzy rules are developed in accordance with the COLREGs rules and regulations. However, expert navigational knowledge is also considered in the fuzzy rules' development process. The course and speed change decisions to avoid the target vessels that have collision course with the own vessel are generated by the defuzzification unit. The inference results from the fuzzy rules' unit are defuzzified by considering the following output FMFs: course change FMF and speed change FMF. These FMFs generate course and speed change decisions that will be executed for collision avoidance in the own vessel. An extensive discussion on fuzzification, fuzzy rules, and defuzzification related to the present approach is presented in [54] and [56].

The SAF module that is modeled as a Bayesian network consists of four nodes/units (see Fig. 3): time until collision estimation (TUCE), collision risk estimation (CRE), collision avoidance action formulation (CAAF), and action delay. The main objective of the SAF module is to transform the parallel collision avoidance decisions that are generated by the PDM module into sequential actions that should be executed in the own vessel. This can be achieved by collecting the collision avoidance decisions and evaluating them using the time until collision with respect to each vessel that has collision course with the own vessel. Then, the final results (i.e., collision avoidance actions) are arranged as a sequential formation involving the course and speed change actions at the respective instants. The inputs of the SAF module are the collision decisions and the collision risk generated by the PDM and CRA modules, respectively. The main objectives of the TUCE and CRE nodes are to estimate the time until collision and the collision risk between the own and target vessels, respectively. The actions delay is designed to formulate the appropriate time interval for executing the speed and course change actions to avoid each collision situation. Therefore, the vessel collision avoidance actions are formulated by the CAAF node, and that is affected by the action delay and the CRE nodes, as presented in Fig. 3. Such actions can be divided into two sections: course and speed change actions which are initially generated as the collision avoidance decisions from the PDM module. Finally, these accumulated actions are implemented in the VCS of the own vessel for collision avoidance among vessels. An extensive discussion on the Bayesian-network-based sequential collision avoidance action formulation is presented in [55] and [56]. The decisions/actions that need to be taken by the own vessel to avoid various collision situations are summarized in Table I.

378


TABLE I COLLISION AVOIDANCE DECISIONS/ACTIONS

Fig. 5. Command and monitoring unit.

B. Hardware Structure

Fig. 4. Ship model in autonomous navigation and collision avoidance maneuvers.

III. EXPERIMENTAL SETUP The experimental setup, which is further discussed in this section, consists of a navigation and control platform, and the vessel model. This model consists of a scaled ship that has been used in this study; and the navigation and control platform has been used for autonomous and manual control of the vessel model. Therefore, the setup has been used to conduct ship maneuvers in autonomous navigation and collision avoidance experiments. The proposed CAS is implemented on the vessel model, which is considered as the own vessel. A. Vessel Model The vessel model considered in this study is presented in Fig. 4 and its characteristics are as follows: overall length of 2.590 m; length between perpendiculars of 2.450 m; breadth equal to 0.430 m; depth of 0.198 m; and estimated trail draft and displacement of 0.145 m and 115.6 kg, respectively. The vessel model is built in single skin glass reinforced polyester with plywood framings and that is controlled by the navigation and control platform, which can be divided into two sections: hardware structure and software architecture.

The hardware structure consists of all sensors and actuators that are used in the navigation and control platform. This structure is further divided into the following two units: the command and monitoring unit (CMU) and the communication and control unit (CCU). The main objective of the CMU is to facilitate manual and autonomous control of the vessel model provided by the human–machine interface (HMI), as presented in Fig. 5. The CCU is implemented on a shore-based station and that consists of several instrumentations such as a laptop computer, a Global Positioning System (GPS) unit, and an industrial WiFi unit. A laptop computer, used as HMI, is connected to an industrial WiFi unit for communicating with the CCU. The computer works as a data display interface as well as an automatic and manual control unit for the vessel model. Furthermore, the above discussed VTMIS is implemented on the laptop computer under MATLAB/LABVIEW software. The VTMIS is simulated to obtain the target vessel behavior that is in a collision course with the vessel model (i.e., own vessel). The data are forwarded to the CAS for collision avoidance decisions/actions. One should note that the CAS is implemented on the vessel model (i.e., own vessel). The GPS unit is used in the CMU for position measurements of the vessel model. The complete GPS system has two units, namely, a base station and a rover station which improve the position accuracy of the vessel model. The base GPS station unit acts as a stationary reference that transmits known stationary position correction signals for the rover GPS station which is located in the ship model. The WiFi unit is used for communication between the ashore-based CMU and the onboard CCU. The proposed CCU is implemented on the vessel model as presented in Fig. 6. The main objective of the CCU is to execute the fuzzy-Bayesian-based decision/action execution process (i.e., the CAS), as described in Section II. That is associated with the course and speed change actions and is facilitated by the following instrumentation: two CompacRIO units, an industrial Ethernet switch (IES), a laptop computer, a GPS unit, an inertial measurement system (IMS), a WiFi unit, and two direct current (dc) motors. Two CompactRIOs with input/output (I/O) modules are used in the CCU. One collects digital data from the IMS and GPS units. Other unit is connected to the steering and speed control subsystems of the vessel model to control the actuations of the rudder and propeller assembled to two dc motors. Furthermore, both CompactRIOs are connected through the IES.


Fig. 6. Communication and control unit.

379

unit consists of the following sensors: a magnetometer, an accelerometer, rate gyro, and a GPS receiver. The IMS is capable of measuring the following: three-axes angles of heading, roll, and pitch; three-axes angular velocities of heading, roll, and pitch; and three-axes linear accelerations of surge, sway, and heave. The internal GPS receiver in the IMS unit measures the vessel model position facilitated with WAAS capabilities. The IES is used in the CCU as a communication gateway among sensors, actuators, and CompactRIO units. Furthermore, the above discussed CAS is implemented on the laptop under MATLAB/LABVIEW software. The CAS formulates the collision avoidance actions that are based on the target vessel collision course and speed information that is given by the shore-based VTMIS (i.e., other laptop computer). Another WiFi unit is used for communication between the ashore-based CMU and the onboard CCU connected through the IES. The proposed vessel model has two control subsystems incorporated: the steering control subsystem (SCS) and the speed control subsystem (SPS). The SCS is associated to the rudder control system, and its main objective is to maintain the appropriate vessel course during its maneuvers. The SPS is associated to the propeller control system, and its main objective is to maintain appropriate vessel speed during its maneuvers. The proportional–integral–derivative (PID) controllers are used for both propeller revolutions per minute (RPM) and rudder positions controls. The course and speed change collision avoidance actions that are generated by the CAS are executed in these subsystems. C. Software Architecture

Fig. 7. Trajectories of collision situation I.

The laptop computer in the vessel model is used to record and store the digital data collected from the IMS and GPS units through the IES connected to the CompactRIO. Another onboard GPS unit is used in the CCU to accurately estimate the position of the vessel model as discussed previously. The IMS

The software architecture in this experimental platform is mainly developed under LABVIEW and MATLAB programs consisting of several loops: a field-programmable gate array (FPGA) loop, a real-time control loop, a CAS loop, and a TCP/IP loop. The FPGA loop aims at collecting data from the sensors (i.e., GPS and IMS units) and controlling the actuations of the steering and speed subsystems that have been programmed under LABVIEW. The associated PID controllers for the steering and speed control subsystems are implemented under a real-time control loop (i.e., the internal deterministic control loop) that has the highest responsiveness, determinism, and priority with comparison to other software loops. The data processing and record saving for the respective sensors are implemented under the internal nondeterministic loop that has lower priority in comparison to the deterministic control loop. The CAS loop consists of the proposed fuzzy-Bayesian-based decision/action execution process for collision avoidance among vessels, generating required collision avoidance actions for the vessel model with respect to the simulated target vessel, with which it is in a collision course. These actions are executed under the real-time control loop associated with the steering and speed control subsystems on the vessel model. The TCP/IP loop is related to the communication between shore-based CMU and the VTMIS, being implemented under wireless communication through the industrial WiFi unit. Furthermore, an extensive discussion on the experimental platform is presented in [59] and [60].

380


Fig. 8. Vessels' positions, speed, and course control of collision situation I.

IV. EXPERIMENTAL RESULTS The collision avoidance experiments were conducted on the lake of “Campo Grande” in Lisbon, Portugal. These experiments involved various autonomous maneuvers and collision avoidance situations by the vessel model. The collision avoidance experiments were conducted on the ship model with the onboard CAS, as described in Section II. However, a scaled version of the CAS has been used during these experiments due to the practical difficulties (i.e., wind and wave conditions) faced by the vessel model. Furthermore, the vessel model position data were collected from the GPS system that has two units (i.e., a base station and a rover station) due to its higher accuracy (i.e., 1 cm). However, the additional sensor data (i.e., IMS sensor) have encountered lower accuracy due to the sensor noise and slow speed conditions of the vessel model. The vessel model with the CAS was represented as the own vessel, and a target vessel in a collision course with the own vessel was simulated. However, it was observed that the formulation of a collision situation between two ships is extremely difficult to achieve due to the ship model sudden course change and speed variations caused by the wind and wave conditions. Therefore, an additional algorithm was developed to simulate the target vessel maneuvers. This target vessel algorithm consists of the following sequential steps: the initial target vessel position should be assigned near the own vessel navigation route; then, the algorithm creates proper collision course between the own and target vessels by

Fig. 9. Trajectories of collision situation II.

considering various target vessel course conditions. As an example, the target vessel course changes from 0 to 360 with 1 intervals under the speed condition that is approximated to the own vessel speed; when the algorithm finds a collision course


381

Fig. 10. Vessels' positions, speed, and course control of collision situation II.

between the two vessels during such course changes, it executes that course with the appropriate speed conditions as the target vessel. Consequently, this target vessel algorithm generates an appropriate collision situation between both vessels, and that information is forwarded to the CAS. Therefore, the own vessel takes appropriate decisions/actions to avoid the collision situation. Several collision situations were created by the proposed target vessel algorithm, and the appropriate actions taken by the vessel model were observed under such conditions. The CAS was implemented on the laptop computer onboard the own vessel, as described in Section II. It is assumed that the target vessel is moving at constant speed and course conditions and does not honor any navigational rules and regulations (i.e., COLREGs). One should note that such speed and course conditions are considered in these experiments to keep the consistence in the collision situation between two vessels [56]. A. Collision Situation I The first set of experimental results of a collision situation between two vessels is presented in Figs. 7 and 8. As presented in Fig. 7, the vessel model (i.e., own vessel) and the target vessel start to navigate from the positions (0 [m], 0 [m]) and (10 [m], 20 [m]), respectively. One should note that the spiral section of the target vessel trajectory, which is near its initial position, rep-

Fig. 11. Trajectories of collision situation III.

resents the algorithm that has been used to capture the collision course between two vessels, as described previously.

382


Fig. 12. Vessels' positions, speed, and course control of collision situation III.

As presented in Fig. 7, the vessel model (i.e., the own vessel) has observed a possible collision situation in which the target vessel is approaching for a crossing situation from starboard. One should note that in accordance with the COLREGs rules and regulations the vessel model is in a “give way” situation, in which it has lower priority for navigation in a collision situation, and the target vessel is in a “stand on” situation. Therefore, the early collision avoidance actions to avoid the collision situations are executed by the vessel model. The respective own and target vessel – coordinates with respect to time are presented in the top plots of Fig. 8. The collision avoidance decisions (see Table I) of altering course to starboard and increasing speed at the first stage, altering course to port and increasing speed at the second stage, altering course to starboard and increasing speed at the third stage, which have been taken by the ship model, are presented in the bottom plots of Fig. 8. B. Collision Situation II The second set of experimental results of a collision situation with two vessels is presented in Figs. 9 and 10. The vessel model and the target vessel start to navigate from the positions (0 [m], 0 [m]) and (20 [m], 10 [m]), respectively. As presented in Fig. 9, the vessel model (i.e., the own vessel) has observed a possible collision situation in which the target vessel is approaching for a crossing situation from port. With respect to the COLREGs rules and regulations, the vessel model is in a “stand on” situation, thus it has higher priority for navigating, and the target

Fig. 13. Trajectories of collision situation IV.

vessel is in a “give way” situation, meaning a lower priority for navigating in a collision situation. Since the target vessel has the “give way” situation and the vessel does not take any actions to


383

Fig. 14. Vessels' positions, speed, and course control of collision situation IV.

avoid the collision situation, the vessel model is forced to take appropriate actions in that sense. It must be noted that the own vessel in a “give way” situation takes collision avoidance actions earlier than in a “stand on” situation. In this context, the distance between two vessels has been considered for the decision-making process in the CAS. Therefore, the vessel in a “stand on” situation could make crash-stop type maneuvers to avoid a collision situation due to inadequate actions from the target vessel. These situations have been categorized as critical collision conditions, and above discussed concepts have been adopted by the CAS, as further described in [48]. The respective own and target vessel – coordinates with respect to time are presented in the top plots of Fig. 10. The collision avoidance decisions (see Table I) of altering course to starboard and reducing speed in the first stage and altering course to starboard and increasing speed at the second stage that have been taken by the vessel model are presented in the bottom plots of Fig. 10. C. Collision Situation III The third set of experimental results of a collision situation with two vessels is presented in Figs. 11 and 12. The vessel model and the target vessel start to navigate from the positions (0 [m], 0 [m]) and ( 10 [m], 20 [m]), respectively. As presented in Fig. 11, the vessel model (i.e., the own vessel) has observed a possible collision situation in which the target vessel is approaching for a head-on collision situation from starboard.

Fig. 15. Trajectories of collision situation V.

However, the target vessel does not take any action in this close encounter situation; therefore, the vessel model is forced to take appropriate actions to avoid the collision situation. Even though

384


Fig. 16. Vessels' positions, speed, and course control of collision situation V.

the vessels should pass port to port in a head-on collision situation in accordance with the COLREGs [7], [9], this close encounter situation with an altering course to starboard by the own vessel could increase the collision risk. Therefore, this safe distance keeping among vessels, especially in close encounter situations, is emphasized by the COLREGs and is implemented by the vessel model in this situation. The respective own and target vessel – coordinates with respect to time are presented in the top plots of Fig. 12. The collision avoidance decisions (see Table I) of altering course to port and increasing speed taken by the own vessel are presented in the bottom plots of Fig. 12. D. Collision Situation IV The fourth set of experimental results of a collision situation with two vessels is presented in Figs. 13 and 14. The vessel model and the target vessel start to navigate from the positions (0 [m], 0 [m]) and (10 [m], 20 [m]), respectively. As presented in Fig. 13, the vessel model (i.e., the own vessel) has observed a possible collision situation in which the target vessel is approaching for a crossing situation from starboard. According to the COLREGs rules and regulations, the own and target vessels are in “give way” and “stand on” situations, respectively. Therefore, the early collision avoidance actions to avoid the collision situations are executed by the vessel model. The respective own and target vessel – coordinates with respect to time are presented in the top plots of Fig. 14.

The collision avoidance decisions (see Table I) of altering course to starboard and increasing speed at the first stage, altering course to port and increasing speed at the second stage, and altering course to starboard and increasing speed at the third stage, which have been taken by the own vessel, are presented in the bottom plots of Fig. 14. Considerable similarities can also be noted on the collision avoidance actions executed by the vessel model in the collision situations I and IV. E. Collision Situation V The fifth set of experimental results of a collision situation with two vessels is presented in Figs. 15 and 16. The vessel model and the target vessel start to navigate from the positions (0 [m], 0 [m]) and (30 [m], 10 [m]), respectively. As presented in Fig. 15, the vessel model (i.e., own vessel) has observed a possible collision situation in which the target vessel is approaching for a head-on collision situation from port. However, the target vessel does not take any action in this close encounter situation, therefore the ship model is forced to take appropriate actions to avoid the collision situation. Even though the vessels should pass in port to port in a head-on collision situation in accordance with the COLREGs [7], [9], this close encounter situation with altering course to starboard by the own vessel could also increase the collision risk. Therefore, the safe distance keeping among vessels, especially in close encounter situations, is emphasized by the COLREGs and is also considered in this situation. The respective own and target


385

TABLE II APPROACHES FOR POSSIBLE COLLISIONS AND RESPECTIVE COLREGS RULES AND REGULATIONS

TABLE III SHIP MODEL COURSE AND SPEED COLLISION AVOIDANCE DECISIONS

vessel – coordinates with respect to time are presented in the top plots of Fig. 16. The collision avoidance decisions (see Table I) of altering course to port and increasing speed at the first stage and altering course to starboard and increasing speed at the second stage, which have been taken by the ship model, are presented in the bottom plots of Fig. 16. In the experimental results, one can observe five possible collision situations that the vessel model (i.e., the own vessel) has identified. This is resumed in Table II along with the proper COLREGs rules and regulations for the vessel model and the target vessel. It is assumed that the target vessel does not honor any navigational rules and regulations (i.e., COLREGs). Even though the purpose of these experiments is for the vessel model to always take appropriate collision avoidance actions, the fact is that in some situations the own vessel may not be the one who has the priority to take collision avoidance actions in the first place, according to the COLREGs rules and regulations. These collision avoidance actions with respect to the COLREGs rules and regulations have been summarized in Table II, and the action stages that have been executed by the vessel model are summarized in Table III for each situation.

V. CONCLUSION AND FUTURE DEVELOPMENT Experimental evaluations on several collision avoidance situations between two vessels have been presented in this study. Considering the experimental results, it can be concluded that the ship model has taken appropriate collision avoidance decisions and actions to reduce the collision risk between both vessels. Therefore, the reported successful experimental results using a vessel model show the superior capabilities of the proposed intelligent-guidance-based collision avoidance system, and this is the main contribution of this study. Furthermore, the tools and techniques presented in the study can be used for the e-navigation strategy, in which one could introduce autonomous navigation and collision avoidance functionalities in the shipping industry. However, the implementation of collision detection and avoidance among multiple vessels in the experimental platform is still a challenge for the future, where the proposed system should be further developed. Therefore, a complete version of the proposed collision detection and avoidance under multivessel situations will be a further development of this study. REFERENCES [1] N. Ward and S. Leighton, “Collision avoidance in the e-navigation environment,” in Proc. 17th Conf. Int. Assoc. Marine Aids Navig. Lighthouse Authorities, Cape Town, South Africa, 2010, pp. 4–10. [2] IMO, “Development of an e-navigation strategy,” Report of the correspondence group on e-navigation, NAV/53/13, 2007. [3] MUNIN, “Maritime unmanned navigation through intelligence in networks,” 2013 [Online]. Available: http://www.unmanned-ship.org/ munin/ [4] R. S. Burns, G. Blackwell, and S. Calvert, “An automatic guidance, navigation and collision avoidance system for ships at sea,” in Proc. IEE Colloq. Control Mar. Ind., Jan. 1988, pp. 3/1–3/3. [5] A. Janex, “Concept of a collision-avoidance system for marine navigation,” in Proc. OCEANS Conf., Sep. 1990, pp. 458–463. [6] L. Perera, P. Oliveira, and C. Guedes Soares, “Vessel detection, tracking, state estimation and navigation trajectory prediction for the vessel traffic monitoring and information process,” IEEE Trans. Intell. Transp. Syst., vol. 13, no. 3, pp. 1188–1200, 2012. [7] IMO, “Convention on the International Regulations for Preventing Collisions at Sea (COLREGs),” 1972. [8] T. Statheros, G. Howells, and K. McDonald-Maier, “Autonomous ship collision avoidance navigation concepts, technologies and techniques,” J. Navig., vol. 61, pp. 129–142, 2008.

386


[9] L. Perera, J. P. Carvalho, and C. Guedes Soares, “Autonomous guidance and navigation based on the COLREGs rules and regulations of collision avoidance,” in Advanced Ship Design for Pollution Prevention, C. Guedes Soares and J. Parunov, Eds. London, U.K.: Taylor & Francis, 2010, pp. 129–142. [10] J. C. Hewlett, “Ship Navigation Simulation Study, Houston-Galveston Navigation Channels,” U.S. Army Engineer District, Galveston, TX, USA, AD-A279 016, 1994. [11] Y. Yavin, C. Frangos, G. Zilman, and T. Miloh, “Computation of feasible command strategies for the navigation of a ship in a narrow zigzag channel,” Comput. Math. Appl., vol. 30, no. 10, pp. 79–101, 1995. [12] K. S. Varyani, A. Thavalingam, and P. Krishnankutty, “New generic mathematical model to predict hydrodynamic interaction effects for overtaking maneuvers in simulations,” J. Mar. Sci. Technol., vol. 9, pp. 24–31, 2004. [13] X. Zhou, S. Sutulo, and C. Guedes Soares, “Computation of ship-toship interaction forces by a 3d potential flow panel method in finite water depth,” in Proc. 29th Int. Conf. Ocean Offshore Arctic Eng., Shanghai, China, 2010, OMAE2010-20497. [14] S. Sutulo and C. Guedes Soares, “A unified nonlinear mathematical model for simulating ship manoeuvring and seakeeping in regular waves,” in Proc. Int. Conf. Mar. Simul. Ship Manoeuv., Terschelling, The Netherlands, 2006. [15] R. Skejic and M. Faltinsen, “A unified seakeeping and maneuvering analysis of two interacting ships,” in Proc. 2nd Int. Conf. Mar. Res. Transp., Naples, Italy, Jun. 2007, pp. 209–218. [16] E. O. Tuck and J. N. Newman, “Hydrodynamic interaction between ships,” in Proc. 10th Symp. Naval Hydrodyn., Cambridge, MA, USA, 1974, pp. 35–58. [17] R. W. Yeung, “On the interaction of slender ships in shallow water,” J. Fluid Mech., vol. 85, pp. 143–159, 1978. [18] E. T. Huang and H. C. Chen, “Passing ship effects on moored vessels at piers,” in Proc. First Symp. CSLC Prevention, Long Beach, CA, USA, 2006. [19] S. Sutulo and C. Guedes Soares, “Simulation of the hydrodynamic interaction forces in close-proximity manoeuvring,” in Proc. 27th Int. Conf. Offshore Mech. Arctic Eng., Estoril, Portugal, 2008, OMAE2008-57938. [20] X. Xiang and O. M. Faltinsen, “Maneuvering of two interacting ships in calm water,” in Proc. 11th Int. Symp. Practical Design Ships Other Floating Struct., Rio da Janeiro, Brazil, 2010, pp. 161–171. [21] G. W. King, “Unsteady hydrodynamic interactions between ships in shallow water,” in Proc. 6th Austral. Hydraul Fluid Mech. Conf., Adelaide, Australia, 1977, pp. 291–293. [22] K. S. Varyani and P. Krishnankutty, “Modification of ship hydrodynamic interaction forces and moment by underwater ship geometry,” Ocean Eng., vol. 33, pp. 1090–1104, 2006. [23] Y. Xu, Z. Zou, M. Liu, and K. S. Varyani, “Study on critical uncontrollable hydrodynamic interaction between ships,” in Proc. 18th Int. Offshore Polar Eng. Conf., Vancouver, BC, Canada, 2008. [24] S. Sutulo, C. Guedes Soares, and J. F. Otzen, “Validation of potential-flow estimation of interaction forces acting upon ship hulls in sideto-side motion,” J. Ship Res., vol. 56, no. 3, pp. 129–145, 2012. [25] X. Zhou, S. Sutulo, and C. Guedes Soares, “Computation of ship hydrodynamic interaction forces in restricted waters using potential theory,” J. Mar. Sci. Appl., vol. 11, pp. 265–275, 2012. [26] T. Gouraly, “Sinkage and trim of two ships passing each other on parallel course,” Ocean Eng., vol. 36, no. 14, pp. 1119–1127, 2009. [27] C. Chauvin and S. Lardjane, “Decision making and strategies in an interaction situation: Collision avoidance at sea,” Transp. Res. F, vol. 11, no. 4, pp. 259–262, 2008. [28] Y. Y. Wang, A. K. Debnath, and H. C. Chin, “Modeling collision avoidance decisions in navigation,” in Proc. 10th Asian Conf. Mar. Simul. Simul. Res., Keelung, Taiwan, Jun. 2010, pp. 191–198. [29] M. R. Benjamin and J. A. Curcio, “COLREGS-based navigation of autonomous marine vehicles,” in Proc. IEEE/OES Autonom. Underwater Veh., Jun. 2004, pp. 32–39. [30] A. Kawaguchi, X. Xiong, M. Inaishi, and H. Kondo, “A computerized navigation support for maneuvering clustered ship groups in close proximity,” Syst. Cybern. Inf., vol. 3, no. 3, pp. 46–56, 2006. [31] A. Miele, T. Wang, C. S. Chao, and J. B. Dabney, “Optimal control of a ship for collision avoidance maneuvers,” J. Optim. Theory Appl., vol. 103, no. 3, pp. 495–519, Dec. 1999. [32] T. Miloh and S. D. Sharma, “Critical maneuvers for avoiding collision at sea,” Inst. Schiffbau der Univ. Hamburg, Hamburg, Germany, Apr. 1975.

[33] S. D. Sharma, “On ship maneuverability and collision avoidance,” in Proc. 2nd West Eur. Mar. Technol. Conf., London, U.K., May 1977, pp. 1–29. [34] T. L. Vincent, “Collision avoidance at sea,” in Differential Games and Applications, ser. Lecture Notes in Control and Information Sciences, P. Hagedorn, Ed. Berlin, Germany: Springer-Verlag, 1977, vol. 3, pp. 205–221. [35] R. Smierzchalski and Z. Michalewicz, “Modeling of ship trajectory in collision situations by an evolutionary algorithm,” IEEE Trans. Evol. Comput., vol. 4, no. 3, pp. 227–241, Sep. 2000. [36] M. Ito, F. Zhang, and N. Yoshida, “Collision avoidance control of ship with genetic algorithm,” in Proc. IEEE Int. Conf. Control Appl., Aug. 1999, pp. 1791–1796. [37] J. Froese and S. Mathes, “Computer-assisted collision avoidance using ARPA and ECDIS,” German J. Hydrogr., vol. 49, no. 4, pp. 519–529, 1997. [38] X. Zeng, M. Ito, and E. Shimizu, “Building an automatic control system of maneuvering ship in collision situation with genetic algorithms,” in Proc. Amer. Control Conf., Jun. 2001, pp. 2852–2853. [39] X. Hong, C. J. Harris, and P. A. Wilson, “Autonomous ship collision free trajectory navigation and control algorithms,” in Proc. 7th IEEE Int. Conf. Emerging Technol. Factory Autom., 1999, vol. 2, pp. 923–929. [40] X. D. Cheng, Z. Y. Liu, and X. T. Zhang, “Trajectory optimization for ship collision avoidance system using genetic algorithm,” in Proc. OCEANS Conf. Asia Pacific, May 2006, DOI: 10.1109/OCEANSAP. 2006.4393976. [41] Y. Liu and H. Liu, “Case learning base on evaluation system for vessel collision avoidance,” in Proc. 5th Int. Conf. Mach. Learn. Cybern., Aug. 2006, pp. 2064–2069. [42] S. Yang, L. Li, Y. Suo, and G. Chen, “Study on construction of simulation platform for vessel automatic anti-collision and its test methods,” in Proc. IEEE Int. Conf. Autom. Logistics, Aug. 2007, pp. 2414–2419. [43] S. Lee, K. Kwon, and J. Joh, “A fuzzy logic for autonomous navigation of marine vehicles satisfying COLREG guidelines,” Int. J. Control Autom. Syst., vol. 2, no. 2, pp. 171–181, Jun. 2004. [44] Y. Xue, B. S. Lee, and D. Han, “Automatic collision avoidance of ships,” J. Eng. Maritime Environ., vol. 223, no. 1, pp. 33–46, 2009. [45] Y. I. Lee and Y. G. Kim, “A collision avoidance system for autonomous ship using fuzzy relational products and COLREGs,” in Proc. 5th Int. Conf. Intell. Data Eng. Autom. Learn., Exeter, U.K., Aug. 25–27, 2004, pp. 247–252. [46] K. Hasegawa, “Automatic collision avoidance system for ship using fuzzy control,” in Proc. 8th Ship Control Syst. Symp., 1987, pp. 234–258. [47] Y. Zhuo and G. E. Hearn, “A ship based intelligent anti-collision decision-making support system utilizing trial manoeuvers,” in Proc. Chin. Control Decision Conf., Jul. 2008, pp. 3982–3987. [48] L. P. Perera, J. P. Carvalho, and C. Guedes Soares, “Fuzzy-logic based decision making system for collision avoidance of ocean navigation under critical collision conditions,” J. Mar. Sci. Technol., vol. 16, pp. 84–99, 2011. [49] G. P. Smeaton and F. P. Coenen, “Developing an intelligent marine navigation system,” Comput. Control Eng. J., vol. 1, no. 2, pp. 95–103, Mar. 1990. [50] P. A. Wilson, C. J. Harris, and X. Hong, “A line of sight counteraction navigation algorithm for ship encounter collision avoidance,” J. Navig., vol. 56, no. 1, pp. 111–121, Jan. 2003. [51] C. J. Harris, X. Hong, and P. A. Wilson, “An intelligent guidance and control system for ship obstacle avoidance,” Proc. IMechE I/J. Syst. Control Eng., vol. 213, no. 14, pp. 311–320, 1999. [52] L. P. Perera, J. P. Carvalho, and C. Guedes Soares, “Solutions to the failures and limitations of Mamdani fuzzy inference in ship navigation,” IEEE Trans. Veh. Technol., 2013, DOI: 10.1109/TVT.2013.228830. [53] L. P. Perera, J. P. Carvalho, and C. Guedes Soares, “Mamdani type fuzzy inference failures in navigation,” in Proc. 9th IEEE Int. Conf. Ind. Inf., Lisbon, Portugal, Jul. 2011, pp. 26–29. [54] L. P. Perera, J. P. Carvalho, and C. Guedes Soares, “Fuzzy-logic based parallel collisions avoidance decision formulation for an ocean navigational system,” in Proc. 8th IFAC Conf. Control Appl. Mar. Syst., Rostock, Germany, Sep. 2010, pp. 295–300. [55] L. P. Perera, J. P. Carvalho, and C. Guedes Soares, “Bayesian network based sequential collision avoidance action execution for an ocean navigational system,” in Proc. 8th IFAC Conf. Control Appl. Mar. Syst., Rostock, Germany, Sep. 2010, pp. 301–306.


[56] L. P. Perera, J. P. Carvalho, and C. Guedes Soares, “Intelligent ocean navigation & fuzzy-Bayesian decision-action formulation,” IEEE J. Ocean. Eng., vol. 37, no. 2, pp. 204–219, Apr. 2012. [57] L. P. Perera and C. Guedes Soares, “Vector-product based collision estimation and detection in e-navigation,” in Proc. 9th IFAC Conf. Manoeuv. Control Marine Craft, Arenzano, Italy, Sep. 2012. [58] L. P. Perera and C. Guedes Soares, “Detections of potential collision situations by relative motions of vessels under parameter uncertainties,” in Sustainable Maritime Transportation and Exploitation of Sea Resources, E. Rizzuto and C. Guedes Soares, Eds. London, U.K.: Taylor & Francis, 2012, pp. 705–713. [59] L. P. Perera, L. Moreira, F. P. Santos, V. Ferrari, S. Sutulo, and C. Guedes Soares, “A navigation and control platform for real-time manoeuvring of autonomous ship models,” in Proc. 9th IFAC Conf. Manoeuv. Control Mar. Craft, Arenzano, Italy, 2012. [60] L. P. Perera, V. Ferrari, F. P. Santos, M. A. Hinostroza, and C. Guedes Soares, “Experimental results on collisions avoidance of autonomous ship manoeuvres,” in Proc. 32nd Int. Conf. Ocean Offshore Arctic Eng., Nantes, France, Jun. 2013, OMAE2013-11265. Lokukaluge Prasad Perera received the B.Sc. and M.Sc. degrees in mechanical engineering from Oklahoma State University, Stillwater, OK, USA, in 1999 and 2001, respectively, and the Ph.D. degree in naval architecture and marine engineering from the Technical University of Lisbon, Lisbon, Portugal, in 2012. He has won Doctoral and Postdoctoral Fellowships from the Portuguese Foundation for Science and Technology in 2008 and 2012 respectively. Currently, he is a Development Engineer at Wärtsilä Finland Oy, Turku, Finland. His research interests are in maritime systems, instrumentation, guidance and control, condition-based monitoring, energy efficiency, and emission control, safety, risk, and reliability.

Victor Ferrari received the B.Sc. and M.Sc. degrees in naval architecture and marine engineering from the University of Genoa, Genova, Italy, in 2006 and 2009, respectively. He is currently working toward the Ph.D. degree in naval architecture and marine engineering at the Centre for Marine Technology and Engineering (CENTEC), Instituto Superior Técnico, Technical University of Lisbon, Lisbon, Portugal In 2011 and 2012, he was a Research Assistant at CENTEC. He is currently Project Manager for Ships Maneuvering at the Maritime Research Institute

387

Netherlands (MARIN), Wageningen, The Netherlands. His research interests are in ship maneuvering and control.

Fernando P. Santos received the Degree and the M.Sc. degree in mechanical engineering from the Instituto Superior Técnico (IST), University of Lisbon, Lisbon, Portugal and from the Faculty of Science and Technology, Universidade Nova de Lisboa, Lisbon, Portugal, in 2002 and 2012, respectively. He also completed an Advanced Training Diploma in Risk Assessment, Safety and Reliability at IST in 2007. He is a Research Assistant in the Centre for Marine Technology and Engineering (CENTEC), IST and is now conducting doctoral studies on modeling and optimization of offshore wind systems reliability and maintenance.

Miguel A. Hinostroza graduated in mechatronics engineering from the Universidad Nacional de Ingenieria (UNI), Lima, Peru, in 2012. Currently, he is working toward the M.S. degree in naval architecture and marine engineering at the Centre for Marine Technology and Engineering (CENTEC), Instituto Superior Técnico, Technical University of Lisbon, Lisbon, Portugal. At CENTEC, he is working on ship dynamics and controls.

Carlos Guedes Soares received the M.S. and Ocean Engineer degrees from the Massachusetts Institute of Technology, Cambridge, MA, USA, in 1976, the Ph.D. degree from the Norwegian Institute of Technology, Trondheim, Norway, in 1984, and the Doctor of Science degree from the Technical University of Lisbon, Lisbon, Portugal, in 1991. He is a Professor of Naval Architecture and Marine Engineering and President of the Centre for Marine Technology and Engineering (CENTEC), a research center of the Technical University of Lisbon, Lisbon, Portugal, which is recognized and funded by the Portuguese Foundation for Science and Technology.