Vessel Behavior Models

We employ agent-based modeling techniques to represent the behavior of each vessel class. Each vessel is implemented as an autonomous agent with its distinct behavior, capable of interacting with the simulated maritime environment and other vessels. Three categories of vessels form the core of the model: (1) merchant vessels, (2) pirate vessels and (3) navy vessels. Each behavior model is based on real-world data. Below, we describe models for each vessel category.

Top

Merchant Vessel Model

Merchant vessels, traveling repeatedly between the world’s large ports, form the bulk of international maritime traffic. Our agent-based model aims to achieve good spatio-temporal statistical correspondence with the real-world traffic distribution -- we do not aim to accurately model physical dynamics of individual vessels. The lifecycle of the merchant vessel is captured below.

Merchant vessel model

Top

Pirate Vessel Model

In the real world, pirates employ different strategies and equipment in their attacks. To reflect this situation as accurately as possible, we have created executable models for the following four pirate strategies:

  1. simple pirate with a small boat without any means of vessel detection except the direct line of sight (5 nm)
  2. radar pirate with a radar extending the vessel detection range to 20-50 nm
  3. AIS pirate with an AIS interception device monitoring AIS broadcasts
  4. Mothership pirate with a medium-size vessel and several small boats able to attack vessel up 1500nm from the Somalian coast.

The last type can be combined with the radar and an AIS interception device to achieve even more complex composite behavior. Finally, all models can be extend with learning capability to model how agents adapt their behavior over time in response to changing circumstances and security measures.

Pirate vessel model

Top

Navy Vessel Model

The lack of data and general complexity of patrolling strategies –- which can vary from active search for pirates to protecting transiting groups of cargo-vessels -– make the task of proper modeling of naval agents very difficult. We thus propose a minimal static model based on the information available and then propose a game-theoretic techniques for optimum patrolling for particular transit scenarios (see Patrolling game).

Naval forces

Top

Merchant Vessel Route Planning

Route planning is an integral part of vessel operation -- we have developed several route planning algorithms that can be used to plan vessel route according to a set of criteria. The basic shortest-route planner is used by all classes of vessels. In its basic configuration, the planner finds a shortest path (accounting for the curvature of the Earth’s surface) and can incorporate additional criteria, such as the risk of a pirate attack. Additional, piracy-risk aware planner have been developed for planning the routes of merchant vessels, based on results on risk assessment and optimal game theoretic routing. The simulation platform provides an extensible hierarchical route planning architecture by which different types of planners can be invoked depending on the region navigated.

Top

Agent Behavior Implementation

We employ an variant of finite state machines (FSM) to implement the behavioral models described above. FSMs provide a good mix of expressiveness, modularity and computational effectiveness. The states of our FSMs correspond to the principal mental states of the vessel agent (such as move, attack, hijacked, patrol etc.). Transitions between the states are defined by unconditional state transitions or by conditional transitions conditioned by external events. A state stores current information when it is inactive and when reactivated, the information can be retrieved to continue in previously interrupted plan, the transitions between the states can be internally or externally triggered, thus allowing interruption of plans or actions.

FSM of a pirate with a mothership and a radar.

FSMs are easily extensible and the implementation of states can be easily reused. It is possible to create abstract skelets of various FSM and then enrich them with a specific behavior. The ability to learn, i.e. store the information can be implemented using internal variables of the respective state.

The usual difficulties of incorporating time into the FSM are overcome by granting the agent each turn a time quantum which is used for an activity performed in a particular state. A disadvantages of FSMs is inability to naturally model concurrent actions on the side of agents, though this has not proven a problem for the maritime traffic simulation.

Publications

Papers

  • Ondrej Vanek and Ondrej Hrstka and Michal Pechoucek: Improving Group Transit Schemes to Minimize Negative Effects of Maritime Piracy. IEEE Intelligent Transportation Systems (to appear). 2014.
    BiBTeX | PDF (430)
  • Ondrej Vanek and Michal Pechoucek: Dynamic Group Transit Scheme for Corridor Transit. In Proceedings of the 5th International Conference on Modeling, Simulation and Applied Optimization (ICMSAO). IEEE Press, 2013.
    BiBTeX | PDF (388)
  • Ondrej Vanek and Michal Jakob and Ondrej Hrstka and Michal Pechoucek: Agent-based Model of Mariime Traffic in Piracy-affected Waters. Transportation Research Part C: Emerging Technologies.. 2013, vol. 36, p. 157–176. ISSN 0968-090X.
    BiBTeX | PDF (422)
  • Michal Jakob and Ondrej Vanek and Ondrej Hrstka and Michal Pechoucek: Agents vs. Pirates: Multi-Agent Simulation and Optimization to Fight Maritime Piracy. In 12th International Conference on Autonomous Agents and Multiagent Systems. 2012.
    BiBTeX | PDF (345)
  • Ondrej Vanek: Security Games with Mobile Patrollers (Extended Abstract). In Proceedings of 10th International Conference on Autonomous Agents and Multiagent Systems (AAMAS). 2011.
    BiBTeX | PDF (7)
  • Ondrej Vanek, Michal Pechoucek, Michal Jakob, Branislav Bosansky a Viliam Lisy: Agentni simulaci proti somalskym piratum. Scientific American, Czech Edition. 2011.
    BiBTeX | PDF (5)
  • Ondrej Vanek, Michal Jakob, Ondrej Hrstka, and Michal Pechoucek: AgentC: Agent-based System for Securing Maritime Transit (Demonstration). In Proceedings of The 10th International Conference on Autonomous Agents and Multiagent Systems. 2011.
    BiBTeX | PDF (10)
  • Ondrej Vanek and Michal Jakob and Viliam Lisy and Branislav Bosansky and Michal Pechoucek: Iterative Game-theoretic Route Selection for Hostile Area Transit and Patrolling. In Tenth International Conference on Autonomous Agents and Multiagent Systems. 2011.
    BiBTeX | PDF (14)
  • Michal Jakob and Ondrej Vanek and Michal Pechoucek: Using Agents to Improve International Maritime Transport Security. IEEE Intelligent Systems. 2011, vol. 26, p. 90-96. ISSN 1541-1672.
    BiBTeX | PDF (13)
  • Ondrej Hrstka and Ondrej Vanek: Optimizing Group Transit in the Gulf of Aden. In Proceedings of 15th International Student Conference on Electrical Engineering (POSTER). 2011.
    BiBTeX | PDF (6)
  • Branislav Bosansky, Viliam Lisy, Michal Jakob and Michal Pechoucek: Computing Time-Dependent Policies for Patrolling Games with Mobile Targets. Tenth International Conference on Autonomous Agents and Multiagent Systems. 2011.
    BiBTeX | PDF (12)
  • Ondrej Vanek and Branislav Bosansky and Michal Jakob and Michal Pechoucek: Transiting Areas Patrolled by a Mobile Adversary. In Proceedings of 2010 IEEE Conference on Compuattional Intelligence and Games. 2010.
    BiBTeX | PDF (51)
  • Michal Jakob and Ondrej Vanek and Stepan Urban and Petr Benda and Michal Pechoucek: Employing Agents to Improve the Security of International Maritime Transport. In Proceedings of AAMAS 2010 Workshop on Agents In Traffic and Transportation. 2010.
    BiBTeX | PDF (59)
  • Stepan Urban and Michal Jakob and Michal Pechoucek: Probabilistic modeling of mobile agents' trajectories. In Proceedings of the International Workshop on Agents and Data Mining Interaction (ADMI 2010). 2010.
    BiBTeX | PDF (50)
  • Michal Jakob and Ondrej Vanek and Stepan Urban and Petr Benda and Michal Pechoucek: AgentC: Agent-based Testbed for Adversarial Modeling and Reasoning in the Maritime Domain (Demo). In Proceedings of The Ninth International Conference on Autonomous Agents and Multiagent Systems. 2010.
    BiBTeX | PDF (58)
  • Ondrej Vanek: Agent-based Simulation of the Maritime Domain. In POSTER 2010, 14th International Student Conference on Electrical Engineering. CVUT, Fakulta elektrotechnicka, 2010.
    BiBTeX | PDF (53)
  • Ondrej Vanek: Agent-based Simulation of the Maritime Domain. Acta Polytechnica. 2010, vol. 50, p. 94-99.
    BiBTeX | PDF (44)
Top

Books

  • Ondrej Vanek and Michal Jakob and Michal Pechoucek: Using Data-Driven Simulation for Analysis of Maritime Piracy. In Prediction and Recognition of Piracy Efforts Using Collaborative Human-Centric Information Systems. IOS Press, 2013, p. 109-116.
    BiBTeX | DOCX (316)
Top

Other

  • Ondrej Vanek: Computational Methods for Transportation Security. . 2013.
    BiBTeX | PDF (431)
  • Ondrej Vanek, Michal Jakob, Ondrej Hrstka, and Michal Pechoucek: Agent-based System for Securing Maritime Transit (May 2011). . 2011.
    BiBTeX | PDF (313)
Top

Reports

  • Michal Jakob and Ondrej Vanek and Branislav Bosansky and Ondrej Hrstka and Vojtech Krizek and Stepan Urban and Petr Benda and Michal Pechoucek: Adversarial Modeling and Reasoning in the Maritime Domain - Year 2 Report. . 2010.
    BiBTeX | PDF (26)
  • Michal Jakob and Ondrej Vanek and Stepan Urban and Petr Benda and Michal Pechoucek: Adversarial Modeling and Reasoning in the Maritime Domain (Year 1 Report). . 2009.
    BiBTeX | PDF (57)
Top