Introduction

BANDIT is an agent-based computational platform, which is designed to evaluate scenarios with an accent on modeling of different types of illegal behavior and on the interaction between agents. The platform consists of an agent behaviour modelling system and a multi-agent maritime simulator. The platform allows to define a number of scenarios through a simple configuration and it offers means to run these scenarios in a single or a batch mode and evaluate the results as single or aggregate data sets respectively.

We utilize the platform to model various contemporary phenomena, primarily focusing on maritime drug smuggling in the Caribbean and in the Eastern Pacific. We also model the migrant activity in the Mediterranean and we estimate areas of high probability of migrant transits.

Check out our new Immigration in Mediterranean simulation demo!

Features

Monte Carlo Simulation

Monte Carlo Simulation

Once you have prepared your agents in the BANDIT platform, you can execute the simulation in a batch mode. The results of the simulation are typically aggregated into easy-to-understand statistics and visualized using a set of heat-maps, traces and waypoints on a globe.

Aerial assets modelling

Aerial assets modelling

The BANDIT is able to simulate not only water surface vessels, but also various aerial assets, such as UAVs or reconnaissance airplanes including their endurance. We also model equipped sensors with their range and probability of object detection.

Environment integration

Environment integration

The BANDIT platform is able to work with various geographical features such as shore lines, navigable water bodies etc. The platform also integrates support for meteorological and oceanographic data from various sources such as NOAA or Copernicus.

Scalable simulation

Scalable simulation

The BANDIT platform is designed to simulate up to thousands of agents in one simulation instance. Each agents operates independently on the rest of the agents, however the platform is capable to model a heavy interaction between the agents if necessary.

Upcoming Features

Game theory

We use game theory to model strategic decision-making of both the adversary as well as the blue team. Game theoretic modeling is a mathematical approach for capturing the reasoning of agents with opposing interests by explicitly considering their set of strategies and possible actions and by modeling their rewards and costs. Finally, by considering other properties of the interaction (such as time span, repetitiveness and observability) we choose the best game model (either extensive or single-turn) and solution concept typically in a form of an equilibrium (such as Nash Equilibrium, Strong Stackelberg Equilibrium or other). The strategies played in the equilibrium correspond to the optimal decision under the constraints and properties considered. We are thus able to estimate the decision making process of the adversary and recommend appropriate action to the blue team.

The Behavior System Model (BSM)

The Behavior System Model (BSM) is a tool to describe the behavior of individual agents in the simulation. We believe that one of the strengths of the multi-agent simulation lies in the possibility of complex behavior modeling of the agents and their rich interaction in the virtual environment. At the current stage of the project the BSMs are defined by a programmer directly in source code of the BANDIT. We are working on a solution where the BSM would be easy to create for a domain expert with little or even no background in computer science by using graphical editor without writing a single line of code.

Use Cases

Immigration Crisis in the Mediterranean

The current situation in the Mediterranean Sea is critical. Tens of thousands of refugees and migrants are trying to reach European shores from the coasts of Africa and Asia. We model the migrant activity using our BANDIT platform. We consider their goals and capabilities, we estimate the parameters of the vessels used for transit and model various realizations with differing fuel level, speed etc., also integrating current oceanographic and meteorological conditions. The results show us most likely routes of the migrants, possible drifting trajectories of damaged boats and best time intervals for transit helping to decide how to allocate naval assets and deploy surveillance drones.

Drug Interdiction in the Central America

The areas of the Caribbean Sea and of the East Pacific are a ground for a large number of illegal drug trafficking activities. We use our BANDIT platform to simulate various behaviors and transit modes of the traffickers. The behavior models also consider the influence of meteorological and oceanographic conditions on the safety of their transit. We simulate a specific scenario thousands of times utilizing the Monte-Carlo approach. From results we are able to estimate the probability of occurrence of the traffickers, their spatial and temporal distribution and also produce an spatio-temporal activity map which could be used for intelligent asset allocation.