Decentralised Task Distribution in Mobile Agent Systems

Occurrences : 

2021

Nombre d'étudiants minimum: 

2

Nombre d'étudiants maximum: 

4

Nombre d'instances : 

3

Faisable à distance: 

Oui

The purpose of this project is to distribute a number of tasks to a set of agents, without using a central controller. A set of agents move randomly (or using Levy walks) through the 2D space. Initially, each agent is assigned a task type randomly. Each agent, upon meeting another agent, notes the task that the other agent is currently doing. After meeting a certain number of agents, the agent estimates the task distribution currently present in the system; and adjusts its own task accordingly, to better reach the task distribution goal (given as input). All agents have an estimate of the total number of agents in the system.

The project consists in modelling the above system and finding the task adjustment algorithm that leads to meeting the task distribution goal (as fast as possible).

Options: Agents may or may not have unique id-s and recognise each other accordingly (e.g. in case two agents meet several times within an interval).

Configurable parameters: the total number of agents; the number of task types; the task distribution; optional: the type of walk (random or Levy); the number of agents to meet (or ticks to wait for) before re-evaluating the current task distribution and updating an agent's own task; <others>.

Implementation platform: NetLogo or Java.