Swarm Agent-Inspired Task Assignment in Mines

Motivation and Scope

With increasing pressure on efficient organization of (underground) mines, new approaches to coordinate transport need to identified and analyzed. Multiagent Systems and Distributed Artificial Intelligence offer ideas that start from a local point of view. Hereby every agent (vehicle, storage pit…) decides about its next action based on its own state and on interaction. All agents together generate a solution without centralized contol. Such approaches based on self-organization promise to offer graceful degradation in presence of failure, which is particularly important in dynamic environments. Yet, there may be a price in performance that needs to be identified.

Our idea of a flexible transportation system is based on autonomous storages pits who evaluate their material filling degree and call for a transport towards other less filled storages only in direction of the exit of the mine. Which vehicle will do the transport to which other storage pit, is up to negotiation. As with any self-organized approaches, calibration of thresholds (e.g. what is the critical ratio to send away material) is of particular importance to create a good solution.

Specific Tasks

· Implementing a specific example scenario of an (underground) mine using the coordination_oru framework. This Java-based framework can handle the collision-free, coordinated movement of vehicles.

· Designing and implementing a task assignment, which is only driven by local resource levels and needs using multi-agent system concepts.

· Integration of the agent-based local task assignment into the example implementation

· Testing and evaluating the overall simulation to evaluate its performance

· If time if left, extend the simulation to use the standard task assignment of coordination_oru and compare its performance to the new self-organizing one.

Necessary Skills

· Good programming skills, preferable in Java

· high algorithmic understanding and analytical skill

· Artificial Intelligence and Distributed Systems courses.

Context and Benefits

This project is initiated by a leading scientist at ABB Sweden interested in Multiagent Systems. Also other Swedish Mining companies showed interest. The thesis project is located at Örebro University due to our expertise in Distributed Artificial Intelligence. The thesis is therefore not a simple AASS-thesis, but has connection to industry. You will not just get important connections at industry, but also learn about Distributed Artificial Intelligence. You will also gather experience in developing complex software.

Contact

Franziska Klügl Franziska.klugl@oru.se