Making the Pilbara Blend: agile mine scheduling through contingent planning
 Photo courtesy of Rio Tinto Iron Ore Pty. Ltd.
The Project
Australian Research Council (ARC) Competitive Research Grant (Linkage Project),
Making the Pilbara Blend: agile mine scheduling through contingent planning
$1.36 million AUD
(ARC: $600,000; Rio Tinto Iron Ore: $762,732) 2012-2015
The Chief Investigators
Associate Profesor Adrian Pearce
and
Professor Peter Stuckey
The Research Fellows
Research Fellows Michelle Blom, Christina Burt and Nir Lipovetzky are presently working on this project.
Research Participation
Relevant research groups include the Intelligent Agent Laboratory (www.agentlab.unimelb.edu.au) and the constraint programming lab (www.nicta.com.au/research/projects/constraint_programming_platform) of the NICTA Optimiation group
(www.nicta.com.au/research/optimisation).
The project
Making the Pilbara Blend: agile mine scheduling through contingent planning is
a collaboration between Rio Tinto Iron Ore and The University of Melbourne, lead by
Chief Investigators Dr. Adrian Pearce and Professor Peter Stuckey.
Detailed project overview
Mine scheduling is a challenging problem for our industry partner Rio Tinto Iron Ore (RTIO) which annually mines more than 200 Million Tonnes of iron ore.
This exciting new ARC Linkage project will tackle a problem they face in making their signature product, the Pilbara Blend, for which Australia is highly regarded internationally.
Making the Pilbara Blend involves taking two distinctly different ore types from multiple mines and dynamically blending them to make a greater value product with highly-consistent quality.
This project tackles a critical problem faced in mine scheduling. An increased need for consistent quality has occurred at the same time as the complexity of modern day mining operations has increased, across multiple mine sites with variable ore grades and increasing infrastructure constraints.
There is a pressing need for more agile mining techniques that maximise net present value (NPV) while accommodating the complexities and uncertainties inherent in modern day mining operations.
Multi-Agent Contingent Planning
In 2010, Rio Tinto approached the University of Melbourne for assistance with their production scheduling, based on multi-agent planning and optimisation techniques developed within the Intelligent Agent Laboratory and the NICTA Optimisation group.
Paradigm Change
This project aims to bring together automated planning techniques with constraint programming to address this paradigm shift.
The project will develop agile scheduling techniques of great economic importance. Carefully planned scheduling has the potential to reduce the need for new infrastructure, minimising environmental impacts and maximising regeneration after mining.
The goal of this project is to bring together automated planning techniques with constraint programming to address this paradigm shift.
The research tackles plan synthesis for the multi-commodity, multi-mine site supply chain problem and promises to lead to improved techniques for collaborative planning and solving hard constraint problems.
The theoretical significance of this project derives from the fundamental challenge posed by the problem underlying dynamic scheduling.
On the one hand, optimisation must take into account nondeterminism - the uncertain outcomes of actions inherent in the problem - and on the other must lead to feasible plans for each mine site - while continuing to solve the necessary constraints for the overall supply chain schedule during the course of continuing (non-terminating) execution.
Research Challenges
Our research agenda tackles some of the most fundamental challenges in automated planning and constraint programming, including
- contingent planning
- partial observability
- non-deterministic planning
- non-terminating execution
- online planning
- continuous constraint satisfaction
One of the aims of the project is to realize a synergy between techniques developed in the constraint and planning fields.
More Information
Please Email Adrian Pearce if you are are interested
Contact: Adrian Pearce
|