An Automated Approach to Generating Haul Roads

For a mining operation, the determination of the best place to build haul roads is typically done by hand, or with the assistance of rudimentary shortest path optimisation algorithms. Straight line, point to point, modelling of costs is often used with distances multiplied by constants that are based on rise of the road between the points. Failings of this approach are that straight line approximations are notoriously poor as they do not take into account areas where a truck cannot travel. In addition, the rise and fall of the road may vary considerably over the truck journey meaning acceleration/deceleration and varying truck speeds are not taken into account. Lastly, the use of rudimentary optimisation algorithms means properties a truck must obey such as minimum turning circles, line of sight restrictions or even that the truck must travel forward at all times are not taken into account. Thus solutions generated may not be feasible, let alone optimal. In this paper we discuss a state of the art parallel Artificial Intelligence tree search technique for solving haul road location problems. As well as the above properties, factors such as road utilisation, cut and fill, and construction costs are taken into account. The combination of advanced modelling and fast computing and optimisation techniques produces significant benefits both in terms of time to generate roads and cost savings. Preliminary results where this method has been used are discussed.