358
available trucks, or digging and dumping capacities). They instead rely on the human dispatch
operator to plan the operations around these disruptions with the production plan in mind,
and for them to be able to provide their attention at the times where changes are occurring.
During particularly busy times, such as the shift changeover and blasts, the operators can easily lose sight on the state of the mine as they struggle to keep up with the large number of manual truck assignments that are required. Ultimately, by considering only a ‘local’ instantaneous
optimisation, they are unable to find ‘globally’ better solutions across a longer time horizon.
The purpose of this paper is to present a new programming model to explicitly consider the
production targets for a full shift, or similar time horizon, and incorporate planned events into a
more ‘global’ or forward-looking optimisation than existing methods. The intent is to allow for
non-steady-state behaviour, and to reduce the cognitive load on operators and reliance on their
attention at specific points in time. The proposed model also enables feedback control on material quantity targets, allowing re-planning after an unexpected disruption or unusual performance, in order to meet the planned targets rather than relying on the operator to adjust setpoints
manually. In this paper, the time horizon considered for the Flow Planner is a 12-hour shift.
The remainder of this paper is structured as follows. In Section 2, we highlight existing
work related to the haulage allocation plan component. We describe the problem domain in
Section 3, and develop our programming model in the same section. The efficiency of the
proposed methodology is judged in Section 4, by testing it on a set of instances and comparing the results to those of an existing technique. This work has been supported by the Rio
Tinto Centre for Mine Automation and the Australian Centre for Field Robotics, University
of Sydney, Australia.
2 LITERATURE
Haul truck dispatching is well studied in the literature as it is a high-cost aspect of mining. Existing commercial solutions, such as DISPATCH by Modular Mining Systems and
SmartMine, in the mining industry are commonly plan-driven systems (Munirathinam and
Yingling 1994). White and Olson (1986) and White, Olson et al. (1993) developed a technique
that forms the initial basis for DISPATCH (Munirathinam and Yingling 1994), and Subtil,
Silva et al. (2011) developed the technique that is the basis of SmartMine. In both studies, the
allocation plan component is a Linear Program (LP) that calculates the instantaneous flow
rates from diggers such that the productivity of the mine is maximised, subject to constraints
such as the number of trucks available, maximum digging rates, the capacity of the dumps,
and material blending requirements at the dumps. Bonates and Lizotte (1988) proposed a simplified approach to defining the haulage allocation, where an LP is used to optimally obtain
the production rates of all operating shovels such that the production is maximised. The LP
is run once a shift to only maximise the production of shovels in both ore and waste without
considering the associated value and operational costs. Soumis, Ethier et al. (1989) developed
an approach where the entire dispatching procedure is implemented over three stages: equipment plan, operational plan and dispatching plan. The combination of equipment plan and
operational plan forms the basis for the haulage allocation plan. Using a heuristic, the equipment plan yields feasible combinations of shovel locations, a subset of which is passed to the
operational plan where a mixed integer programming model (MIP) performs the optimisation of production. A number of other existing approaches to the allocation plan component
aim to obtain the optimal number of trucks required to service the shovels (Ta, Kresta et al.
(2005), Burt and Caccetta (2007), and Upadhyay and Askari-Nasab (2016)).
3 PROBLEM DESCRIPTION AND SOLUTION METHODOLOGY
Mining operations are typically planned as a set of production targets for a given period. For
long-term plans these can be quite coarse, however at the time-scale of shifts or days, these
production targets typically include: the blocks to be extracted in the period; the dig units
available trucks, or digging and dumping capacities). They instead rely on the human dispatch
operator to plan the operations around these disruptions with the production plan in mind,
and for them to be able to provide their attention at the times where changes are occurring.
During particularly busy times, such as the shift changeover and blasts, the operators can easily lose sight on the state of the mine as they struggle to keep up with the large number of manual truck assignments that are required. Ultimately, by considering only a ‘local’ instantaneous
optimisation, they are unable to find ‘globally’ better solutions across a longer time horizon.
The purpose of this paper is to present a new programming model to explicitly consider the
production targets for a full shift, or similar time horizon, and incorporate planned events into a
more ‘global’ or forward-looking optimisation than existing methods. The intent is to allow for
non-steady-state behaviour, and to reduce the cognitive load on operators and reliance on their
attention at specific points in time. The proposed model also enables feedback control on material quantity targets, allowing re-planning after an unexpected disruption or unusual performance, in order to meet the planned targets rather than relying on the operator to adjust setpoints
manually. In this paper, the time horizon considered for the Flow Planner is a 12-hour shift.
The remainder of this paper is structured as follows. In Section 2, we highlight existing
work related to the haulage allocation plan component. We describe the problem domain in
Section 3, and develop our programming model in the same section. The efficiency of the
proposed methodology is judged in Section 4, by testing it on a set of instances and comparing the results to those of an existing technique. This work has been supported by the Rio
Tinto Centre for Mine Automation and the Australian Centre for Field Robotics, University
of Sydney, Australia.
2 LITERATURE
Haul truck dispatching is well studied in the literature as it is a high-cost aspect of mining. Existing commercial solutions, such as DISPATCH by Modular Mining Systems and
SmartMine, in the mining industry are commonly plan-driven systems (Munirathinam and
Yingling 1994). White and Olson (1986) and White, Olson et al. (1993) developed a technique
that forms the initial basis for DISPATCH (Munirathinam and Yingling 1994), and Subtil,
Silva et al. (2011) developed the technique that is the basis of SmartMine. In both studies, the
allocation plan component is a Linear Program (LP) that calculates the instantaneous flow
rates from diggers such that the productivity of the mine is maximised, subject to constraints
such as the number of trucks available, maximum digging rates, the capacity of the dumps,
and material blending requirements at the dumps. Bonates and Lizotte (1988) proposed a simplified approach to defining the haulage allocation, where an LP is used to optimally obtain
the production rates of all operating shovels such that the production is maximised. The LP
is run once a shift to only maximise the production of shovels in both ore and waste without
considering the associated value and operational costs. Soumis, Ethier et al. (1989) developed
an approach where the entire dispatching procedure is implemented over three stages: equipment plan, operational plan and dispatching plan. The combination of equipment plan and
operational plan forms the basis for the haulage allocation plan. Using a heuristic, the equipment plan yields feasible combinations of shovel locations, a subset of which is passed to the
operational plan where a mixed integer programming model (MIP) performs the optimisation of production. A number of other existing approaches to the allocation plan component
aim to obtain the optimal number of trucks required to service the shovels (Ta, Kresta et al.
(2005), Burt and Caccetta (2007), and Upadhyay and Askari-Nasab (2016)).
3 PROBLEM DESCRIPTION AND SOLUTION METHODOLOGY
Mining operations are typically planned as a set of production targets for a given period. For
long-term plans these can be quite coarse, however at the time-scale of shifts or days, these
production targets typically include: the blocks to be extracted in the period; the dig units