Dragline Excavation Sequencing

It has long been recognised that optimising the positioning and material movement tasks of dragline operation has the potential to reduce excavation times and improve spoil management. The challenges to building a system capable of realising these benefits are many and varied, and include the complexity that emerges from the infinite ways in which the machine can be applied to an excavation task, encoding the situational awareness required to characterise the available set of tasks relative to the state of the excavation, addressing the competing objectives of compliance, productivity, and spoil fitting, competing with operator judgement on what the operation requires and addressing the inevitable deviation of execution from the intended task.

This project has wrestled with these challenges to develop and trial a prototype operator assist technology for dragline excavation sequencing. The technology provides guidance to operators on where to position the machine, where to dig material, and where to place the dug material. Implicit in this guidance are decisions on how the larger excavation tasks are segmented and how different strategies act to improve the efficiency of the terrain transformation provided by draglines.

The project framed the problem as computing a sequence of excavation tasks along a receding horizon to optimally progress an excavation. Computational models were used to interpret machine and terrain perception data relative to the strip design and machine configuration, and to simulate the impact of excavation decisions on the future state of the operation. A decision-tree framework was used to construct potential sequences that were constrained by operating conditions and the available excavation strategies, and to search and rank those decision sequences that indicate a benefit to operation. Starting from an initial operating state (terrain, machine position, and strip design), the root of the tree is branched according to the available decision set, with each decision combining a repositioning action and an excavation action. Each decision-point in the tree provides an operation state from which further branching of decisions can occur. The combinatorial complexity of the decision space was mitigated through a range of mechanisms, including: applying heuristics (or accepted behaviours) to reduce the available decision set at each branching point; using a receding horizon approach that limits the depth of the tree and recomputes at changes in operation; and by abstracting the material movement simulation at different levels of detail to reduce the computational cost of evaluating decision sequences.

The simulation capability of the system was used to identify the extent of the potential benefit from optimal excavation sequencing. Comparison of operation data with simulated excavation fed by optimised excavation tasks identified two areas of opportunity for increasing the productive output of draglines.

The first opportunity relates to achieving consistent operation, at or near the maximum productive capability of the machine. Though operators must manage both machine motions and the accumulation of duty, the large variation in cycle times for similar operation cycles translates to an opportunity to reduce variation and reduce average cycle times. Combined with faster swinging, a margin of up to 40% over typical operation is identified, though this value carries the uncertainty of model simplifications and mismatch between model parameters and reality.

The second opportunity for a production gain comes through the combination of faster operation cycles and 'optimal task sequencing'. A 10% improvement to productivity appears possible by optimising the positioning, digging, and spoiling decisions made by operators. This opportunity was the primary target of the operator assist developed in this project. To realise that benefit requires two key technical capabilities: (i) the situational awareness to characterise the operating context, identify the excavation strategies most suited, and evaluate how the current state impacts the short and long term objectives of the operation; and (ii) a framework to efficiently search the population of potential excavation sequences.

The project developed a prototype system with the capability to: (i) compute excavation sequences consistent with operation constraints and established excavation strategies; (ii) operate on-board a dragline as part of the production workflow; (iii) display guidance to assist operators in executing the computed sequence; and (iv) adapt to operator execution and deviation from the computed sequences.

The production trial brought focus to two key challenges facing the operator assist: (i) building operator trust in the sequencing capabilities to mitigate the reservations borne from operator experience and judgement; and (ii) the level of situational awareness required to feed the decision-making process. In practice, both of these challenges hinge on the capability to accurately perceive and comprehend the operating options and to make accurate predictions over short and long term horizons. Operators assess the evolving trade-off between productivity and spoil management based on experience and break down their excavation tasks accordingly. In addition to positioning decisions, operators also discretise their work into consumable blocks. The length and shape of these blocks impacts their capability to utilise the available spoil room. The sequencing system must make the same judgements, but with the benefit of enhanced situational awareness, it has the capacity to drive excavation to the limit of that warranted by the state of the operation and the demands of the strip.

The pursuit of delivering an operation benefit during the trial was negated by operator uncertainty on the impact of 'optimised' decisions and insufficient situational awareness to support both short and long term forecasts while covering the full decision space used by operators. However, the trial represented the first time a task-sequencing operator assist has provided guidance to dragline operations and represents a significant and important step toward realising value from the technology. The path forward to deliver on the potential, and to extend the application to broader dragline scenarios, is well defined.