Shovel Load Assist Project

This project, also known as SLAP, commenced in 2007 with the vision to develop an autonomous mining shovel that was able to:

· automatically dig material without requiring human involvement;

· load this material into trucks in such a way that the tray capacity is fully utilized and the load is equally distributed;

· manage the dig face and floor to ensure compliance with the mine plan and the maintenance of a favorable bench for excavation;

· have sufficient awareness to avoid self-collisions, e.g. track strikes;

· have sufficient situational awareness to know the activities of equipment operating in its surrounds and use this knowledge in its own decision making, including avoiding collisions with other equipment, notably the haul-trucks being loaded and bulldozers doing clean up work around the machine;

· monitor machine productivity, e.g. cycles times, payloads, and so on;

· employ the on-board sensors used for automation to build local maps of terrain that feed into the mine plan and allow informed automated decision making;

· determine the tactical plan to be used in excavating blocks of material including decisions about when to reposition, where to move to, and what material to take from a position;

· manage overall activity in the load area, including scheduling  of  work,  e.g.   the arrival and dispatch of trucks from loading positions, scheduling floor cleanup, and so on;

· be capable of performing ancillary functions such as management of trailing electrical cable,

· automated parking of the machine to allow access, etc.; and

· have a human-machine interface that enables its effective use.

The participants in the project were the Smart Machines Group at the University of Queensland, the Australian Centre for Field Robotics at the University of Sydney, both through the CRCMining, CSIRO, and Joy Global Surface Mining Equipment (previously P&H Mining Equipment).  

The strategy adopted for working towards the vision of an autonomous mining shovel was by a series of stepping stone technologies, each of which would contribute to the overall goal whilst having a value proposition to the open-cut coal sector as a useful technology in its own right.  The name of the project deliberately included the word 'assist' to capture the notion that while the long term goal was full automation of mining shovels, the pathway to achieving this was to be by a series of semi-automation technologies that would assist the operator achieve higher safety and productivity.  

The project identified and developed five key stepping stone technologies

· TrackShield: A safety related operator assist technology that helps operators avoid self collisions, focussing on collisions between the shovel dipper and tracks.

· TruckShield: A safety-related operator assist technology that provides an engineered layer of protection against high-energy collisions between shovels and trucks.

· DozerShield: A safety-related operator assist technology provides an engineered layer of protection against high-energy collisions between shovels and the clean-up dozer.

· AutoSwing: An operator assist technology providing a semi-automation layer that allows the swing, dump and return phases of the shovel cycle to be performed autonomously.

· AutoFill: An assistive technology providing a near-full-automation layer that allows the dig, swing, dump and return phases of the shovel cycle to be performed autonomously.

These five technologies represent increasing levels of automation, with AutoFill providing near complete automation of typical truck loading activities. The project was conducted in four phases: The first three phases focussed on the development of Trackshield, TruckShield, DozerShield and AutoSwing. The first phase developed proof-of-concept systems on a P&H2100BLE shovel that was installed at Bracalba Quarry, near Caboolture, for the purpose of the project. Phase 2 addressed and closed  knowledge and technology gaps identified in the work of Phase 1. Phase 3 trialled the technologies in a production environment at Lake Lindsay mine in Central Queensland.  AutoFill was developed and tested as Phase 4 of the SLAP project at Bracalba Quarry on the P&H2100BLE.  Throughout, the project conducted its activities within the functional safety lifecycle frameworks of AS61508 and AS62061.

Progress on the first technology, TrackShield, had been made in a precursor project - C15038 Collision Control and Avoidance for Electric Mining Shovels - which addressed the problem of avoiding self collisions. This technology continued to mature during the SLAP Project to become a product now commercially available from Joy Surface Mining for their mining shovels with sales of the product globally.

Truckshield is an extension of the TrackShield concept with the purpose of managing the inherent risks associated with loading of haul trucks by electric mining shovels.

A significant research question that the SLAP project addressed was accurate and reliable determination of the position of trucks (and other equipment) relative to the shovel.  A  redundant truck pose estimation system was developed that used three technologies that had a minimal set of common-cause failure modes to determine the position of trucks relative to the shovel: (i) inertially aided high-precision GNSS trilateration; (ii) ultra-wideband ranging transceivers mounted to trucks and the shovel; and (iii) shovel mounted LiDAR sensors that determined truck position from point clouds generated by the sensor. The system requires agreement of two of the three systems and is capable of estimating the six dimensional position and orientation of trucks being loaded with sufficient accuracy that no point on the body of the truck is more 0.3m from where the system estimates it to be, with availabilities of 99.99% when the truck was stationary and 0.5m with availabilities of 98.97% when the truck was in motion.

Field trials of TruckShield conducted on a P&H4100XPB and Komatsu 830e truck at Lake Lindsay mine in Central Queensland demonstrated that TrackShield can address the risks of truck strikes with no observable detrimental impact on cycle times when compared with manual operation.

Dozershield was a further extension of the TruckShield/TrackShield concept to clean-up equipment operating around the shovel.  Currently, operations must rely on either procedural or alarm based solutions to manage the risk of dozers performing clean-up around shovels. Again, a significant challenge is determining the six-dimension spatial pose of the dozer relative to the shovel and also predicting its likely future motions.  

The project has demonstrated, through field trials of DozerShield at Lake Lindsay mine using a P&H4100XPB Shovel and a Caterpillar D11 Dozer doing clean-up work, that a hard engineering control can be applied to reduce the risk of collision between shovel and dozer. Because DozerShield inhibits shovel motion when a sufficiently high probability of collision is detected, it introduces operational delays. Extrapolating the activation delays observed during trials suggests DozerShield would introduce delays of approximately 400s per shift.  However, these larger than expected delays were due to the large number of activations observed and DozerShield delays are expected to be considerably less when paired with higher compliance to SOP and advanced warning of the dozer operator.   

TruckShield, DozerShield and AutoSwing are designed to reduce the frequency and severity of collisions.  These technologies alter the health and safety risk profile of shovel operation but also have the potential to influence value by reducing equipment downtime, reducing repair costs, etc.  The expected value of these savings is directly linked to the frequency of the event occurring and the potential that the technology has to reduce that frequency.  The integrity and functionality that the technology provides thus influence both value and risk in a closely coupled manner.  

For AutoSwing, however, there are also capabilities that the technology provides that influence value without a corresponding impact on risk, e.g. faster swing times. Field trials of AutoSwing at Lake Lindsay mine showed that The autonomous loading of trucks is feasible and results obtained predict a 11.5% reduction in swing time equivalent to 1.4 second per loading cycle. Moreover, smoother machine response observed in swing armature current with predicted benefits to machine duty.

AutoFill extends the functionality of AutoSwing by adding an autonomous digging capability.  Under control of AutoFill, the shovel can complete fully autonomous truck loading passes without operator intervention. The scope of AutoFill includes: dig selection, digging, swinging, dump point selection, dumping, and returning to tuck. During trials of Autofill it was demonstrated to be robust in detecting the presence of the truck on its arrival, identifying a dig strategy (i.e. where to dig from), executing this strategy, monitoring the material excavated after each dig, planning where to dump to equalize tyre loading and updating the excavation sequence for a truck being loaded in real-time to accommodate for variations.

The value proposition for TruckShield, DozerShield and AutoSwing was investigated   as  part of the project scope. DozerShield and AutoSwing present strong value propositions while TruckShield is an attractive solution along the path to automation. Using parameters typical of a Central Queensland Coal mine, the per-shovel, net present value of these technologies was estimated to be $10.7m with a payback period of 3 years in unconstrained production setting and $1.5m with a payback period of 7 years in a constrained production setting.

Through the SLAP project, what, in 2007, seemed an ambitious vision for a future autonomous mining excavator now seems within the reach of existing knowledge and technology.  Achieving it, will of course take a continued effort by stakeholders, with the likely pathway being by through realizing of incremental stepping stones over time, e.g. DozerShield, TruckShield, AutoSwing, AutoFill, then the autonomous mining shovel.