Nevada Gold Mines reviews its mining truck automation journey at Arturo

In mid-2017 Barrick Gold in partnership with Premier Gold Mines contracted Autonomous Solutions Incorporated (ASI) to complete an autonomous operation Proof-of Concept (POC) at Nevada Gold Mines (NGM), Goldstrike – Arturo Operation. The POC directed ASI to supply a successful retrofit of five Komatsu 930E-2 haul trucks for fully autonomous operation, and to provide and implement their proprietary Mobius control software. By June of 2018, the first truck began testing. This project and the lessons learned along the way are being presented this week by J . Oxborrow from the Nevada Gold Mines Open Pit Engineering team at this year’s SME MineXchange Annual Conference in Phoenix, Arizona, being attended by IM.

Oxborrow’s paper says the safe implementation of an Autonomous Haulage System (AHS) on 15+ year old haul trucks required a vast number of changes to procedures, in both operations and planning. Considerations to employee perception and acceptance of the project were among the most important items during the implementation process. Operators were engaged from the onset of the project to ensure project acceptance, and that all concerns were addressed.

“ASI was willing and able to be agnostic to not only the OEM equipment manufacturer but also to the dispatch system used at the mine, where each of the other vendors would require a conversion to a specific dispatch system at an additional cost. Also crucial was the ASI ability to retro-fit non-drive-by-wire trucks, and an accelerated timeline to match the availability of equipment to perform the POC.”

“An important design consideration of the autonomous system that was discussed thoroughly in the risk assessments was that when the truck was in manual mode, the autonomous system components had no effect on the original equipment manufacturer (OEM) design. Similarly, the system was designed so that when in autonomous mode any manual input to the OEM brake or steering system would cause the autonomous system to release control of the vehicle.
Once the autonomous kit was installed on the first haul truck testing and tuning of the components began. From the initial risk assessments, it was determined that a test track isolated from other operations would be required. The test track was constructed to allow the simulation of a mining environment including shovel load areas with cable trees, dump edges, stack dumps, roads, ramps, and a large open area to allow for various control tuning operations.”

An NGM supervisor was used as the testing coordinator, this supervisor was responsible for controlling access to the test track. While testing was taking place, a roadblock was used at the entrance to the track. The personnel maintaining the roadblock only allowed vehicles past that had made positive communication with the supervisor at the test track after the supervisor had ensured the autonomous vehicle was locked out.

At the start of each day of testing at the test track, ASI, Sedna, and NGM employees met and completed a daily test plan and team risk assessment. The team risk assessment was the controlling document for the tests to take place that day. The testing supervisor was responsible for ensuring that all personnel that entered the test track were aware of the tests and potential hazards that would be taking place. At the end of each day ASI would report the test results to the project team.

“The main risk mitigation during testing was the use of a ‘safety rider.’ The safety rider sat in the operator’s seat of the truck while an ASI engineer sat in the passenger seat. If at any time the truck began to do any unsafe action, the safety rider would override the autonomous system and bring the truck to a safe stop. Specialised training took place for Safety Riders so that they could see how the equipment would respond. To evaluate the risk associated with the design of the ASI system hardware and software a thorough review of the system was performed using the system theoretical process analysis (STPA) process.” The STPA process was extensive and resulted in several changes to the system design and new trainings to be developed.

“The switch to automated haulage required many new standard operating procedures (SOP) to be developed to ensure worker safety and to realise efficiency gains. There were also many existing procedures that required modification for use in an autonomous setting. Procedures needing to be written and those needing revision were identified during the risk assessment process. A small group of NGM project team members, ASI, and Sedna representatives developed a detailed flow diagram for each procedure. At this point the employees selected as control room operators, maintenance technicians, and shift supervisors were given the flow diagrams and given the task of creating written procedures for each of the flow diagrams. NGM, ASI, and Sedna employees involved in creating the flow diagrams were available to the operators throughout the process of creating the written procedures to answer any questions or concerns that the operators may have had. In all 21 new procedures were developed and 5 existing procedures were highly modified.”

All procedures were written with the intent that as the project progressed, knowledge increased, and the system capabilities evolved that they would be modified. Procedures were periodically reviewed and modified. Throughout the project. The biggest example of this was the introduction of vehicle to everything (V2X) technology which significantly changed the processes for manned and autonomous vehicle interaction as well as the process for switching from autonomous mode to manual mode and vice versa. Operators are encouraged to bring suggestions forward to help optimize and improve processes.

“From the on-set of the project NGM sought to utilise its own employee to operate and maintain the system. Maintenance personnel were trained on how to maintain new control systems installed on the trucks and equipment operators were trained on how to monitor and operate the system software.
Among the employees brought on-board the project was a dedicated trainer. After the creation of the SOPs the trainer developed a site access training class that was required for all employees and contractors to gain access to the automation area. In the first 5 months over 400 employees and contractors received autonomous hazard training.”

A key component of operator training was the use of simulators to familiarise the operators with the use of the Mobius control software. A preliminary map of the mining area was created and uploaded into Mobius, operators were then rotated between monitoring trucks, managing maps, and running the simulated shovel interface. The simulator training was crucial in the operators learning how path planning occurred, how close to map edges and obstacles the trucks would get, and best practices for the placement of cable trees. The simulator training also allowed planning and user interface issues to identified early in the project.

As production commenced and the control operators became more proficient in using the system, a training program was started in which non-project employees were put through simulator and hands on training. This training was termed autonomous ‘boot camp.’ By increasing non-project employee exposure to the autonomous system more interest in automation was generated. Since the completion of the boot camp program those who participated in the training have been used as fill-in control room operators, and many have asked about full-time assignment to the project.

“The most critical control room training was in the mapping process. Accurate maps of autonomously drivable areas and autonomous exclusion zones are critical, not only for preventing equipment damage, but to ensure employee safety in the autonomous area. From the start of the project, maintenance personnel worked closely with ASI and Sedna as the kits were installed on the trucks, receiving hands on training as the kits were installed. When operations began, the maintenance technicians were instrumental in helping identify weak points in the design of the system and recommending changes to increase the durability.”

When the first few hardware bugs were straightened out, a maintenance technician was moved to a maintenance trainer role. The maintenance trainer developed a training course for all maintenance employees. The course consisted of a presentation which gave a high-level overview of the automation hardware installed on the truck with an explanation of how it interacted with the OEM system. After the presentation, the trainees were taken to an autonomously equipped vehicle and shown how the hardware works, including how interface controllers read inputs from sensors and sent outputs to actuators. After the training was complete mine maintenance personnel were all much more comfortable and competent when working around the sensitive electronics on the modified trucks.

The paper says mine planning is a critical factor in the overall efficiency of an autonomous system. “The autonomous system is not able to adapt to changes in its surroundings and will not use judgment to make decisions. As such detailed planning is required for the autonomous system to be efficient. Changes to the current short, and mid-range mine planning processes were required to fully realise the benefits of automation. Short range planning originally created a shovel sequence that was a generalised tonnage estimate for the upcoming week out to three months. As autonomous haulage was implemented the shovel sequence needed to become much more detailed involving planning for shorter shovel moves and how the shovel would advance across the dig face. Shapes for the load area are sent to a map store on a server accessible by the control room operators who pull in the new load area shapes as they are needed.
After the load area expands past a certain point, trucks begin to plan in less desirable ways as the free space of the load area allows the Mobius planning algorithm to perform shortest path planning. To better control the path planning of the truck the road network must be extended onto the bench as the shovel progresses and the load area shape must be cut back.”

“Prior to short range planning cable routing, several moves would result in production inefficiencies and in some cases encroach on the truck’s minimum required operational space. The autonomous system requires a consistent, low-latency wireless communication network to operate. The planning for the wireless network has become a critical task in the operation of an autonomous haulage fleet. Changes to the mine terrain effect the signal strength across the entire operational area. A radio frequency engineer works with mid and long-range planning to predict coverage gaps as the pit and dumps advance. Radio access point moves are planned and coordinated with operations to minimise the potential effects of the move.”

“New coverage models are created each week from surfaces created by survey from drone flight data. These coverage models are then validated with data from the operating vehicles. As more data from the vehicles is corrected the model accuracy has been improved. As anomalies in coverage are encountered, data can be collected to identify the source of the problem. A spectrum analyser was purchased and training of supervisors and maintenance personnel on its use allows vast amounts of diagnostic data to be collected from throughout the operational area. For more detailed analysis of the wireless system, a signal analyser can record every aspect of the radio signal, from the phase, frequency, and amplitude, to each individual data symbol and how it is modulated.”