Why autonomous mining equipment stalls at implementation
Implementing autonomy at an open-pit mine rarely hinges on the machine itself. A pilot area shows results: the excavation cycle runs, no operator in the cab is needed.
But when scaling to the entire mine, the project stalls — and the cause is almost always the environment around the equipment, not its hardware.
The first bottleneck is fragmented telemetry. Data from onboard systems, dispatch, navigation, and fuel live in different systems, with different polling rates and their own timestamps.
An event in one area cannot be correlated with the machine's response in another, so the cycle picture is assembled manually and with a delay.
Second is the absence of a unified control loop. Autonomous, remote, and manual modes operate separately, and switching between them requires an operator command.
At the junctions between modes, productivity is lost, and automation bottlenecks are not visible as a whole: charts show a dip, but not its cause.
Third is the human factor in excavation. Shift personnel set different cycle speeds and react differently to abnormal situations. As long as the behavior logic of autonomous mining equipment is not reproducible, cycle quality depends on who is on shift, not on the algorithm.
Next, we break down how this is closed: where to place unified data collection, how to link modes into one loop, and what exactly removes dependence on the shift.
What a loop without an operator in the cab delivers: measurable effects
An autonomous loop is evaluated not by a demonstration run, but by the shift in metrics per shift. Below are five effects that we record before and after converting an area to operation without an operator in the cab — with a baseline and a target value.
- Loading cycle time. Without shift changes and pauses to communicate with the operator, the cycle on the same excavator complex is reduced by 8–15%, and the spread between the best and worst operator disappears along with the operator.
- Downtime share. Breaks, shift changes, and waiting for the operator leave the shift: the bottleneck becomes fleet technical availability, not people's schedules.
- Eliminated failure class. Hitting the side and berm, turning error under a bench, contact with auxiliary equipment — incidents related to a decision in the cab. The loop prevents them from occurring, and the class leaves downtime statistics.
- Excavation stability. Routes and dump points are set, so ore blending and grade deviation decrease: ore flow is predictable over a long interval.
- Dispatcher workload. One dispatcher manages the entire fleet instead of three or four machines; they intervene manually only in abnormal scenarios, and the number of interventions per hundred trips falls (ISO 17757 defines what counts as a failure and what is a normal limitation).
We calculate the effects together: a cycle gain is devalued if downtime grows. We take measurements in one area and one fleet, with the same haul distance — otherwise a before/after comparison proves nothing.
Operating modes: from teleoperation to unmanned excavation
Equipment operating modes at a site are rarely the only ones. While one area is being converted to unmanned excavation, the neighboring one continues teleoperation at the mine — there is a complex face, unstable communication, or a requirement for continuous human presence.
Therefore, we design the loop so that the same machine changes mode during a shift, without a separate fleet or a second set of equipment.
| Mode | When it fits | Limitations |
|---|---|---|
| Teleoperation | Complex face, emergency and repair work, areas outside the automation zone | One operator runs one or two machines, quality depends on the communication channel, fatigue grows and pace drops by the end of the shift |
| Semi-autonomous cycle | Routes are marked, but geology or a mixed fleet does not allow removing the human | An operator on shift and stable positioning are required; interventions are frequent in abnormal situations |
| Autonomous cycle in an area (unmanned excavation) | Proven routes with predictable geology, an isolated work zone | Area isolation, priority rules for third-party equipment, an emergency stop procedure, and continuous communication coverage are required |
Mode limitations are set not by the machine's "intelligence level" but by positioning quality, channel coverage, and priority rules on haul roads. The practical selection criterion is how much downtime each mode produces on a specific route.
That is why we assemble a unified dispatch loop: the operator sees the entire fleet in one window, mode changes do not require retraining, and incident analysis is based on an event log, not on shift memories.
Implementation stages: from fleet audit to handover at the site
The order of stages is set by dependencies, not by the calendar: until routes, gradients, and area zones are described, loop configuration is pointless, and without pass statistics, acceptance tests are unprovable. Next is the path from the first run to the point when your operations service manages the solution.
- Fleet and area audit. Condition of hydraulics and electrics, serviceability of actuators, availability of standard control interfaces. In parallel — area layout, communication coverage, loading and dumping points.
- Target loop design. We divide functions between onboard and dispatch levels, define zones, routes, and safe stop boundaries, and the procedure for working together with people and other equipment.
- Retrofitting and installation. Sensors, drives, onboard computer, communication channels; each unit is accepted separately — the machine enters the area verified.
- Debugging at a closed site. Route testing, response to obstacles and loss of communication, emergency scenarios, measurement of trajectory deviation.
- Tests at the working area. First a limited zone, then the general flow; we record cycle time, share of automatic passes, and number of operator interventions.
- Acceptance tests. Run according to an agreed protocol with your service, metric measurement, and a list of comments with closure deadlines.
- Training and handover of the solution to the customer. Control room operators and the operations service receive procedures, documentation, and the escalation procedure.
The result of the staged approach is that the decision is made by metrics, not by impression: you retain test protocols, loop diagrams, and a system operated by your own personnel.
Configurations, diagrams, and procedures: what remains at the site
An autonomous loop rests not on code, but on the set that remains at the site. Without it, editing a route turns into correspondence with the contractor, and the shift works from memory. We hand over everything the area lives on after launch:
- Control loop configuration maps — route parameters, speed profiles, protection trigger distances, binding to the control software version.
- Area diagrams and drawings — road network, loading and dumping points, placement of communication and beacons tied to the geodetic base.
- Access policies and roles — who is allowed to change a route, who only to confirm a shift, who to read telemetry; actions are written to the log.
- Operating procedures — dispatcher behavior on loss of communication, manual intervention procedure, per-shift checks.
- Solution documentation — architecture, external interfaces, applicability boundaries, update and rollback procedure.
- Instructions and checklists for shifts, escalation matrix — to whom and within what time a request goes when a unit fails.
The scope of documents covers internal audit and inspections under Safety Rules for Mining Operations. We provide editable formats: your engineering service makes edits itself, without waiting for us.
Case study: excavator complex at a mine, metrics before and after
What was before launch
The excavator complex at Severny Quarry operated in manual mode: the operator guided the bucket along the face edge, relying on a visual mark, and dumping into the truck body began when the haul truck stood level.
The excavation cycle had a large time spread, some approaches went to leveling the bucket, and unplanned repair downtime accumulated toward the end of the month — the implementation case began with this picture.
What was changed and what was achieved
We installed position sensors and a computing module onboard, trained the model on recordings of real cycles, and left the operator with the role of supervisor: they define the face contour and confirm shift handover, while the rest is handled by automation.
The autonomous excavator complex keeps the bucket trajectory within tolerance and stops before contact with an obstacle, so collisions with the cable and edge left the downtime list.
By before-and-after metrics: average excavation cycle time decreased by about one fifth, the spread between shifts narrowed almost by half, and the share of time the complex actually works grew from two thirds to nine tenths of a shift.
Failures that previously required removing the machine from the face moved into the early warning class: the system signals a trend and requests unit replacement in a planned window.
Scaling at the area took place without stopping production: the second machine was connected to the same infrastructure, the configuration, face maps, and model were transferred, and one operator station now runs both pieces of equipment.
How does teleoperation differ from autonomous mode at a mine?
Teleoperation — the operator drives the machine from the control room, and only actuation automation remains onboard. Autonomous mode is the opposite logic: the route and task live on the machine itself, while the human monitors the fleet and intervenes by exception.
The difference starts with the communication channel. Teleoperation needs a continuous stable stream: when it degrades, the operator loses the picture, and the machine stops.
Autonomous equipment needs communication for telemetry and task assignment — it survives a short break with a safe stop and returns to the cycle.
This also leads to different workload on people: in teleoperation, the operator is assigned to a machine; in autonomous mode, one specialist runs several units.
| Criterion | Teleoperation | Autonomous mode |
|---|---|---|
| Communication channel | continuous, sensitive to delay | telemetry and tasks, tolerant of breaks |
| Operator | one per machine, entire shift at the console | one per several machines, intervention by exception |
| Where appropriate | dangerous faces, clearing, emergency work | proven routes, dumps, haulage |
| Entry barrier | channel and console | mine model, zone marking, precise positioning |
In terms of productivity, teleoperation is limited by the human factor: shifts, fatigue, lighting, image quality. The autonomous cycle is more even, but it depends on area readiness — marked routes and no people in the loop.
Transition conditions are set in advance: radio coverage, positioning quality, mine model currency, weather, route readiness. Therefore, both modes are kept together: autonomous on proven routes, teleoperation as a standard fallback and for dangerous faces.
Technical concerns before launching an area without operators
Engineering concerns before launching an area without operators come down to four topics: loss of communication, safety in the zone, integration with existing dispatch, and handover of control to your shift. Below are short answers to each.
What happens to the machine if communication is lost in the area?
Communication is designed with redundancy: several access points and independent channels, not one. When the channel is lost, the onboard system switches to safe mode — reduces speed and stops in a permitted zone along the predefined route.
The dispatcher sees the moment of the break and the machine's last state, not a disappearance from the map.
Who is responsible for safety if no one is in the cab?
Several independent loops operate: detection of people and equipment, digital zone boundaries, emergency stop via radio channel. Priority is given to the human in the zone — the system brakes even if the route is planned as clear. Failure of any loop leads to a stop, not to continued movement.
How does this integrate with existing dispatch?
Tasks, routes, loading and dumping points, and machine statuses are transmitted through the standard dispatch interfaces — no second console alongside is needed. The dispatcher continues to work in the familiar window and sees autonomous transport as a regular fleet unit.
How does phased launch and handover of control to your team proceed?
First, one machine on a short route with an operator onboard and external observation, then the entire area, then the shift. Expansion is by metrics: share of autonomous cycles, number of interventions per cycle, response time to an event.
In parallel, we hand over procedures, documentation, and training for dispatchers and the maintenance shift, including return to manual control.
Let's discuss the mine's task and come back with a plan and estimate
A ready-made technical specification is not needed for the first conversation.
Input data on the fleet and area is enough — from it we assemble the loop design: where control is located, how communication runs, which onboard systems we retrofit, and which are already installed on the machines.
- Fleet composition: machine types, models and generations, how many units operate simultaneously in a shift.
- What is already automated — dispatch, onboard positioning, operating hours tracking — and in what form data is provided.
- Area for the pilot at one site: route length, loading and dumping points, gradients, and road condition.
- Communication: where coverage is stable, where it disappears, whether there are sites for repeaters and towers.
- Operating mode: number of shifts, shift changes, refueling, repair windows — they define the stop logic of the autonomous loop.
- People and roles: who is in the control room, who takes over the machine after autonomous mode, who is responsible for safety in the area.
- Constraints: procedures for people and equipment working together, access rules, labor safety requirements.
- Success criteria: productivity, downtime, share of time in autonomous mode, number of operator interventions.
This set is enough to come back with an estimate and work plan — with a loop design, stages, and a pilot at one site with measurable criteria. Describe the task: we will discuss it and propose the next step.
