2. Underground Haulage
• Underground metal mines.
• Haulage of broken ore
• Unstable roof - Hazard Environment
• Line of sight remote control
• Demand for safety decreases productivity
• Existing Technologies:
• Buried wires, lights on the tunnel roof
• Reflective beacons
• Expense, maintenance, speed
• Vehicle weight: > 30 tons
• Speed: upto 20 km/h
• Payload: ~ 10 tons
• Cycle time: 1-5 minutes
• Operates 24x7
• Human controlled
3. Research Addressing an Industry Need
• Existing Technology
• Manual
• Line-of-Sight Remote
• Tele-Remote
• Automation
• Automation
• Without significant infrastructure
• Use natural features to navigate 1996 Sensor Trials - Lasers
Consortium formed in 1998 to build research prototype in 12 Months for
$AU1.6M that only requires existing mine infrastructure and can be
retrofitted to any commercial underground haulage vehicle.
4. A Solution for Limited Infrastructure
in a Partially Unknown Environment
Local Feature-Based
Opportunistic
Localization
Constrained Active
Contours
Predictive steering
and speed control
Custom Built
Computer Hardware
Distributed Realtime
Middleware
Speed & Heading
Throttle, Gear,
Brake, Steering
Operational
Control System
(Articulated Hydraulic)
Laser
Hints
Tactical Reactive Guidance
(Robust Wall Following)
Mission GUI
Map
Strategic
Topological Navigation
(Nodal Map)
TeleRemote
CoPilot
AutoPilot
5. Topological Navigation with
Opportunistic Localization
Forward 1st gear
Turn left @ 20m
Forward 2nd gear
Turn right @ 30m
Forward 2nd gear
Turn left@ 25m
Forward 3rd gear
Turn Left @ 10m
Centre @ 50m
Stop @ 60m
Route MapMine Map Nodal Map
6. Robust Feature Detection for Localization
• Navigation relies upon distance along route
• Odometry is used as primary sensor
• Odometry drifts / vehicle slip - must be corrected
• Long tunnels have few unique features for localization
• Skeletonization of tunnel provides intersection topology
• Centre of intersection is “invariant” to incoming pose
We only need to know
where we are when we are
at an intersection, which
provides us the information
that tells us where we are!
Opportunistic localization.
7. Robust Wall Following with
Constrained Active Contours
Active contour to generate the recommended vehicle path
Active contour contains nodes that are free to move in 2D
Minimize repulsion/attraction and curvature energy
8. Robust Wall Following with
Constrained Active Contours
Add potential energy to nodes
Unstable due to complex interaction between Ad-Hoc forces
9. Robust Wall Following with
Constrained Active Contours
Constrain motion of nodes to horizontal rails (1D)
Problem - it can’t go around corners
10. Robust Wall Following with
Constrained Active Contours
`
Constrain motion of node to radial rails
Advantage - laser data is in radial coordinates
11. Robust Wall Following with
Constrained Active Contours
`
Re-seed top node into right hand band
12. Local Testing - A Research Necessity
300m shade cloth tunnel with rough ground conditions
13. Field Test at Operating Mine
Transport 1000kms and commission in 2 days
15. From Research to Product
• Hazard and Operational Analysis
• Independent Safety Systems - Isolation
• Product Development
• Integrate into mine communications infrastructure
• Make improvements to user interface
• Commercial Release, April 2003
Minegem User Interface 2003Embedded computer on LHD
17. A Much Needed Technology
for the Mining Industry
• Safety
• Fatality rate is plateauing
• Mines are going deeper
• Production
• Limited manpower availabilty
• Productivity blocked by OH&S
• Automation
• Safety without loss in productivity
• Improved vehicle health
• 24/7 mixed operations
• “Unlimited” growth potential
18. A Project with Impact on the Industry
• Prediction made for the next 8 years (CIE 2004)
• 80 loaders to be automated (currently 15)
• Saving $1 per tonne of ore moved
• Each machine moves 0.5M tonnes per year
• 100% decline in morbidity & 80% decline in injury rates
• Competitive Advantage
• Retrofit to existing fleet, and other OEM fleet.
• Can operate at different levels of autonomy.
• Uses relative rather than absolute navigation (unknown mine)
• Weakness - don’t solve all problems, change the problem.
• Future Impact
• Caterpillar holds 24% of the world’s market
• They sell 800 to 1000 LHD’s per year worldwide
• Expect more than 50% to be automated in the medium term
• Increasing demand for underground mining.
CIE - The Centre for International Economics
MINEGEM is an autonomous navigation system for a class of underground loaders used in the mining industry that aims to improve safety for mine personnel while at the same time achieving an increase in productivity. The underpinning navigation algorithm development was carried out by CSIRO from 1996 to 1999 and consists of a relative navigation technique that uses 2D laser scanners as its primary sensor. After a successful technology transfer period, Caterpillar launched the MINEGEM product in 2004. The system has been sold around the world and is now in everyday use moving millions of tonnes of broken ore while operators sit safely in comfortable control rooms.
A “choke point” for underground metal mines (often referred to as metaliferous mines) is the haulage of the broken ore from the area that it is liberated from the rock mass to the initial crushing plant. In all metaliferous mines in the world the first stage of this process is achieved using underground front-end loaders, or Load Haul Dump (LHD) vehicles. The blasted ore lies in areas (stopes) containing unstable and unsupported roofs. LHDs must be driven into these areas to collect the ore – one bucket load at a time. Initially, drivers sitting in the driving seat drove LHDs into the stopes. However, every now and again, a driver would be unlucky enough to have a roof collapse with perhaps a few hundred tonnes of rock on top of their vehicle. With a changing attitude to the value of life, this practice became unacceptable and so instead the driver would drive the LHD to the opening of the stope, would then dismount and continue the ore collection (bogging) operation using a line-of-sight remote control system. This attempt at improving worker safety was not entirely successful, as operators would accidentally drive their remote controlled LHD into themselves, crushing them against the immovable tunnel wall. Operators would also receive injuries from falling when mounting and dismounting their LHDs.
The next logical step to increase safety was to remove the operators from the LHD’s immediate area altogether and extend the remote control idea to the entire bogging and transport (tramming) phases. In the 1990s, this occurred in the form of tele-operation where an operator could drive the LHD and bog the ore using feedback from video cameras located on the vehicle. This now made the operation safe for the personnel, but had the side effect of reducing productivity, since it was impossible to remotely drive an LHD as fast as a machine driven by an on-board operator. The task had been reduced to a video game with very little feedback to the operator. In addition to the reduction in tramming speeds, LHDs suffered more damage, as they were now far more likely to hit tunnel walls and maintenance costs increased.
The solution to the dual goals of safety and high productivity was automation. It was clear that if LHDs could be automated, then operators may become supervisors; located on the surface, in a clean office – totally out of harms way and the tramming process could become fast. Automated vehicles have the potential to be driven faster than manually driven machines as the time lag between perception and action (when compared to humans) can be significantly reduced. A third goal of a reduction in damage and natural wear and tear on the LHD could also be aimed for as vehicles could be programmed to be driven within the manufacturers design limits.
The innovation from the CSIRO team was the development of an algorithm that could use the output from a 2D laser scanner (SICK LMS) to both steer the LHD and locate (localise) it within the system of mine tunnels. A technique based on constrained active contours (“snakes”) was developed that could determine the correct steering angle for the LHD based on the free space in front of the vehicle. The behaviour of the snake was influenced by a number of steering “hints”, such as “keep left”, “turn right at the next intersection”, etc, that were issued based on the LHD’s rough location in the mine tunnel network. Precise localisation was not required and so the system is considered a relative navigation system. Only a rough location along a particular tunnel is required to load the driving hints and this rough position estimation is achieved using so-called “opportunistic localisation” where the system recognises features in the 2D laser scans at tunnel intersections. The combination of a relative navigation and steering system with the opportunistic localisation results in an infrastructure free navigation solution. The only infrastructure required is a radio communications system (WiFi in this case).
The navigation technique resulted in a patent:
“System for relative vehicle navigation”, Elliot Duff, Jonathan Roberts, Peter Corke, Jock Cunningham (Patent granted in USA, Canada, Australia, New Zealand, Chile, Spain, South Africa).
Over engineered
Sensing and actuation
Access to complete machine control
Competitive solutions
Visual servoing (CSIRO)
Global mapping (ACFR)
Hierarchical
control (layered) - piggy back on top a working systems
navigation (tactical and strategic)
Anthropomorphic
If humans can do it!
Based up rally drivers
Performance
4 years / 1000s hours / 10Kms of tunnel
3 different vehicles (30 to 60 tonnes)
Algorithm does not require modification
Autonomous tramming = human
Still requires Tele-operation of digging
Commercialization (3 years)
Independent safety system
Robust hardware & software
User interface
Productivity trials at ODO and NPM
Hangups / oversize rocks / road maintenance
Mix manual / remote / copilot / auto modes / fleet
Commercial system has been released
A trial of likely navigation sensors was carried out at Mt Isa Mine. Data were collected from a 2D laser scanner, a bearing only laser and retro-reflectors, an IMU, steering and wheel encoders and a Doppler radar. All sensors were mounted on a production LHD and data were collected for a typical tramming route. The output of this trial was a report and a number of publications that indicated the likely success of a system developed using 2D laser scanners.
4 x PLS laser rangefinders
Transmission encoder
Doppler ground speed radar
Engine management system (RPM)
Articulation angle
Brake pressure
6-axis INS
Navigations Sensors
2D laser scan
Bearing only laser
IMU encoders
Doppler radar
Scanning Laser
Test Suitability in environment
Very early adopters
In June 1999, the experimental LHD was transported to Northparkes mine in New South Wales where it was demonstrated operating autonomously within the mine within a day of being delivered. The trip was completed by the LHD driving autonomously out of the mine, along 3km of service tunnels.
Drives at 18km/h
No guidance infrastructure
12 months development time for prototype (to 1999)
Licensed to Caterpillar
Currently production trialling at 2 sites in Australia
Operator acceptance is very good
First commercial field robotic system
Source: MCA, Safety performance report of the Australian minerals industry 2005-2006.
Employment has grown by 66 per cent in the last 5 years
Impact and relevance to Industry
A 2004 report compiled by The Centre for International Economics (CIE) on the economic benefits of robotics mining to the Australian economy concluded that:
• “A conservative estimate is that the technology will save 10 lives over 8 years.”
• “I
He said that worldwide, Caterpillar produces 24 per cent of the market for mining haulage trucks.
“These loaders are designed and built to suit very specific underground mining conditions," he said.
"They have to be able to fit comfortably within tunnels and they must be robust and durable enough to withstand tough conditions such as rock falls.”
Caterpillar’s Underground Mining Loaders range in price from US$500,000 to US$1 million each.
The MINEGEM system has now been installed at a number of mines around the world including mines in Europe, Australia and South East Asia. A total of twelve vehicles are now operating using the MINEGEM system.
• Stawell gold mine in Victoria, Australia reported a 65% productivity improvement of line-of-sight remote control operation (they skipped tele-operation) with a 20-30% reduction in maintenance costs [Personnel Communication with mine management]. These figures were based on the first eleven months of operation. The automated machines now move 50% of all the ore at Stawell and have demonstrated an availability rate of between 85% and 90% [Perspectives on Modern Mining, 2007].
• Olympic Dam mine in South Australia reported an increase in productivity of 40% using two MINEGEM capable LHDs in 2003 [Personnel Communication with mine project manager]. They reported that in one block (area of the mine) the automated system could be operated for 6.7 hours per shift compared to 4.8 hours in manual operation.
At its time of launch, the MINEGEM LHD navigation package was the only commercially available automated vehicle system available for the underground mining industry. Since then, another product has been released that is a direct competitor to MINEGEM. Automine, developed and sold by Sandvik, is based on the same 2D laser scanners but uses an absolute navigation technique rather than the relative navigation technique used by MINEGEM. It is therefore suspected that the Automine system will not be as adaptable and easy to maintain as the relative navigation-based MINEGEM.
http://www.mining-technology.com/features/feature1209/