QLD Mining and Energy Bulletin
Issue link: http://ebook.aprs.com.au/i/31843
RECENT DEVELOPMENT IN TECHNOLOGY AND ITS IMPACT ON MINE AUTOMATION BY ELLIOT DUFF T here is no doubt that automation is the future of mining operations in Australia. Successful development and deployment of automation technology will allow surface and underground mining to be carried out more effi ciently, safely, and with less human intervention than is currently required. However, the performance and nature of this technology will rely upon the developments made in a number of related areas: communications, logistics, human factors, geosensing, and fi eld robotics. Field robotics or robotics for unstructured and dynamic environments such as mining, (the area of expertise of the author) is experiencing unprecedented growth around the world with Australia playing a leading role. Over that last 18 years, CSIRO has been involved in numerous automation projects, investigating, developing and demonstrating the feasibility of automating mining equipment, such as automation of: highwall miners, dragline swing, Load Haul Dump (LHD) vehicle, explosive loading, rope shovel, coal longwall process, tele- robotic rockbreaker, hot metal carrier). Today, CSIRO’s Minerals Down Under Flagship (MDU) is researching the technology challenges that will face the “Future Mine”. To help in this research, the automation of mining equipment can be divided into three tasks: • Navigation – the task of controlling the movement of a vehicle from one point to another (i.e. the task of haulage: haul trucks and LHDs). • Manipulation – the task of handling objects; picking, insertion, modifi cation, or destruction (i.e. the tasks of drilling, explosive loading or rock-breaking). • Excavation – the task of digging, moving or removing material (i.e. the control of draglines, shovels, dozers, front and backhoe loaders – Figure 1). These tasks can be further divided into four processes (Figure 2): 68 QLD Mining and Energy Bulletin Winter 2011 1. Sensing – hardware that provides information about the state of the robot and its environment (e.g. encoders, inertial, GPS, video, Lidar, radar). 2. Perception – software that acquires and processes the sensing data into a model of itself and the environment that the robot can understand. 3. Control – software that commands the actions of the robot (e.g. optimal path planning, obstacle avoidance). 4. Actuation – hardware that produces an action (e.g. electric motors, pneumatic and hydraulic actuators). Traditionally, the greatest challenge in fi eld robotics has been the selection of sensors that will survive in the harsh outdoor environment. In such an environment, it is necessary to install additional sensors (with different modes of operation) to ensure that robots have the ability to detect the failure of their own sensors and actuators (providing redundancy and closed loop control). Given the signifi cant advances made over the last decade in the reliability, sensitivity and cost of these sensors, the focus of research has turned to perception. Perception is an ability that humans take for granted. In the unstructured environment of a mine (as opposed to the structured environment of a robotic factory), perception is far more complex than initially expected. To help clarify the problem, perception can be broken into four components: localisation, mapping, moving object tracking and change detection (Figure 3). Localisation Localisation is the ability to know one’s own pose (i.e. position and orientation). Whilst basic localisation can be performed with conventional Global Navigation Satellite System (GNSS) in an open pit mine, it is often insuffi ciently accurate for autonomous navigation (i.e. it has a precision of less than 20 metres). With RTK (Real Time Kinematic) differential correction this precision can be enhanced to two centimetres, but this cannot be guaranteed if there are insuffi cient satellites visible in the sky or with multipath interference caused by refl ections from pit walls and other large vehicles. These problems can be overcome with two solutions: • Data fusion where an Inertial Measurement Unit (IMU) is able to provide data that can help the GNS reject signals that come from multipath sources and maintain a localisation solution whilst in GPS denied areas (e.g. down in the pit). The accuracy of the solution is dependent on the cost of the IMU (e.g. laser ring gyros can easily exceed $100k). • Pseudo-Satellites - GNSS transmitters that can be installed around the rim of the open-pit. The transmitters can be placed in such a manner as to ensure complete coverage. However, the pseudo-satellites are not permitted to transmit on the conventional GPS bands, thus making the system incompatible with commercial, off-the-shelf GNS receivers. An alternative solution is to install non-GPS localisation infrastructure in the pit. This infrastructure could use radar, lasers, RF tags, or alternative time-of-fl ight localisation techniques – very much the same way that airports and harbours track the motion of aircraft or ships in their fi eld-of-interest. Such technology could also provide localisation underground. Mapping Mapping is the ability to continuously measure the 3D terrain (roads, pit, tunnels, etc.). Mapping of the mine can be conducted with conventional surveying systems (e.g. static 3D laser ranging systems with RTK GPS). However, these systems do not provide the temporal fi delity that is required for autonomous navigation or excavation. Optimal path planning is signifi cantly infl uenced by changes in the terrain at the “end points” (i.e. where the ore is being loaded or dumped). There are a number of solutions to improve the temporal fi delity of the mine map: RESEARCH REPORT