Problems for MISG
The list of problems will be updated as problems are submitted and further information about the present problems and additional problems becomes available. Please keep watching these pages. 网易体育 five problems will be selected depending on the expertise and interests of the participants.
- Problem: Merging of Image Data from Multiple Sensors.
Industry: Anglo America Corporation
Industry Representative: MIchael Sears
Moderator: Steven Damelin
Student Moderator:
Description:
There is now a wide variety of remote sensing instruments available which produce data ranging from satellite radar to airborne hyperspectral. Individual satellites produce multiple data sets which may not be accurately co-registered and where the pixel ground sizes (IFOV) varies as well as the data type. An example is ASTER which produces data from separate spectrometers in the visible, infrared, and thermal regions of the spectrum, leading to incorrectly geographically registered data of varying pixel sizes.
The idea of using high spatial resolution panchromatic data to merge with, and hence to sharpen, lower spatial resolution data is well known as pan sharpening. An example from Quickbird satellite data is given below. The image on the left is the original colour data at the original 2,5 metre resolution. In the image on the right the data has been merged with 0,6 metre panchromatic data to give a much sharper result.
The problem becomes much harder in the situation where the data sets are from different sensors, possibly of different types, and possibly collected at different times. Below is an interesting example showing that it is possible to do this. The data set is a merge between ASTER data (offering good multispectral resolution) and EROS (high spatial resolution) data. Note that here a geometric correction was also required.
The problem proposed for discussion is to investigate the possibilities ? with various types of data ? of:- Automated co-registration;
- Data merging;
- Data representation.
-
Problem: Strategic placement of roadblocks in the fight against crime in a South African city
Industry: Gauteng Department of Community Safety
Industry representative: Lebogang Motlhabane (with colleagues from the School of Computational and Applied Mathematics)
Moderator:
Student Moderator:
Description:
It has proved effective in the fight against crime to concentrate on the search of motor cars. The geographical location of roadblocks is of obvious importance but not of direct interest here. We focus on the local problem of roadblock types and competition between the strategies of the police and the criminal.The problem will be posed in two levels.
Level 1. Traffic flow effects of the two strategies of stop-every-car versus random sampling of cars.
Level 2. To estimate the payoff matrix in a Stackelberg game between a dominant player (the police) and a respondent player (a criminal), based on the location of roadblocks on a representative street map. -
Problem: Optimal mining replacement rate for underground tabular platinum reef deposits
Industry: Platinum mining
Industry representatives: Cuthbert Musingwini and Richard Minnitt
Moderator: Montaz Ali
Student Moderator:
Description:
Underground mining of inclined tabular platinum reef deposits in South Africa initially starts from surface by sinking an inclined access shaft called a decline. The decline is located approximately in the centre of the deposit. From the decline, a network of lateral and inclined tunnels (also called development), is excavated in waste rock immediately below the reef horizon, to gain access to the reef. The reef contains the valuable mineral while waste rock does not. Once the reef is accessed, it is blocked out into mining production areas called stopes. The stopes are then extracted to produce the mineral. The sequence of extracting the stopes starts from the decline and progresses outwards to the reef boundary and also starts from top to bottom. The figure below illustrates this mining sequence.
The mining sequence is sometimes disrupted when available stopes are inadequate to replace producing stopes that become depleted. This can be due to factors such as intersecting disruptive geological structures (weak planes) or encountering logistical problems. Parts of a stope falling within a geological structure cannot be extracted. Mines create a contingency (or operating flexibility) to counter the effect of these potential disruptions by keeping development 12-24 months ahead of producing stopes, in the form of fully prepared stopes that are available, but not producing (i.e. a kind of buffer). This will ensure that there are available stopes to replace producing stopes that become depleted with time. However, keeping development ahead of stoping is costly - money must be spent to excavate and keep the development open while no production of ore is realised from the development effort. It is therefore necessary to find an optimal mining replacement rate that will ensure mines get the benefit of adequate operating flexibility while minimising development costs (i.e. maximise Net Present Value and simultaneously maximise flexibility). Can an optimal mining replacement rate be determined? Could this be an unconstrained optimisation problem or a problem of the nature of Economic Order Quantity (EOQ)?
Two papers related to the subject are:
1. Musingwini C, Minnitt R C A and Woodhall M (2007) ?Technical operating flexibility in the analysis of mine layouts and schedules?, Journal of the Southern African Institute of Mining and Metallurgy, Vol. 107, No.2, pp129-136.
2. Lindon L F, Goforth D, van Wageningen A, Dunn P, Cameron C, Muldowney D (2005), ?A parallel composite genetic algorithm for mine scheduling?, in Proceedings of ASC 2005 ? The 9th IASTED International Conference on Artificial Intelligence & Soft Computing, Sept 12-14, 2005, Benidorm, Spain.
-
Problem: Autonomous robot motion
Industry: Mobile Intelligent Autonomous Systems Group, CSIR, Pretoria
Industry Representatives: Dr Simukai Utete and Professor J R Raol
Moderator: Mapundi Banda
Student Moderator:
Description:
This study group problem considers decision-making strategies for outdoor motion by an intelligent autonomous system. An autonomous vehicle must be capable of planning its motion from a start position to a destination or goal point as well as detecting and avoiding obstacles encountered on its route.For a review of motion planning methods, see Kavraki and LaValle (2008). Some approaches explicitly decompose the motion planning problem into the two constituents of path to goal and local obstacle avoidance ? for example, Janglova (2004).
The problem considered here is that of a robot allowed to move independently in an open environment without human intervention. Some physical characteristics of the robot such as its size and velocity are known. In addition, the terrain that the robot is allowed to traverse is given. The robot must navigate in such a way that it avoids static and dynamic objects. Dynamic objects include traffic on the roads where the robot is allowed to move and people. The robot has to make decisions when it approaches such objects, sometimes in very limited time, since some objects might cross its path unpredictable. It is necessary to determine an efficient way of representing routes traversable by the robot and objects in the path of the robot in order to generate an optimal or sub-optimal route/response for the robot.
The problem can be posed as a decision problem. Using sensor data collected locally in real-time, how best can the robot plan its route to a goal point while responding rapidly to what it encounters in its path.
In summary, interest is in the design of novel methods for rapid-response behaviour appropriate for a robot navigating a real-world landscape with static and dynamic obstacles. Potential applications of such strategies include robot use in work areas where decisions about action on detecting other vehicles or people in performing a task are critical ? for example, in mining environments (Nebot and Baiden 2007; AcuMine).
References:
Kavraki, L. and LaValle, S. Motion Planning. Chapter 5. Springer Handbook of Robotics. Siciliano, B. and Khatib, O. (Eds.) Springer, 2008.
Janglova, D. Neural networks in mobile robot motion. International Journal of Advanced Robotic Systems, Vol. 1, No. 1, pp. 15-22, 2004.
Nebot E. and Baiden G. (Guest Eds.). Editorial, Journal of Field Robotics, Special Issue on Mining Robotics, Vol. 24, No. 10, pp. 801-802, 2007.
AcuMine. http://www.acumine.com -
Problem: Pressure-velocity Coupling
Industry: Pebble Bed Modular Reactor (Pty) Limited (PBMR)
Industry Representative: Onno Ubbink
Moderator:
Student Moderator:Description:
PBMR is a gas cooled pebble bed nuclear reactor. The reactor core contains about 450 000 pebbles. The flow through the reactor core is modelled with a porous medium approach and subsequent bypass flows with a one-dimensional systems approach. Both methodologies boil down to a flow model based on the Navier-Stokes equations where flow resistances are provided by means of experimental correlations. Changes in the flow resistance between the different flow regions are large and discontinuous.
Conventional codes achieve pressure-velocity coupling by means of a staggered-grid approach, which limits the implementation to a structured grid environment. In order to model the complex 3D geometry of the PBMR reactor the use of arbitrary unstructured grids are preferred. The co-located pressure velocity grid approach with a Rhie & Chow interpolation scheme for pressure-velocity coupling is normally used on unstructured grids. Direct implementation of this approach gives a typical spike in the velocity profile. There are some remedies to resolve the problem by means of momentum source averaging or special pressure interpolation. However, none is general or stable enough for our application.We are looking for a general stable methodology that naturally takes care of the pressure-velocity coupling at such domain discontinuities without the introduction of the velocity spikes.
References:
Ferziger, Joel H., Peric, Milovan; Computational Methods for Fluid Dynamics; 3rd rev. ed.; 2002; ISBN: 978-3-540-42074-3
-
Problem: Pebble bed / Reflector Boundary Treatment
Industry: Pebble Bed Modular Reactor (Pty) Limited (PBMR)
Industry Representative: Onno Ubbink
Moderator:
Student Moderator:Description:
PBMR is a gas cooled pebble bed nuclear reactor with a core that contains about 450 000 pebbles surrounded by a solid graphite reflector. Heat removal in the pebble core is by means of a combination of convection, radiation and conduction. Radiation is modelled with an effective conductivity approach. Heat removal in the graphite reflector is by means of conduction.The gas coolant flow in the pebble core is modelled with a porous medium approach that enables a course mesh approach to be used. The numerical modelling can be done with a finite volume approach based on either a vertex centred or a cell centred nodalisation. Depending on the flow conditions the two approaches might give large temperature differences in the solid graphite reflector that reduce with mesh refinement ? the cell centred approach converges to the vertex centred approach with mesh refinement. For reasons of consistency and to be compatible with other numerical methods employed by PBMR, the cell centred nodalisation is preferred. There are remedies for the 1 dimensional case available to let the cell centred approach act like the vertex centred approach, such that both methodologies produce the same results for the same mesh resolution.
We are looking for a general stable methodology in 3 dimensions that naturally takes care of the heat transfer phenomena at the pebble bed / reflector boundary for the cell centred nodalisation on course meshes.
References:
http://www.pbmr.co.za
Ferziger, Joel H., Peric, Milovan; Computational Methods for Fluid Dynamics; 3rd rev. ed.; 2002; ISBN: 978-3-540-42074-3