Masters in Computer Science Alumni

Media sharing amongst teenagers is a well-observed process all across the world, our research continues this path, and in particular, our aim is to analyze media sharing and usage of African youth in townships. Specifically the purpose of this research project is to analyze how teenagers make use of public situated displays that store user generated content. From this research we will determine some guidelines to which mobile phone application developers could use to better cater for the low-income teenage market and also develop technology to facilitate media sharing activities.

My research seeks to answer the question of whether anomaly detection techniques can be used to enhance the capabilities of Security Information and Event Management (SIEM) systems. Anomaly detection is a promising approach in the field of network security. My research will evaluate which techniques perform best/worst in terms of performance and effectiveness and will give recommendations in this regard. Testing will be carried out at a financial services provider where the SIEM system “Tivoli Security Operations Manager” is in use. This company will provide test data and is willing to integrate anomaly detection into the SIEM system to perform a field trial.

This project aims to expedite the processing of Astronomical Radio Interferometry UV data to produce multi-frequency data and images useful to Astronomy researchers with the use of GPU cards. A GPGPU implementation of Direct Fourier Transform method of Image Synthesis will be developed, optimized, and tested against standard Fast Fourier Transform and multi-core implementations. Auto-tuning will be implemented if deemed necessary. This project will probably be extended to include other GPU related enquiries, that of 3D Direct Fourier Transforms, Deconvolution methods or both.

The application of High-Level Constraints to Fuzzy Crowds has been shown to be feasible for small crowds but impractical for crowds with more than a hundred agents. However the special effects industry often requires crowds with thousands of agents. Using CUDA and multi-threading techniques we have scaled the simulation vertically achieving speedups of over 30x on a single node. Using MPI we hope to scale the simulation horizontal achieving linear speedup with the number of nodes in the cluster. Image: See attachment.

In Artificial Intelligence (AI) finding the shortest path in a graph is a classic problem that has many efficient solutions. To speed up path finding in known environments, there are many techniques which preprocess the graph. However, none of these techniques have been applied to the Weighted Region Problem. The weighted region problem is finding the shortest path from a source to a destination in a planar subdivision in which each region has a weighting. The aim of my research is to apply different preprocessing techniques to the weighted region problem and evaluate them.

Due to the small number of people in rural areas telephone companies do not find it cost effective to install telephone infrastructure there. Mesh potato (MP) is a solution. A MP is a plain old telephone(POT) connected to a wireless node/access point and allows individuals to make phone calls much like a land line. It is still in its infancy stages and has room for improvement. We seek to improve on it at the routing protocol. The protocol used is called better approach to mobile as-hoc networking or just B.A.T.M.A.N. In this project we are investigating the development of a new quality of service (QOS) routing scheme that will enable the MP to generate a better QoS in this network.

The Lattice Boltzmann Method is used to simulated fluids with moving obstacles. These simulations are implemented on the GPU using CUDA to vastly reduce the required simulation time. The simulation includes the liquid's free-surface, tracked using the Volume of Fluid method. The Lattice Boltzmann Method is a Eulerian (grid-based) method that is linear with respect to time, and requires no global communication, which makes it particularly well-suited to parallel architectures. Moving obstacles are simulated using Bullet Physics which interfaces with the fluid simulation when calculating fluid-obstacle momentum transference.

The aim of this project is to develop a method of procedurally generating spacecraft that would be suitable for use in video games and film special effects. Our work is based on existing shape grammar methods, but we will extend these with features such as voxel-space interpretation and symmetry. It is likely that these extensions will be applicable to other areas of procedural generation, such as the shape grammars used to create buildings. We have also planned a thorough set of tests for evaluating our work, and ensuring it meets the desired requirements.

The process of extracting ore or minerals from the ground remains one of fundamental interest as it generates a lot of resources. However, miners often undergo a lot of risks, and a lot of hazardous events such as blasting often arise. The increasing mortality rate of miners remain an issue of great concern to the miners and the environment at large. Autonomous robots would play a vital role in this area. Hence, the need to provide supportive navigational strategy to aid the preventive safety operations performed by autonomous robots

We calculate triangulated surfaces of binding interfaces in non-redundant protein-protein complexes from the PDB with our software, Piech. Our large data set and fine-grained surface calculations enables a more statistically significant survey than was previously possible. We find tha homo- and heterodimeric complexes have similar, very broad distributions of interface area, with an "average" interface size of around 1,200Å. We identify distinctive interface residue propensity patterns for different classes of proteins, particularly at the core of the interface. Hydrophobic residues are strongly favoured at the core regions of both hetero-and homomers, but not as favoured in immune protein complexes. This work evaluates the largest number of protein-protein interfaces to date.

This research deploys a P2P application in a mobile ad hoc environment. The P2P network will consist of limited-resource mobile phone nodes that form a temporary network anywhere without much infrastructure investment which is typical mobile ad hoc networks (MANETs). The P2P network will use the Freenet steepest-ascent hill-climbing search mechanism to search for and share resources. This search mechanism will be compared to other search mechanisms used in popular P2P networks such as Gnutella. Through system evaluation the searching and routing metrics as well as the upload and download times of nodes in the network will be compared to similar properties in a Gnutella network.

In this project we will endeavor to develop an indigenous knowledge management system for indigenous knowledge for a selected local community, as a proof of concept. As current systems do not support the social-cultural and communication structures of communities, my study focus is to discover and collect ICT design ideas and structures from the community for the purpose of storage and retrieval of indigenous knowledge and map the life style into an appropriate ICT indigenous knowledge retrieval architecture. The retrieval techniques and processes will be assembled from the identified rural community as a result of mapping existing knowledge transfer and communication patterns onto the system. As part of a long-term collaborative research project, the indigenous knowledge retrieval architecture will then become part of the indigenous knowledge management system to be developed.

The collection of heritage data is a time-consuming and expensive process and is often dependent on project funding. However this is not sustainable. Therefore it is necessary to find cost-efficient and time-efficient ways of preserving heritage data. A solution may be to exploit social networks and the way in which people interact. An application on a social network may provide a means to avoid the cost, decrease time and increase scale of operation of heritage preservation by motivating users to supply and process the data. This project aims to use a Facebook application for the purpose of gathering heritage data and useful metadata.

Forests are an important part of many scenes in computer games and simulations. Due to the high complexity involved, various tricks have to be employed in order to facilitate their use in games and simulations. These tricks often have a negative impact on a game's visual appearance. This project deals with techniques to mitigate these effects by using procedural generation and level-of-detail tools. Specifically we will be investigating L-Systems and various tree placement strategies to generate realistic forests within reasonable performance limits.

The aim of this research will be to develop a technique that will allow users to securely store private data on a public cloud such as Amazons S3 and analyse the impact such a system will have on terms of a performance overhead. The system must allow for sharing of data, assign certain rights to other users and revoke them, and it must be able to validate the integrity of the file. Along with being able to store data the concept of Searchable Encryption and its application in this context will also be explored.

In many domains such as infrastructure management, business process monitoring, crisis management and other monitoring activities, systems are characterised by large numbers of sensors collecting data from a variety of information sources. The information is collected in real-time and thus there is an interest for live performance analysis and reporting.

This calls for data mining methods for recognising, predicting, reasoning and controlling performance of systems. In recent years, soft computing methods and algorithms and methods have been applied to data mining, to identify patterns and new insight into data.

Three soft computing techniques were chosen, namely, Artificial Immune Systems, Bayesian Belief Networks and Neural Networks.

In this project, soft computing techniques are applied to different datasets including the Wi-Fi network monitoring data, Kenya drought data and the pollution monitoring data. In this way, we can determine which of the techniques and algorithms work best under which circumstances.

The objective of this project is todevelop interference mitigating techniques that will allow Wireless sensornetworks to communicate effectively and reliably in collocating environmentswhere other sources of wireless devices exist and which are capable ofdisrupting communication in a sensor network. The first step in mitigatinginterference is the ability to detect it. Thus the first phase of this project isdedicated to selecting tools and parameters that can effectively detect thepresence of interference. These are then used in developing algorithms whichcan cognitively adjust the operating behavior of the sensornodes (in order to improve communication performance) based on the interferenceexperienced in the working environment. The last phase of this project involvesrunning experiments to test the effectiveness of the mechanisms utilized to mitigate interference on sensor networks.

Research in digital literacies and mobile technology use in low-income communities of South Africa has revealed the ubiquity of the mobile Internet in such settings. Exceptional adoption rates demonstrate the value that such technology offers by providing affordable communication for many who were previously unconnected. This project will extend our understanding of this behaviour by directly comparing two systems for information search deployed on different platforms: the conventional web and the mobile medium. Both controlled experiments and a longitudinal study will be employed in order to measure usability as defined by ISO 9241-11 and to gather data about actual use.

The main aim of this project is to improve the quality of the communication between isolated networks using a data mule. The data mule is a combination of a vehicle and an electronic device, with the purpose of picking up data from one network when in close range, buffering it and dropping it off in another network wirelessly.

In order to achieve the main aim, we will develop a new algorithm. An integral part of the development includes the use of data mules that prioritizes on efficiency (packet loss and latency) and intelligence (adaptability of communication in changing situations) of data transmission between isolated networks.

Open access and open resources have gained much attention from the world in the last few years. The interest in sharing information freely by the use of the World Wide Web has grown rapidly in many different fields. Now, information is available in many different data forms not only documents because of an evolution in technology. This research focuses on Open Educational Resources (OERs) and how educators supply their resources to the public. The aim is to build a repository that is able to handle the different complex forms of educational resources and a desktop application that eases the processes needed to be taken by a user to share his/her educational resources. The solution proposed is centered on the front-end application which handles the complex objects on a users desktop. The desktop application will handle all the needed processes with minimal interaction with the user to ingest learning objects into the repository.

Interoperability is the capability of heterogeneous systems to communicate and exchange data with each other, using a set of pre-defined formats that will allow the systems to interpret the data exchanged correctly and use each other's services successfully. The current suite of high-level interoperability protocols is made of either simplistic and easy to implement protocols that lack some efficiency or efficient but hard to implement protocols. The implementation of more that one protocol requires the knowledge of different sets of vocabularies and rules.

The ever-increasing number of online digital repositories require an easy to implement but still efficient and standardised protocol framework, that will act as an incentive to interoperability and open access. The aim of this project is to provide an experimental set of protocols developed on top of a standardised framework.

We generate realistic terrains by combining user-sketched curves and real landscape data in a patch-based synthesis to produce terrains with features that reflect user constraint curves and the characteristics of a real terrain. This process involves the manipulation of large landscape data which is extremely time-consuming and so a GPU implementation is used to increase performance.

With the ever-increasing volumes of research documents produced over the years, in the form of electronic thesis and dissertations, the existing information management systems cannot sacle well and hence dynamic scalability of cloud computing can be taken advantage off to offer this service. This research thus, aims to develop techniques for building scalable digital information management systems based on efficient and on-demand use of generic grid-based technologies. In particular, existing resources like the Amazon Elastic Compute Cloud (Amazon EC2), Amazon Simple Storage Service (Amazon S3) and Amazon SimpleDB were used in this study.

This research aims to situate digital storytelling within rural African communities. We design and build a mobile digital storytelling system on top of insights that arose out of ethnography and a design workshop held in a rural village in the Eastern Cape of South Africa. We use this system to further situate digital storytelling by probing how rural African communities interpret digital storytelling and how the practice fits into the oral culture and social fabric of such communities.

Digital Library systems are still immature software, known to be difficult to deploy and use. One of the difficulties in deployment is customisation of the interface, and integration into an existing website. Digital Library systems don't tend to use templating systems to allow web designers to re-skin the system, like popular Content Management Systems do. This project aims to integrate a Digital Library system into a Content Management System, to provide the advantage of templating and more easily integrate with an existing Content Management System-driven site.

The focus of this research is on the rapid creation of motion graphs for use in skeletal character animation. To speed up the graph creation process, we have focussed on the comparrison of the frames of animation that are being used in the graph. Instead of comparing every frame of aniamtion to every other frame of animation, we have divided the space that the skeleton occupies into a regular 3D grid. By grouping the positions of the skeleton's joints into buckets based on the cell they are in, we have found speedups to the comparrison process of up to 125 times.

Draggable tag clouds are a proposed extension to the standard tagging metaphor that exploits the spatial proximity between tags. Tag clouds are static, visually weighted renditions of clusters of terms that interface to the prevalent concepts of collections of information. Draggable tag clouds may act as a search user interface that affords the user the opportunity to rearrange tags within the tag cloud space as a means to approximately express their information need. The closer a tag appears to the cloud’s centre, the greater its significance to the user. To examine the utility of this proposed extension, the draggable tag cloud search user interface was linked to Google News, allowing users to search news content by continuously modifying the tags’ positions.

Currently Presence theory is used to explain the sense of presence a user experiences in a virtual environment; this theory does not take into account many psychological aspects. Flow theory however does and is used to explain social psychological and behavioural aspects. Two questionnaires were selected which would best represent this compatibility in a game: Fallout 3 (by Bethesda Game Studio); for Flow theory: the FSS-2 – General and for Presence theory: the ITC-SOPI. The data from these questionnaires would be analysed for any similarities and differences and contribute to a more well-rounded explanation of an user's experience(s) in a game.

ICT4Dproject aims to aid NGOs working with micro-entrepreneurs in South Africa. In particular, I am working with an NGO called Triple Trust Organisation in an attempt to improve their effectiveness in working with spaza shop owners in Khayelitsha and the surrounding area. Since computers are not common in these areas, cell phones will be used as the technology platform. The project will fit into the research area of small-screen form design, with some emphasis on touch-screen devices." A representative image: is attached. It is not a picture of mine, because I seem to have misplaced the pictures I took.

In the field of task automation, multiple robots are often coordinated to accomplish tasks more efficiently. In certain situations, a distributed algorithm run onboard all robots is best used for task allocation. The aim of this project is to construct a simulation which uses a distributed auction algorithm to allocate path plans to Unmanned Aerial Vehicles (UAVs) in order to survey irregular areas of uneven terrain. Additionally, the simulation should be robust against communication faults and produce solutions which are physically feasible.

The video game industry is maturing. Video game production required powerful and robust tools to ease development. Game engines are specialised middleware that take care of much of the technical details allowing the game creators to concentrate on game creation. A game engine is a valuable tool in game programming education, but they are made for performance and utility, not education. A modular game engine would allow students to quickly make games using built in functionality, but also allows them to alter an extend the current engine.

The notebooks in the Bleek and Lloyd collection contain handwritten stories that metaphorically encode the Bushman culture. These notebooks, however, only exist as scanned images and therefore the stories they contain cannot be searched, indexed or compared. This research seeks to investigate how accurately the Bushman stories can be automatically converted from images to text, through a process known as transcription, and also to explore the various techniques for doing this with a strong focus on the complex diacritics that are used in the Bushman script.