Honours Projects 2009

Projects supervised by Joachim Gudmundsson from NICTA

Algorithms for analysing movement data

Technological advances of location-aware devices, surveillance systems and electronic transaction networks produce more and more opportunities to trace moving individuals. Consequently, an eclectic set of disciplines including geography, sports, database research, animal behaviour research, surveillance, security and transport analysis show an increasing interest in movement patterns of entities in various spaces over various time scales.

This is a new and exciting area where the group at NICTA is one of the world leaders. Research is done in close collaboration with all the researchers, and students are expected to contribute to the positive and international atmosphere in the group.

The most crucial step of all the below projects is the development of efficient algorithms. Since spatio-temporal patterns are very complex compared to patterns that have been considered before, it is unlikely that any approach neglecting the spatial information would be able to generate successful algorithms. The aim is to develop algorithms for a new set of real-world problems using tools from the fields of algorithms, computational geometry and data mining. The projects have a strong mix of theory and programming.

For more information please see www.dmist.net

Project 1: Queries on the trajectory of soccer players
One of the recent application areas is team sports, where players’ positions on the field are tracked with very high accuracy. Currently the use of this data is only used for trivial statistics, e.g., how far did player X run during the game. The aim of this project is to take a first step towards a system that would support advanced queries in large sets of trajectories. For example, if a coach or commentator draws a path, or a set of paths, on a screen then one should report all occurrences of this path among the trajectory data.

Project 2: Detecting group movement patterns
Interesting patterns for moving objects involve some subset of the objects that have the same, or similar, behaviour. Two simple examples are "encounter" (e.g. a large enough group of lions meet in the same region) and "flocking" (e.g. a large enough group of gnus is moving along paths close to each other for a certain time). Also, in some applications repetitive patterns (migration patterns of birds or commuting patterns of people) are of interest, such as: recurrence, concurrent recurrence, regular recurrence. The aim of this project is to develop fast approximation algorithms for these kinds of movement patterns.

Project 3: Recognising manoeuvres
In both in defence and sports applications it is of interest to recognise predefined movement patterns of groups. This would be to detect different movement strategies and evaluate how effective they are, or to monitor how well individuals follow a practised plan. An example would be to detect and categorise group formations and manoeuvres. Again, due to the large data sets it is crucial to develop fast and efficient approximation algorithms to obtain useful software.