Bio-inspired Robotics (iRat):
Bio-inspired computational paradigms for artificial learning and intelligence are used as the basis for the modelling of the artificial thinking systems. Featuring elements with complex biological behaviours, these paradigms are used to empower autonomous robots with adaptive and reactive cognitive behaviours. In addition, the group has also developed a robot, the iRAT (intelligent Rat Animat Technology), specifically to embody the bio-inspired neural based cognition models.

This integrated approach leads to an increased understanding of the neural, behavioural and information processing bases of complex and intelligent systems. Insight from neurocognitive systems will be used to develop computational models, autonomous robots and intelligent software agents, which in turn will lead to a deeper understanding of the relationship between neurocognitive mechanisms and their behaviour in whole systems.

A key aspect in investigating how information is understood in biological systems is to study the creation, evolution and application of languages. Combining with robotics, we examine the effects of language acquisition and learning among embodied agents, in particular the representation of spatial and temporal concepts. These concepts include where and when events, objects and agents are located in space and time.

Besides inventing their own languages to representing these concepts, the robots also converse with each other in order to better understand and establish concepts relating to their world, allowing for more effective communication between agents. Furthermore, they are able to demonstrate their understanding of these concepts by their ability to perform behavioural and navigational tasks.

Click here to find out more about Lingodroids

Discursis is a new communication analytics technology that allows a user to analyse text based communication data, in the form of conversations, web forums, training scenarios, and many more.

Discursis uses natural language processing algorithms to automatically processes transcribed text to highlight participant interactions around specific topics and over the time-course of the conversation.

Discursis can assist practitioners in understanding the structure, information content, and inter-speaker relationships that are present within input data. Discursis also provides quantitative measures of key metrics, such as topic introduction; topic consistency; and topic novelty.

Click here to find out more about Discursis

Charting Visualization Theory
A theoretical approach to visualization allows us to understand how different visual representations provide insights. The category model of visualization developed by Vickers, Faith and Rossiter (2013, TVCG) is based on semiotics and mathematical category theory. This VFR Category can also be used to show the scope and relationships between existing visualization theories of visualization. An interactive chart of this process can be viewed here.



DVS data sets
For spiking networks to perform computational tasks, benchmark data sets are required for model design, refinement and testing. Classic machine learning benchmark data sets use classi cation as the dominant paradigm, however the temporal characteristics of spiking neural networks mean they are likely to be more useful for problems involving sequence data. To support these paradigms, we provide data sets of 11 moving scenes, each with multiple variations, recorded from a dynamic vision sensor (DVS128), comprising high dimensional (16k pixels) and low latency (15 microsecond) events.

Click here to view the data sets introduced in our paper in ICANN2014

please cite: Gibson, T. A. et al.(2014). Event-Based Visual Data Sets for Prediction Tasks in Spiking Neural Networks. In Artificial Neural Networks and Machine Learning–ICANN 2014 (pp. 635-642). Springer International Publishing.

Machine Learning, Data analysis and Visualization
There is an abundance of data associated with many important problems in science, commerce, our environment and man-made systems. This creates a demand for techniques that can be used to model and understand large and complex datasets. Typical applications include prediction and classification, anomaly detection and support systems to assist in understanding the data. An example current area of application is in the analysis of healthcare systems data in collaboration with healthcare professionals.



Metaheuristic Optimization and Evolutionary Algorithms
Optimization problems (e.g. minimizing cost, maximizing efficiency) are of fundamental and practical importance. Metaheuristics attempts to solve challenging optimization problems with minimal assumptions, incorporating and combining techniques inspired by nature, statistical modeling, artificial intelligence and mathematics. Our research is particularly focused on developing techniques to better understand the relationship between difficult optimization problems and algorithm performance.



'Harlie' stands for Human and Robot Language Interaction Experiment.
It is the name of an Android smartphone application that our research team has recently developed.
Harlie can call you on your smartphone to have a chat with you and ask for voice samples.

We are now calling for participants, please click here for more details