Integrated Ambient Intelligence and Behaviour Inference for Empathetic Robotic Assistants

As our population’s average age rises, we increasingly face the problem of the cost of caring for the elderly or infirm. Technology can help in simple, mechanistic ways, but the elderly may be wary of allowing more intrusive care eg via robots into their lives. In order to make such technology more acceptable, it has to be able to react to the moods and behaviours of the elderly or disabled person. This project aims to combine Ambient Intelligence and Human Behavioural Recognition to determine the physiological and psychological state of the elderly person, so that a robotic assistant can provide an appropriate and empathetic response. The task will involve gathering and interpreting data from a highly heterogeneous set of sources – passive sensors in the home, IoT devices, cameras for facial mood recognition and standard biometric sensors such as electrodermal activity sensors. The first challenge is the development of a modelling technique that can cope with sources that may be occasionally faulty or absent, have different time periods, reliability and accuracy, or may conflict. The inferencing from these sources will involve ontological or other expressions of the aspects of the environment and person, and a range of computational intelligence techniques to handle temporally and spatially distributed data. The second challenge is in identifying or developing a robust and fast technique for interpreting facial expressions and expressing that temporal information in a form suitable for combining with the other data sources. Potentially Fuzzy Cognitive Maps will be explored as a way of interpreting the Facial Action Units extracted from camera footage. The third challenge lies in developing appropriate responses for the robotic assistant that can support the physical and psychological state of the elder, including both short term behaviour and long term trends. Techniques to evaluate the effect of the response will be necessary, so that the system can adapt to the elderly person’s behavioural changes. This project will require a student with an interest in the investigation of Computational intelligence techniques, and the ability (and confidence) to adapt or combine existing techniques to generate new, more appropriate algorithms to handle complex and real-time data sets.

Specific qualifications/subject areas required of the applicants for this project
(e.g. 2:1 BSc Hons. Chemistry):

UK 1st Class / 2.1 Bachelor’s degree (or UK equivalent according to NARIC) and/or UK Master’s degree with a minimum of a merit (or UK equivalent according to NARIC) in Computer Science, Computer Systems or Mathematics. Good mathematical and programming skills for algorithm development and evaluation are essential.